1

Coping with Congestion Understanding the Role of

Transportation Demand Management Policies on

Commuters

Meeghat Habibian1 Mohammad Kermanshah2

1 Dept of Civil Engineering Sharif University of Technology Tehran Iran phone +989122145196 fax

+982166022716 Email Habibianautacir2 Dept of Civil Engineering Sharif University of Technology Tehran Iran phone +982166164187 fax

+982166022716 Email mkermansharifedu

Abstract

This paper examines the role of transportation demand management (TDM) policies on commuters mode

choice in the city of Tehran The analysis is based on the results of a stated preferences survey developed

through the design of experiments approach Five policies covering increasing parking cost increasing fuel

cost cordon pricing transit time reduction and transit access improvement are assessed in order to study

their impact on commuters consideration of six modes of transportation to travel to work A multinomial logit

model was developed for the 366 commuters who regularly commute to their workplace in the center of the

city In addition to a number of commuting and contextual variables the model shows that the

implementation of single policies and the interaction of multiple policies are significant in affecting

commutersrsquo mode choice The marginal value of policies is presented and three graphs are provided to show

the implementation of the model

Keywords Transportation demand management policy logit model stated preferences marginal

effect main effect interaction

2

1 Introduction

Congestion due to driving a car is a common problem for megalopolis citizens because it imposes

environmental and social costs such as air and noise pollution depletion of energy road casualties

and daily delays Among these outcomes delay on the roads is reported as the most pervasive and

costly problem (de Palma amp Lindsey 2001) Because of limitations on expanding transportation

networks policymakers and transportation planners around the world have attempted to reduce

congestion by proposing transportation demand management (TDM) policies over the past several

decades During the past two decades TDM policies such as increasing the vehicular operating cost

and promoting public transit ridership have been a centerpiece of public policy (Cao amp Mokhtarian

2005)

The necessity of TDM in worldwide megalopolises and the variety with which researchers approach

the subject led to the development of numerous policies and programs to manage travel demand

Litman defines TDM as a general term for strategies that result in more efficient use of

transportation resources (Litman 2010) The generality of the TDM concept and its reliance on

technology for implementation leads to the development of several policies yearly For example

Marshal et al addressed a comprehensive range of 64 travel reduction policies in 1997 (Marshall et

al 1997) and Litman defined a list of 49 TDM policies in 2003 (Litman 2003) with an increase to

more than 80 policies after seven years (Litman 2010)

To manage travel demand a TDM policy should be chosen based on its features A TDM policy has

several features including coerciveness type of behavioral change temporal scale spatial scale

market-based vs regulatory mechanism influence on latent and manifest demand technical

feasibility and costs (Loukopoulos 2005) Among the studies in which researchers categorized the

policies based on the policiesrsquo features (eg (Marshall amp Banister 2000 May et al 2003 Louw amp

Maat 1999)) is a study that uses a coerciveness-based classification (Steg amp Vlek 1997) This

approach which has been also explored in some other studies (eg (Thorpe et al 2000 Stradling et

3

al 2000)) classifies the policies based on their coerciveness towards mode change and also calls

them pull or push policies Pull policies encourage the use of non-car modes by making them

attractive to car users Transit-oriented development street reclaiming and bus rapid transit

development are examples of pull policies Some studies describe these policies as ldquocarrotrdquo policies

(OFallon et al 2004 Golob amp Hensher 2007) Inversely push policies are those that discourage car

usage by making it less attractive Road pricing parking pricing and cordon pricing are a few

examples of push policies Push policies also have been called ldquostickrdquo policies in some studies

(OFallon et al 2004 Golob amp Hensher 2007)

Each city consists of different types of individuals who are affected differently by TDM policies On

the other hand variation of the aforementioned features affects each individual travel pattern in

unique ways Thus in separate studies different policies are determined to be effective policies In

other words it is not possible to detect a policy as a dominant policy in changing the mode of car

users and some studies recommend that a variety of TDM policies should be simultaneously

implemented (Marshall amp Banister 2000 Meyer 1999) In fact implementing more TDM policies

may cover more individual trips and may be more effective Vieira et al found that adopting more

than one TDM policy which they called multi-instrumentality could possibly overcome some of the

identified weaknesses of and eventually enhance the strengths of single implementations of policies

(Vieira et al 2007)

Furthermore a variety of decisions made by public and private organizations affect transportation

systems In fact these small or large decisions which are often made without consideration of their

indirect and long-term outcomes affect the travel pattern of individuals For example urban fringe

communities may offer tax discounts and impose lax environmental standards to attract retail

businesses and industry even though they create more automobile-dependent land-use patterns

(Litman 2010) This issue shows that individuals are often faced with many pull and push motives for

4

or against their car usage To analyze the effects of such decisions on individuals travel patterns one

should be aware of the interactions in addition to the policies

The above discussions show that introducing more than one policy to manage the transportation

demand of a megalopolis is a common problem faced by their policymakers This paper is focused on

modeling the role of multi-TDM policies on commuters mode choice especially in regards to the

interactions of these policies The model provides a number of graphs that enhance transportation

planning for the city of Tehran After describing the research context this paper describes the stated

choice design and the stated preferences survey Then the developed mode choice model is

explored followed by presentation of the graphs The conclusion summarizes the findings and

discusses the implications of the results

2 Research context

Although there are many studies that look at the impact of a single TDM policy on a society such as

studies on congestion pricing (Arentze et al 2004) park and ride (Parkhurst 2000) and parking

pricing (Visser amp Van der Mede 1986 Hensher amp King 2001) only some studies focus on the impact

of multiple policies Among these a few look at the simultaneous implementation of TDM policies

In the context of multiple policies Stradling et al detected motorists who are ready to reduce their

car usage and how they should be helped to change (Stradling et al 2000) In fact individual

reasons for car usage and sensitivity to a number of pull and push policies were detected through a

postal questionnaire survey of English car drivers Those authors verified the difference between pull

and push policies through a factor analysis approach The researchers also found the ranking of the

TDM policies stated by the drivers Another study performed by Mackett focused on pull policies and

personal actions that might attract motorists out of their cars and into transportation alternatives

for short trips (Mackett 2001) He examined various events that could attract car drivers to use an

alternative mode and their associated actions He classified these actions as collective actions

(actions that may be undertaken by the government or other organizations) or non-collective

5

actions (other actions) and assessed the role of policies coerciveness on driver mode change by

assuming that the collective policies are more coercive than the non-collective policies By focusing

on a smaller community and introducing a push policy (fuel pricing) in addition to some pull policies

Kingham et al examined the travel behavior of two companies employees (Kingham et al 2001)

through a survey They studied each employees perception of hisher mode choice during the trip

to work and investigated the potential of transferring car trips to other modes in the presence of

studied policies They also examined the importance of policies that would encourage employees to

use public transit or car-share to travel to work

Although the above studies focused on individualsrsquo perceptions of car-use for daily trips by adopting

some assumptions on the mode choice model of Shiraz Iran Zareii examined the results of

implementing five push policies in the city in terms of total travel time and amount of CO2 emission

(Zareii 2003) Because this study determines car usage by transferring the imposed cost of each

policy to extra time (calculated via the individuals time-value) push policies were of interest

Loukopoulos et al attempted to obtain quantitative estimates of the size of car usage reduction

goals and frequency of implementation of adaptation alternatives (Loukopoulos et al 2004) They

assessed two push policies and a pull policy for different trip aims and examined the cost-

minimization principle in relation to five adaptation alternatives OFallon et al explored the

potential effect of 11 policies on the respondents decision to choose to drive a car to work or school

during the morning peak period in three cities of New Zealand through a stated preferences survey

(OFallon et al 2004) They also reported the marginal effect of each of the studied policies and

recommended a study with fewer policies to explore the possible impacts of combinations of

specific policies Washbrook et al examined the role of seven policies on mode choice (Washbrook

et al 2006) Although the design of this study focused on the policiesrsquo main effects the results were

used to estimate commuter response to various policy combinations of charges and incentives

6

Because the aims of the above studies were to look for the best studied policy by assessing its

impact on car usage they did not deal with the effects of simultaneous implementation of TDM

policies Pendyala et al assessed five TDM policies by adopting an activity-based micro-simulation

model system (AMOS) to simulate changes in individual travel patterns (Pendyala et al 1997) In

their survey they also assessed combinations of specific policies in four transportation control

management scenarios and determined the possible impacts in those scenarios Thorpe et al

presented the individuals attitudinal responses to three push and one pull TDM policies in two case-

study cities in the UK Cambridge and Newcastle (Thorpe et al 2000) They examined the

relationship between the perceived effectiveness and public acceptance of alternative TDM policies

and showed that the public acceptance order of generic TDM policies is improving public transit

road-user charging zone-access controls and increased parking charge This study concluded that

there was evidence of interaction effects between levels of public acceptance of TDM policies when

considered separately and in combination with other policies Further these effects could be

investigated more rigorously with a stated preferences experimental design of alternative TDM

packages which allow the investigation of both main and interaction effects1

Eriksson et al examined the acceptability of one push policy (raised tax on fuel) and two pull policies

(improved public transport and subsidized renewable energy) individually and as packages

combining one push and one pull policy (Eriksson et al 2008) By proposing a model of factors

predicting acceptability of TDM policies they concluded that while the pull policies are perceived to

be effective fair and acceptable the push policy and the packages are perceived to be ineffective

unfair and unacceptable By removing one of the pull policies (ie subsidizing renewable energy)

these authors further assessed the expected car usage reduction in response to other two policies

(Eriksson et al 2010) By focusing on improved public transport raised tax on fuel and their

combination as a package the results showed that the combination was more effective than the

1 In a few studies in choice modeling researchers also examined the second order interactions of attributes in the models(eg (Mogas et al 2006))

7

individual policies Vieira et al explored the concept of multi-instrumentality as a procedure of

policy integration and implementation whereby a systematic search for complementary policies was

sought when planning and designing one (or several) core policy(s) aiming to fulfill one particular

policy more effectively (Vieira et al 2007) They defined criteria to assess the TDM policies and

analyzed four improvement mechanisms in each pair of policies Based on meta-studies they

defined the potential improvement between different types of policies By defining synergy concept

as a benefit of integration May et al reviewed a number of examples to assess the concept and

found little evidence of synergy in outcome indicators (May et al 2006)

Based on the above discussion assessing individual behavioral response to more than one TDM

policy is an interesting issue within the TDM context The following three issues are addressed in this

paper developing a model to investigate the role of TDM policies in commuters mode choice

exploring the role of effective parameters on the consideration of each mode of travel and

suggesting a method to determine the results of implementing two TDM policies simultaneously In

this paper the stated preferences approach is used to model the car users mode choice using the

design of experiments principles

3 Stated preferences

The five policies selected for the city of Tehran consisted of three push and two pull policies The

policies were increasing parking cost increasing fuel cost cordon pricing into an odd-even zone2

transit (bus or subway) time reduction and transit access improvement The latter two were

described by setting measures in favor of the public transit vehicles in streets and intersections

decreasing the time of boarding and alighting at the stations and increasing the number of transit

lines and stops in the city

2 This zone explored in the next section

8

Parking costs fuel costs and public transit time policies are designated with three levels and cordon

price and public access time are designed with two levels Table 1 shows the policies and their levels

All push policies had fixed values for their levels for pull policies because there were variations in

the transit time and transit access time for individuals proportional values of the current state were

used which is different for each individual The term no change in Table 1 refers to the current value

of a policy that each individual already experiences The mean values are also presented in Table 1

for a better description of current state

In preparing a questionnaire for the stated preferences part the design of experiments approach

was adopted Full factorial design is the most general type of design in this approach which

introduces all combinations of all levels in the modeling process In other words full factorial design

produces 108 possible choice sets (33322) This design allows the investigation of all

interactions as well as the main effects in the model On the one hand fewer choice sets are

available when ignoring the effects of higher-order interactions and on the other hand these

interactions have a negligible role in the variance (Louviere et al 2000 Hensher et al 2005) thus

fractional factorial design methods have been proposed

Table 1- Policies and their levels

Measure Type Numberof levels

Description of levels Mean Value

Increasing parking cost Push 3 No change 4000 7000 Rials3 h 71 RialshNACordon pricing Push 2 25000 50000 Rialsday

Increasing fuel cost Push 3 No change 3000 5000 Rialsliter 1470 RialsLiterTransit time reduction Pull 3 No change 15 30 percent shortage 385 minTransit access improvement Pull 2 No change 25 percent shortage 11 min

Efficient design a type of fractional factorial design was used in the study and a design with 895

efficiency was adopted which allows assessing all two-way interactions of policies as well as the

3 10000 Rials are almost equal to 1 US dollar

9

main effects with only 36 choice sets4 (See (Rose amp Bliemer 2009) or (Kuhfeld 2009) for more

details on efficient design) To avoid a time-consuming questionnaire 36 choice sets (scenarios)

were randomly ordered and divided into six separate questionnaire types coded as 1 to 6 Each of

the questionnaires had six scenarios and each scenario consisted of five policies

4 Survey

Two push policies are currently being implemented in the city of Tehran The first is car-free

planning in the CBD area of the city and the second one is an odd-even scheme based on the last

digit of car plates that attempt to enter a zone which is about three times larger than and includes

the CBD area Based on their occupation a few people can drive to the CBD area with a license

called permission A stated preferences survey was assigned for the morning car commuters to the

odd-even zone but they were asked to ignore these two policies to find the accurate sensitivity of

individuals to the study policies The odd-even zone is selected as study area for the two following

reasons 1) because of odd-even control respondents are familiar with the fringes and they can

better imagine the entrance pricing area and 2) respondents are familiar with the limits that they

face half of the week and are thus aware of the alternative existing modes Compared to the CBD

area this zone covers more car commuters and the entrance restriction is more imaginable for this

zone than the former one Respondents were interviewed face-to-face in their workplaces midway

through the year 2009 The interviews were enhanced with a special card to better define the

scenarios

For this study 2196 scenario observations from 366 individuals were adopted The sample included

308 men (ie 841) and 58 women (ie 159) The figures are close to the employment

percentages in the city according to the Iranian Center of Statistics (ICS) This source indicates that

825 of Tehran employees are men and 175 are women (Iranian Center of Statistics (ICS) 2009)

Because this study focuses on car-using commuters comparisons between the sample and city data

4 Efficient design is also adopted in other studies such as managed lanes (Burris amp Patil 2009)

10

especially regarding educational distribution were impossible Table 2 presents demographics of the

sample

Table 2- Demographics gender marital household (HH) size employee type age HH employee(s)

Amount Percent

Gender Male 308 841Female 58 159

Marital Single 100 273Married 266 727

HH Size 1 4 112 86 2353 129 3524 90 2465 42 1156+ 15 41

Age 18~29 122 33330~39 146 39940~49 58 15950~59 32 8760+ 8 22

HH employee(s) 1 156 4262 159 4343 41 1124+ 10 27

The first part of the questionnaire is dedicated to gathering the occupation state home and job

locations the distance between these locations round-trip time (from home to workplace and then

workplace to home) and all car trip characteristics in the previous day or the day before it based on

plate number It was necessary that the respondents drive hisher car in the day studied to complete

the trip diary portion of the questionnaire5 The general reasons for car usage and the scenarios

formed the next portion In each scenario every respondent was asked the question How would

you travel to the workplace if all of these changes were in place on the day studied For example

one may have to pay 4000 Rialsh for parking 50000 Rials per entrance to the cordon the same

amount in transit access and fuel cost and a 15 percent decrease in transit time simultaneously

Depending on individual responses six main options were distinguished6 These choices were still

5 In designing the questionnaire the general form of questionnaire which has mentioned in OFallons study was adopted6 In the pre-test survey 14 modes is distinguished

11

drive a car (C) walk to the station and catch public transit (WampR) drive to a public station and catch

public transit (DampR) ride a motorcycle (MC) catch a taxi7 (T) and catch a taxi by phone (T_T) DampR is

somewhat different than the more familiar ldquoPark amp Riderdquo In fact in the fringes of the odd-even

zone there were no specialized parking lots dedicated to this purpose and commuters considered

Drive amp Ride because they were not allowed to pass the fringes

After each scenario if the respondent changed hisher mode the reason(s) for the change were

asked It could be a sole policy or a bundle of them Furthermore travel-related information was

sought These data were not part of the stated choice but they might have important influences on

individual choices These data consisted of car dependency (need to drive someone or move freight

in the trip) parking place type and average weekly parking costs car and motorcycle ownership and

number of household driving licenses

Depending on the individuals activity in that day three types of activity patterns were detected

Pattern 1 described individuals who had no stop in their commute Pattern 2 was for individuals who

had at least one stop on their way to or from work and pattern 3 was for the individuals who went

to another workplace in their daily activities

Finally for the sake of data generalization and the examination of household characteristics gender

age and household type employment status and education level were also asked

5 Mode choice model

In order to detect the policies that affect individual mode choice the logit modeling approach was

adopted In this model one can determine if the interaction of two policies affects the mode choice

In the calibration step 152 variables were defined and their effects on consideration of each mode

were examined

7 Taxis in Iran are somewhat different than taxis in other countries of the world In fact taxis in Iran are not hiring by oneperson or a group of people at a time Taxis allow passengers to board or alight along their path with respect to theircapacity In other word this mode is functioning similar to transit vehicles but the stops are not predefined

12

51 Model structure

Initially a multinomial logit (MNL) model is developed (Figure 1a) By selecting a number of tree

structures based on recognizing differences in the variances associated with unobserved influences

we find that the greatest similarity in variance profiles is associated with public transport modes as

opposed to non-public modes (Figure 1b) This structure has two nests one including Car (C) and

Motorcycle (MC) as private modes and the other including Walk and Ride (WampR) Drive and Ride

(DampR) Taxi (T) and Tel-taxi (T_T) as non-private modes The result of this nested logit (NL) model is

shown in Table 3

Although it is not a statistically significant improvement overall on the MNL model the statistically

significant inclusive value8 (IV) of 0889 for non-public modes relative to the fixed parameter value of

10 for public modes suggests that there is a structural advantage in selecting the NL specification

The normal test of a statistically significant difference between NL and MNL is an IV parameter

relative to 10 calculated using a Wald-test via equation 1

)1(Wald-test = (IVparameter ndash 1)std error

a The MNL structure

b Final nested structure

Figure 1- Model structure

8 Also called scale parameter

Alternatives

MCCar T_TT WampR DampR

Alternatives

MCCar T_TT

Public

WampR DampR

Private

13

We have (0889-1)2508 =-075 which would be rejected at the usual acceptable significance levels

This suggests that the NL model could be collapsed into an MNL form

Table 3- Nested logit (NL) model resultValueParameter

0889IV (Private)1000IV (nPrivate)

-2668335L( )-4057684L(0)

0342sup2

After the calibration process the variables that were statistically significant were identified and are

presented in Table 4 Table 5 presents the final model of the study with a goodness of fit of 031 for

the six studied modes For a general review of the model calibration results the effective factors can

be grouped under the following three categories TDM policy characteristics commuting trip

characteristics and household socio-economic characteristics which are all treated as alternative-

specific variables

52 Model results

Car (C)

It is expected that push policies impel car-drivers to choose other modes Table 5 shows that cordon

pricing and increase in parking cost cause individuals to choose not to use their car This is in line

with other studies suggesting that these policies are effective to discourage car usage (Hensher amp

Rose 2007 OFallon et al 2004) In addition the interaction between the policies of fuel cost

increase and increase in parking cost shows similar car usage discourage effect Because fuel cost is

related to the distance between home and work locations and parking cost is related to work time

the time that an individual spends out of the home is negatively affected by hisher likelihood to use

a car

14

Table 4 - Definition of the significant variables

AbbreviationVariableTransportation demand management measures

Measures

ParkingParking cost increase Rials per hour

CordonCordon price Rials per entranceAccessTransit access time shortage percent

Interaction of push measures

ParkampFuelParking cost and fuel cost simultaneous effectsCordonampFuelCordon pricing and fuel cost simultaneous effects

Interaction of pull measures

PT_timeampaccessPT time reduction and access improvement simultaneous effectsCommuting trip characteristics

Trip distanceDistance between home and workplaceTrip timeTravel time between home and workplace

Exp FuelLikelihood of unsubsidized fuel use (self-reported on a Likert scale)NtripsNumber of daily tripsPattern2Commuting with 1+ stop(s) in go or return

Pattern3Commuting with 2 workplacesFirst trip timeStart time of first trip

PnocarwkLikelihood of going to work in absence of that car (self-reported)PTnwaccNon-walk access to transit (yes=1)First NaccoNumber of passengers in first trip

PassengerAny passenger on that day (yes=1)Park_paymentParking payment in last weekNhempfullNumber of full employees in HH

CardependencyBoardalight a passenger or move freight in the trip (yes=1)D car ownBe the owner of the used vehicle (yes=1)

Car accCar accessibility in household (number of cars to number of HH driving licenses ratio)NmotorcycleNumber of motorcycles owned by HHD home placeHome Location is in study area (yes=1)

PermissionPermission to enter to study area (yes=1)ComfortI use my car because it is comfortablePoor_PTI use my car because transit is not good

HH socio-economic characteristics

FemaleGender (Female=1)Age lt30Age younger than 30 (yes=1)Age 30_39Age between 30 to 39 (yes=1)Job_durationNumber of years that individual has been at hisher job

Emp_fullFull-time employee (yes=1)Edu BSDegree of education is BSc (yes=1)Edu BS+Degree of education is higher than BSc(yes=1)

D childlt=18Child younger than 18 in HH (yes=1)

15

Table 5 ndash The mode choice model

Tel-Taxi(T_T)

Motorcycle(MC)

Drive amp Ride(DampR)

Taxi (T)Walk amp Ride(WampR)

Car (C)Mode

Variable-471756-37067-147911Constant

Transportation demand management measure variables00019-00045Cordon

-000072Parking-004308Access

-28443D-05Parkampfuel-32475D-06Cordonampfuel

00029Pt_timeampaccess

Commuting trip characteristics-04709Trip distance

-02163-00831Trip time-96755163655Exp fuel-16253Ntrips

-114779Pattern2-71008Pattern3

00282-00270First trip time-02439-01549Pnocarwk

-11322992883-32765PTnwacc-133701First Nacco

-7778-73782Accompany-00049000010Park_payment

201646195554Nhempfull-160144ComfortCar1

-206142DependencyCar1-16101883385-121224DependencyCar2

42176Poor_PTCar1-24988Poor_PTCar2

- -27221D car own70960-39136Car acc

1 -71112-156123Nmotorcycle-1436322762D home place

2 78826Permission

HH socio-economic characteristics149490Female

297584-24548Agelt30-136490Age30_39

079430366303585Job_duration-108743Emp_full-203468-64900Edu BS

10932856687-4499984445Edu BS+102271D childlt=18

-2677366L( )-3849556L(0)0305sup2

112127178592580607N

Note = Positive significance at 1 5 10 level

As expected individuals with higher income are more likely to use their car This is indicated in the

model by the positive signs of individuals who use fuel with fixed (unsubsidized) cost and individuals

16

who pay more in parking charges in the previous week of study Negative sign of Pnocarwk variable

shows that the commuters who stated that their commute depends on car availability are more

likely to use their car Individuals in households with more full-time employees are more likely to use

their car which may be the result of higher household income Not surprisingly commuters who

have permission are more likely to maintain car usage Among the household socio-economic

parameters greater job experience (Job_duration) and higher graduate levels (EduBS+) increase the

probability of car usage

Public transit accessed by walking (WampR)

Access time to transit negatively impacts WampR choice which is expected This result is similar to

findings for the city of Sydney (Hensher amp Rose 2007) The negative coefficient of first trip time

indicates that individuals are more likely to use WampR in the early morning This result seems to

reflect the better weather for walking and faster speed of WampR mode early in the morning

Obviously individuals who are not able to access transit stations via walking (PTnwacc) are less likely

to consider this mode Furthermore serving passengers on daily trips is also a deterrent to using

WampR

Initially assessing the individuals who stated that their car usage is due to poor public transit service

(Poor_PT) led to an unexpected result in favor of considering WampR By introducing to this variable

the number of household cars as a proxy for household income (Poor_PTCar1) the model shows

that of the previously mentioned individuals those who have lower income are the ones who have

to consider WampR The result is understandable as these individuals may have no alternative when

they have to change their mode (they also are not likely to consider other modes) Individuals with

higher levels of income who have to use their car during before or after work (Dependencycar1+)

are not likely to use WampR

The greater the number of motorcycles in a household the less likely commuters is to consider

WampR There appears to be a competition between motorcycle and PT for access to the city center

17

Better PT services in the center of the city in terms of coverage and frequency increases the

likelihood that its residents will consider WampR This is verified by the positive sign of the

D_home_place variable Commuters with greater job experience (Job_duration) in their workplace

are more likely to use this mode Although individuals with higher levels of education are not likely

to use WampR as education level increases avoidance of WampR decreases

Taxi (T)

Table 5 shows that none of the studied policies are significant in considering taxi usage It seems that

taxi usage considering its function in Iran as a non-private and non-public mode of transport is not

affected by pull or push policies A negative sign for taxi travel time indicates that individuals are not

likely to use this mode for longer trips This seems reasonable given that longer trips are more

expensive Commuters who are more likely to use fuel with no subsidy are not likely to use taxis As

mentioned before they prefer to use their car A higher number of trips in a day are also a deterrent

to considering taxi usage which may be due to increased cost for more trips Results show that an

individual with more daily trips avoids using taxis Commuters who are employed in more than one

workplace (Pattern 3) are not likely to use taxis This may be due to the fact that they have a lower

level of income which forces them to dedicate more time on the job

Initial results showed that individuals who stated that their car usage is due to poor public transit

service (Poor_PT) are not likely to use taxis This result was far from our expectations By introducing

to this variable the number of household cars as a proxy for household income the model shows

that the previously mentioned individuals who have higher income (Poor_PTCar1+) are the ones

who are not likely to consider taxis Furthermore because such individuals are not considering any

other modes they may treat taxi usage as a kind of PT mode with poor service

As expected greater access to cars in a household (Car_acc) lessens the likelihood of considering

taxis as an alternative Furthermore individuals in households with more motorcycle ownership are

less likely to consider taxis It seems like there is a competition among motorcycles and taxis for

18

access to the city center Younger commuters are less likely to use taxis and individuals with at least

master degrees do consider this mode in addition to their car

Public transit accessed by Drive (DampR)

This mode is affected by the simultaneous interaction of transit time and transit access

(PT_TimeampAccess) which is reflected in the fact that individuals prefer to use this mode for longer

trips Comparing this mode and WampR the first trip start time affects the consideration of this mode

differently Later morning commuters prefer to use their car to access PT modes Such commuters

may have higher income levels or managerial jobs Obviously individuals who are not able to access

PT stations by walking (PTnwacc) are likely to use DampR Serving passengers in daily trips is also a

deterrent in considering this mode which is similar to WampR but with a lower coefficient

Commuters with higher income levels who depend on their car during before or after work

(Dependencycar1+) are likely to use DampR Individuals who use their own car are less likely to use

this mode which is unexpected As a city center develops better PT network coverage and residents

have smaller distances to their workplaces they are unlikely to use DampR This is proven in the model

by a negative sign for D_home_place

Motorcycle (MC)

Increasing fuel cost and cordon pricing simultaneously discourage motorcycle usages Although fuel

cost is expected to reduce motorcycle usage to some extent its combined effect with cordon pricing

also reduces motorcycle usage However this variable is not as strong as other policy variables

=10)

Of the studied modes motorcycle usage is affected by the most commuting variables This may be

due to the fact that this mode is not common Commuting distance has a negative effect on

motorcycle usage which is expected It is worth noting that trip distance appears only in this mode

which may be a reflection of the role of distance in regards to the safety risk in considering this

19

mode Commuters with more stops to serve passengers while commuting (Pattern 2) are not likely

to use this mode which may be due to the poor passenger service of this mode

Individuals who state that commuting is independent of the mode (Pnocarwk) are not likely to use

MC By looking at the (First_Nacco) negative sign this could stem from the fact that the more

passengers there are on the first trip the less likely individuals are to consider MC Regarding the

low capacity of MC and its safety concerns such commuters avoid using this mode Commuters who

pay more parking charges (Park_payment) are less likely to use MC which is expected Individuals

who are dependent on their car during before or after their work time are not likely to use MC

even if they have lower levels of income (DependencyCar1) Individuals who use their own car

(D_car_own) are less likely to use this mode As expected individuals who live in households with

more motorcycle ownership are more likely to use this mode The positive sign of (Permission)

indicates that commuters who have permission to enter the study area do consider MC Because

such commuters generally provide that permission just for car usage this result is unexpected

As with commute variables of all the studied modes MC is affected by the greatest number of

socio-economic variables As expected young commuters (Agelt30) are more likely to use this mode

Commuters with Bachelor of Science degree are less likely to use this mode among others Full time

employees (Emp_full) are less likely to consider MC whereas commuters with more experience in

their jobs prefer to use it Results show that individuals who live in a household with children

younger than 18 are more likely to consider using a car

Tel-Taxi (T_T)

Results show that cordon pricing causes higher probability of using T_T In fact individuals who use

T_T as a mode with similar level of service as cars9 are more willing to pay the cost and make use of

the mode It is worth noting that the effect of cordon pricing in pushing commuters from car usage

9 As this mode does not have driving stress and parking search time in some cases it may have more amount of utility thana car does

20

(000045) is greater than its effect on pulling them to Tel-taxi (000019) This is because of the

possibility of considering other non-car modes

Because consideration of this mode is a function of its operation travel time (Trip_time) appears as

a deterrent in this mode utility function Table 5 shows that individuals are more sensitive to the trip

time when using T_T mode versus taxi which is expected due to their relative costs

The greater the number of full time employees in a family (Nhempfull) the higher the probability of

considering T_T by its commuters which may be due to the higher income level of these

households This is verified by the greater likelihood of using T_T rather than taxis by such

commuters Individuals with higher levels of income who depend on their car during before or after

work time are less likely to use T_T Commuters with lower income levels who state that they use

their car for the sake of comfort (Comfortcar1) are less likely to use T_T which may be due to its

cost Although such individuals do not consider any other modes they specifically avoid T_T Greater

access to cars in a household leads to greater likelihood of T_T usage which could be due to the

higher income level of a household As mentioned before such individuals even avoid taxis

Females who drive to their workplace are more likely to use T_T It seems like this part of society

considers this mode when desiring to avoid the difficulties of driving Younger commuters are less

likely to use T_T and individuals between 30 and 39 years of age are specifically avoiding this mode

Results show that university graduated commuters are more likely to use this mode

6 Marginal effects

To explore the effects of each policy on mode choice and to answer the second issue raised at the

beginning of this paper the marginal effects approach can be adopted Although the coefficients of

the models utility functions show the drivers behavior when facing one or more policies the

marginal effects of policies or their interactions may appropriately show the results of their

implementation More specifically the marginal effect for this study is interpreted as the change in

21

probability given a unit change in a variable ceteris paribus In this section the variable is defined as

a specific policy or a specific interaction of two policies Table 6 presents the marginal effects of the

studied policies and their interactions with mode choice The results are shown in the form of trip

percentages transferred away from the car to the studied modes and the probability-weighted

sample enumeration approach is adopted to find the values It is worth noting that this table is fully

compatible with Table 5 but the marginal effects that were less significant than 90 percent have

been removed

Table 6 - Marginal effects of policies (percent)

Tel-Taxi(TT)

Motorcycle(MC)

Drive ampRide(DampR)

Taxi (T)Walk ampRide(WampR)

Car (C)Mode

Variable-000088Cordon-000140Parking

-09069Access-0000001ParkampFuel

00040PT_TimeampAccess

Table 5 shows that a 1 Rial increase in cordon pricing causes a 000088 percent decrease in car

usage Furthermore a 1 Rial per hour increase in parking cost decreases the car usage probability to

00014 percent By assuming 8 hours for the average parking duration the daily marginal value of

parking cost converts to 000018 percent These values show that cordon pricing is more effective in

forcing individuals not to use their car than increasing parking cost with the same value Results also

show that a 1 Rial increase in parking cost and fuel cost simultaneously decreases the probability of

choosing car usage by 0000001 percent Table 5 also shows that a 1 minute decrease in transit

access time would result in a 09 percent increase in probability of choosing this mode It also shows

that each 1 minute decrease in transit travel time and transit access leads to a 0004 decrease in the

probability of choosing the DampR mode

22

The above results show that just one of the policy interactions is not significantly different from

none (ie CordonampFuel) and push policies have a significant role in impelling car users to consider

other modes

7 Planning

To demonstrate the application of the model some of its planning aspects for the city of Tehran are

presented here In fact the model is able to assess the effect of all studied variables which were

grouped into three tiers on commuters mode choice In this section we assess the effect of TDM

policies by adopting their pair-wise combinations for the city Table 6 shows that pull policies in this

study did not impose a major effect on car usage Because the average access time of the sample

was 11 minutes improving it by 50 percent (about 5 minutes) increases the probability of WampR to

509069100 which is equal to a 048 percent increase in considering this mode (and decrease in all

other modes) On the other hand decreasing transit time about 50 percent (about 19 minutes based

on sample average) and similarly improving access time decreases the probability of considering

DampR to 0004195100 which is equal to a 038 percent decrease in considering this mode (and

increase in all other modes) Therefore in this section we focused on the push policies which

appear in the utility function of car mode It is worth noting that in assessing each combination of

policies all other variables were assumed to be fixed as the current state

71 Parking pricing and cordon pricing

The average hourly parking charge of individuals in the sample is 71 Rials This charge is 3000 Rials

per hour in the curb lane on some streets about 2000 Rials per hour in parking lots and free in

alleys and other streets The average amount shows that most of the commuters benefited from

free parking spaces in alleys and non-controlled streets Figure 2 depicts the result of implementing

this policy and the cordon pricing policy simultaneously In line with marginal effects it can be seen

that in implementing each policy separately cordon pricing is more effective than parking pricing in

the planning range It is worth noting that commuters in this study are less sensitive to parking

23

pricing than cordon pricing One reason may be that they thought in any case they would find free

parking space Figure 2 shows that the effect of each policy depends on the level of the other policy

In other words the effect of the simultaneous implementation of two policies is not equal to the

sum of their separate individual effects

Figure 2- Effect of implementing parking pricing and cordon pricing

72 Parking pricing and increasing fuel cost

Figure 3 shows the simultaneous effects of parking pricing and increasing fuel cost policies Although

both of these policies are not effective separately their simultaneous effect is significant It can be

seen that their synergy appears in the higher levels of both policies

Figure 3 - Effect of implementing parking pricing and increasing fuel cost

24

73 Cordon pricing and increasing fuel cost

The effect of implementing cordon pricing and increasing fuel cost is shown in Figure 4 It can be

seen that these two policies have no interaction effect in decreasing car usage This result is verified

by this interactions insignificant marginal value in Table 6

Figure 4- Effect of implementing cordon pricing and increasing fuel cost

8 Conclusions

This study examined the role of TDM policies in individual mode choice for work trips in the city of

Tehran Five policies including increasing parking cost increasing fuel cost cordon pricing in the odd-

even zone of the city transit time reduction and transit access improvement were investigated Of

the five policies the former three were push policies and the latter two were pull policies The

design of experiments approach was used to design the questionnaire that would capture the stated

preferences of car commuters

A number of nested structures based on recognizing differences in the variances associated

with unobserved influences were examined and finally the MNL as a superior model for this study

was developed

25

The results show that for car users who faced TDM policies all of the five studied TDM

policies were factors in the mode choice process The model shows that in considering car usage

push policies play a main role and pull policies do not

Results also show that the interactions of policies are also significant The model shows that

the interaction of parking cost policy and fuel cost policy is significant in prompting car users to

consider other modes It also shows that the interaction of cordon pricing and increase in fuel cost is

effective to discourage motorcycle usage Furthermore the interaction of transit time reduction

policy and access improvements policy is significant in considering DampR as a transportation mode to

workplace

Pull policies in the study were expected to attract individuals to transit modes (pull policies

of this study are related to transit modes) This was verified by the model

Results show that although taxi usage is usually treated as an alternative for urban trips it

was not directly affected by the studied policies This may be due to its special function in Tehran as

a non-public and non-private mode

The results also show that except for the interaction of cordon pricing and increase in fuel

cost other interaction effects have significant marginal effects on mode choice

Results of the model show that to implement a single policy cordon pricing is the most

effective in decreasing car usage

This study assessed workplace commuters Thus future studies should explore the

individuals mode choice with other trip aims Additionally including more policies especially pull

policies in favor of non-transit modes is reasonable to rigorously assess the model

It can be seen that the studied modes are affected differently by the TDM policies and their

interactions The impact of the interaction of policies appeared in the consideration of three out of

six studied modes and also in two out of three developed graphs This issue needs more attention

within the transportation demand management context and further studies should be performed

26

Design of this study is based on the consideration of all two-way interactions with some

degrees of deficiency in D-efficient policy Focusing on the significant interactions of this study is

recommended to perform other studies with more D-efficient values

Further understanding of the impacts of TDM policies might be provided in future research

by focusing on the amount of their synergy in prohibiting car usage

9 References

Arentze TA Hofman F amp Timmermans HJP (2004) Predicting multi-faceted activityndashtravel adjustmentstrategies in response to possible congestion pricing scenarios using an Internet-based stated adaptationexperiment Transport Policy 11 p31ndash41

Burris M amp Patil S (2009) Estimating the Benefits of Managed Lanes Technical Report Texas TexasTransportation Institute University Transportation Center for Mobility

Cao X amp Mokhtarian PL (2005) How do individuals adapt their personal travel A conceptual exploration ofthe consideation of travel-related strategies Transport Policy 12(3) pp199-206

de Palma A amp Lindsey R (2001) Transportation Supply and congestion In International Encyclopedia of theSocial and Behavioral Sciences 1st ed Elsevier p15882ndash15888

Eriksson L Garvill J amp Nordlund AM (2008) Acceptability of single and combined transport policy measuresThe importance of environmental and policy specific beliefs Transportation Research Part A 42 p1117ndash1128

Eriksson L Nordlund AM amp Garvill J (2010) Expected car use reduction in response to structural traveldemand management measures Transportation Research Part F 13 p329ndash342

Golob TF amp Hensher DA (2007) The Trip chaining Activity of Sydney Residents A Cross-Section Assessmentby Age Group with a Focus on Seniors Journal of Transport Geography 15 p298ndash312

Hensher DA amp King J (2001) Parking demand and responsiveness to supply pricing and location in theSydney central business district Transportation Research A 35 pp177-96

Hensher DA amp Rose JM (2007) Development of commuter and non-commuter mode choice models for theassessment of new public transport infrastructure projects A case study Transportation Research Part A41 p428ndash443

Hensher DA Rose JM amp Greene WH (2005) Applied Choice Analysis A Primer New York CambridgeUniversity Press

Iranian Center of Statistics (ICS) (2009) Information of Iranian States [Online] Available at HYPERLINK httpwwwamarorgirUploadModulesContentsasset16tehrantehparthtmlhttpwwwamarorgirUploadModulesContentsasset16tehrantehparthtml [Accessed 13 November2009]

Kingham S Dickinson J amp Copsey S (2001) Travelling to work Will people move out of their cars TransportPolicy 8 pp151-60

Kuhfeld WF (2009) Marketing Research Methods in SAS Experimental Design Choice Conjoint andGraphical Techniques SAS 92 Edition NC USA SAS Institute Inc Cary Available on the Webwwwsupportsascomtechsuptnotetnote_stathtmlmarket

Litman T (2003) The online TDM Encyclopedia mobility management information gateway Transport Policy10 pp245-49

Litman T (2010) Online TDM Encyclopedia [Online] Available at HYPERLINK httpwwwvtpiorghttpwwwvtpiorg [Accessed 21 Augest 2010]

3

al 2000)) classifies the policies based on their coerciveness towards mode change and also calls

them pull or push policies Pull policies encourage the use of non-car modes by making them

attractive to car users Transit-oriented development street reclaiming and bus rapid transit

development are examples of pull policies Some studies describe these policies as ldquocarrotrdquo policies

(OFallon et al 2004 Golob amp Hensher 2007) Inversely push policies are those that discourage car

usage by making it less attractive Road pricing parking pricing and cordon pricing are a few

examples of push policies Push policies also have been called ldquostickrdquo policies in some studies

(OFallon et al 2004 Golob amp Hensher 2007)

Each city consists of different types of individuals who are affected differently by TDM policies On

the other hand variation of the aforementioned features affects each individual travel pattern in

unique ways Thus in separate studies different policies are determined to be effective policies In

other words it is not possible to detect a policy as a dominant policy in changing the mode of car

users and some studies recommend that a variety of TDM policies should be simultaneously

implemented (Marshall amp Banister 2000 Meyer 1999) In fact implementing more TDM policies

may cover more individual trips and may be more effective Vieira et al found that adopting more

than one TDM policy which they called multi-instrumentality could possibly overcome some of the

identified weaknesses of and eventually enhance the strengths of single implementations of policies

(Vieira et al 2007)

Furthermore a variety of decisions made by public and private organizations affect transportation

systems In fact these small or large decisions which are often made without consideration of their

indirect and long-term outcomes affect the travel pattern of individuals For example urban fringe

communities may offer tax discounts and impose lax environmental standards to attract retail

businesses and industry even though they create more automobile-dependent land-use patterns

(Litman 2010) This issue shows that individuals are often faced with many pull and push motives for

4

or against their car usage To analyze the effects of such decisions on individuals travel patterns one

should be aware of the interactions in addition to the policies

The above discussions show that introducing more than one policy to manage the transportation

demand of a megalopolis is a common problem faced by their policymakers This paper is focused on

modeling the role of multi-TDM policies on commuters mode choice especially in regards to the

interactions of these policies The model provides a number of graphs that enhance transportation

planning for the city of Tehran After describing the research context this paper describes the stated

choice design and the stated preferences survey Then the developed mode choice model is

explored followed by presentation of the graphs The conclusion summarizes the findings and

discusses the implications of the results

2 Research context

Although there are many studies that look at the impact of a single TDM policy on a society such as

studies on congestion pricing (Arentze et al 2004) park and ride (Parkhurst 2000) and parking

pricing (Visser amp Van der Mede 1986 Hensher amp King 2001) only some studies focus on the impact

of multiple policies Among these a few look at the simultaneous implementation of TDM policies

In the context of multiple policies Stradling et al detected motorists who are ready to reduce their

car usage and how they should be helped to change (Stradling et al 2000) In fact individual

reasons for car usage and sensitivity to a number of pull and push policies were detected through a

postal questionnaire survey of English car drivers Those authors verified the difference between pull

and push policies through a factor analysis approach The researchers also found the ranking of the

TDM policies stated by the drivers Another study performed by Mackett focused on pull policies and

personal actions that might attract motorists out of their cars and into transportation alternatives

for short trips (Mackett 2001) He examined various events that could attract car drivers to use an

alternative mode and their associated actions He classified these actions as collective actions

(actions that may be undertaken by the government or other organizations) or non-collective

5

actions (other actions) and assessed the role of policies coerciveness on driver mode change by

assuming that the collective policies are more coercive than the non-collective policies By focusing

on a smaller community and introducing a push policy (fuel pricing) in addition to some pull policies

Kingham et al examined the travel behavior of two companies employees (Kingham et al 2001)

through a survey They studied each employees perception of hisher mode choice during the trip

to work and investigated the potential of transferring car trips to other modes in the presence of

studied policies They also examined the importance of policies that would encourage employees to

use public transit or car-share to travel to work

Although the above studies focused on individualsrsquo perceptions of car-use for daily trips by adopting

some assumptions on the mode choice model of Shiraz Iran Zareii examined the results of

implementing five push policies in the city in terms of total travel time and amount of CO2 emission

(Zareii 2003) Because this study determines car usage by transferring the imposed cost of each

policy to extra time (calculated via the individuals time-value) push policies were of interest

Loukopoulos et al attempted to obtain quantitative estimates of the size of car usage reduction

goals and frequency of implementation of adaptation alternatives (Loukopoulos et al 2004) They

assessed two push policies and a pull policy for different trip aims and examined the cost-

minimization principle in relation to five adaptation alternatives OFallon et al explored the

potential effect of 11 policies on the respondents decision to choose to drive a car to work or school

during the morning peak period in three cities of New Zealand through a stated preferences survey

(OFallon et al 2004) They also reported the marginal effect of each of the studied policies and

recommended a study with fewer policies to explore the possible impacts of combinations of

specific policies Washbrook et al examined the role of seven policies on mode choice (Washbrook

et al 2006) Although the design of this study focused on the policiesrsquo main effects the results were

used to estimate commuter response to various policy combinations of charges and incentives

6

Because the aims of the above studies were to look for the best studied policy by assessing its

impact on car usage they did not deal with the effects of simultaneous implementation of TDM

policies Pendyala et al assessed five TDM policies by adopting an activity-based micro-simulation

model system (AMOS) to simulate changes in individual travel patterns (Pendyala et al 1997) In

their survey they also assessed combinations of specific policies in four transportation control

management scenarios and determined the possible impacts in those scenarios Thorpe et al

presented the individuals attitudinal responses to three push and one pull TDM policies in two case-

study cities in the UK Cambridge and Newcastle (Thorpe et al 2000) They examined the

relationship between the perceived effectiveness and public acceptance of alternative TDM policies

and showed that the public acceptance order of generic TDM policies is improving public transit

road-user charging zone-access controls and increased parking charge This study concluded that

there was evidence of interaction effects between levels of public acceptance of TDM policies when

considered separately and in combination with other policies Further these effects could be

investigated more rigorously with a stated preferences experimental design of alternative TDM

packages which allow the investigation of both main and interaction effects1

Eriksson et al examined the acceptability of one push policy (raised tax on fuel) and two pull policies

(improved public transport and subsidized renewable energy) individually and as packages

combining one push and one pull policy (Eriksson et al 2008) By proposing a model of factors

predicting acceptability of TDM policies they concluded that while the pull policies are perceived to

be effective fair and acceptable the push policy and the packages are perceived to be ineffective

unfair and unacceptable By removing one of the pull policies (ie subsidizing renewable energy)

these authors further assessed the expected car usage reduction in response to other two policies

(Eriksson et al 2010) By focusing on improved public transport raised tax on fuel and their

combination as a package the results showed that the combination was more effective than the

1 In a few studies in choice modeling researchers also examined the second order interactions of attributes in the models(eg (Mogas et al 2006))

7

individual policies Vieira et al explored the concept of multi-instrumentality as a procedure of

policy integration and implementation whereby a systematic search for complementary policies was

sought when planning and designing one (or several) core policy(s) aiming to fulfill one particular

policy more effectively (Vieira et al 2007) They defined criteria to assess the TDM policies and

analyzed four improvement mechanisms in each pair of policies Based on meta-studies they

defined the potential improvement between different types of policies By defining synergy concept

as a benefit of integration May et al reviewed a number of examples to assess the concept and

found little evidence of synergy in outcome indicators (May et al 2006)

Based on the above discussion assessing individual behavioral response to more than one TDM

policy is an interesting issue within the TDM context The following three issues are addressed in this

paper developing a model to investigate the role of TDM policies in commuters mode choice

exploring the role of effective parameters on the consideration of each mode of travel and

suggesting a method to determine the results of implementing two TDM policies simultaneously In

this paper the stated preferences approach is used to model the car users mode choice using the

design of experiments principles

3 Stated preferences

The five policies selected for the city of Tehran consisted of three push and two pull policies The

policies were increasing parking cost increasing fuel cost cordon pricing into an odd-even zone2

transit (bus or subway) time reduction and transit access improvement The latter two were

described by setting measures in favor of the public transit vehicles in streets and intersections

decreasing the time of boarding and alighting at the stations and increasing the number of transit

lines and stops in the city

2 This zone explored in the next section

8

Parking costs fuel costs and public transit time policies are designated with three levels and cordon

price and public access time are designed with two levels Table 1 shows the policies and their levels

All push policies had fixed values for their levels for pull policies because there were variations in

the transit time and transit access time for individuals proportional values of the current state were

used which is different for each individual The term no change in Table 1 refers to the current value

of a policy that each individual already experiences The mean values are also presented in Table 1

for a better description of current state

In preparing a questionnaire for the stated preferences part the design of experiments approach

was adopted Full factorial design is the most general type of design in this approach which

introduces all combinations of all levels in the modeling process In other words full factorial design

produces 108 possible choice sets (33322) This design allows the investigation of all

interactions as well as the main effects in the model On the one hand fewer choice sets are

available when ignoring the effects of higher-order interactions and on the other hand these

interactions have a negligible role in the variance (Louviere et al 2000 Hensher et al 2005) thus

fractional factorial design methods have been proposed

Table 1- Policies and their levels

Measure Type Numberof levels

Description of levels Mean Value

Increasing parking cost Push 3 No change 4000 7000 Rials3 h 71 RialshNACordon pricing Push 2 25000 50000 Rialsday

Increasing fuel cost Push 3 No change 3000 5000 Rialsliter 1470 RialsLiterTransit time reduction Pull 3 No change 15 30 percent shortage 385 minTransit access improvement Pull 2 No change 25 percent shortage 11 min

Efficient design a type of fractional factorial design was used in the study and a design with 895

efficiency was adopted which allows assessing all two-way interactions of policies as well as the

3 10000 Rials are almost equal to 1 US dollar

9

main effects with only 36 choice sets4 (See (Rose amp Bliemer 2009) or (Kuhfeld 2009) for more

details on efficient design) To avoid a time-consuming questionnaire 36 choice sets (scenarios)

were randomly ordered and divided into six separate questionnaire types coded as 1 to 6 Each of

the questionnaires had six scenarios and each scenario consisted of five policies

4 Survey

Two push policies are currently being implemented in the city of Tehran The first is car-free

planning in the CBD area of the city and the second one is an odd-even scheme based on the last

digit of car plates that attempt to enter a zone which is about three times larger than and includes

the CBD area Based on their occupation a few people can drive to the CBD area with a license

called permission A stated preferences survey was assigned for the morning car commuters to the

odd-even zone but they were asked to ignore these two policies to find the accurate sensitivity of

individuals to the study policies The odd-even zone is selected as study area for the two following

reasons 1) because of odd-even control respondents are familiar with the fringes and they can

better imagine the entrance pricing area and 2) respondents are familiar with the limits that they

face half of the week and are thus aware of the alternative existing modes Compared to the CBD

area this zone covers more car commuters and the entrance restriction is more imaginable for this

zone than the former one Respondents were interviewed face-to-face in their workplaces midway

through the year 2009 The interviews were enhanced with a special card to better define the

scenarios

For this study 2196 scenario observations from 366 individuals were adopted The sample included

308 men (ie 841) and 58 women (ie 159) The figures are close to the employment

percentages in the city according to the Iranian Center of Statistics (ICS) This source indicates that

825 of Tehran employees are men and 175 are women (Iranian Center of Statistics (ICS) 2009)

Because this study focuses on car-using commuters comparisons between the sample and city data

4 Efficient design is also adopted in other studies such as managed lanes (Burris amp Patil 2009)

10

especially regarding educational distribution were impossible Table 2 presents demographics of the

sample

Table 2- Demographics gender marital household (HH) size employee type age HH employee(s)

Amount Percent

Gender Male 308 841Female 58 159

Marital Single 100 273Married 266 727

HH Size 1 4 112 86 2353 129 3524 90 2465 42 1156+ 15 41

Age 18~29 122 33330~39 146 39940~49 58 15950~59 32 8760+ 8 22

HH employee(s) 1 156 4262 159 4343 41 1124+ 10 27

The first part of the questionnaire is dedicated to gathering the occupation state home and job

locations the distance between these locations round-trip time (from home to workplace and then

workplace to home) and all car trip characteristics in the previous day or the day before it based on

plate number It was necessary that the respondents drive hisher car in the day studied to complete

the trip diary portion of the questionnaire5 The general reasons for car usage and the scenarios

formed the next portion In each scenario every respondent was asked the question How would

you travel to the workplace if all of these changes were in place on the day studied For example

one may have to pay 4000 Rialsh for parking 50000 Rials per entrance to the cordon the same

amount in transit access and fuel cost and a 15 percent decrease in transit time simultaneously

Depending on individual responses six main options were distinguished6 These choices were still

5 In designing the questionnaire the general form of questionnaire which has mentioned in OFallons study was adopted6 In the pre-test survey 14 modes is distinguished

11

drive a car (C) walk to the station and catch public transit (WampR) drive to a public station and catch

public transit (DampR) ride a motorcycle (MC) catch a taxi7 (T) and catch a taxi by phone (T_T) DampR is

somewhat different than the more familiar ldquoPark amp Riderdquo In fact in the fringes of the odd-even

zone there were no specialized parking lots dedicated to this purpose and commuters considered

Drive amp Ride because they were not allowed to pass the fringes

After each scenario if the respondent changed hisher mode the reason(s) for the change were

asked It could be a sole policy or a bundle of them Furthermore travel-related information was

sought These data were not part of the stated choice but they might have important influences on

individual choices These data consisted of car dependency (need to drive someone or move freight

in the trip) parking place type and average weekly parking costs car and motorcycle ownership and

number of household driving licenses

Depending on the individuals activity in that day three types of activity patterns were detected

Pattern 1 described individuals who had no stop in their commute Pattern 2 was for individuals who

had at least one stop on their way to or from work and pattern 3 was for the individuals who went

to another workplace in their daily activities

Finally for the sake of data generalization and the examination of household characteristics gender

age and household type employment status and education level were also asked

5 Mode choice model

In order to detect the policies that affect individual mode choice the logit modeling approach was

adopted In this model one can determine if the interaction of two policies affects the mode choice

In the calibration step 152 variables were defined and their effects on consideration of each mode

were examined

7 Taxis in Iran are somewhat different than taxis in other countries of the world In fact taxis in Iran are not hiring by oneperson or a group of people at a time Taxis allow passengers to board or alight along their path with respect to theircapacity In other word this mode is functioning similar to transit vehicles but the stops are not predefined

12

51 Model structure

Initially a multinomial logit (MNL) model is developed (Figure 1a) By selecting a number of tree

structures based on recognizing differences in the variances associated with unobserved influences

we find that the greatest similarity in variance profiles is associated with public transport modes as

opposed to non-public modes (Figure 1b) This structure has two nests one including Car (C) and

Motorcycle (MC) as private modes and the other including Walk and Ride (WampR) Drive and Ride

(DampR) Taxi (T) and Tel-taxi (T_T) as non-private modes The result of this nested logit (NL) model is

shown in Table 3

Although it is not a statistically significant improvement overall on the MNL model the statistically

significant inclusive value8 (IV) of 0889 for non-public modes relative to the fixed parameter value of

10 for public modes suggests that there is a structural advantage in selecting the NL specification

The normal test of a statistically significant difference between NL and MNL is an IV parameter

relative to 10 calculated using a Wald-test via equation 1

)1(Wald-test = (IVparameter ndash 1)std error

a The MNL structure

b Final nested structure

Figure 1- Model structure

8 Also called scale parameter

Alternatives

MCCar T_TT WampR DampR

Alternatives

MCCar T_TT

Public

WampR DampR

Private

13

We have (0889-1)2508 =-075 which would be rejected at the usual acceptable significance levels

This suggests that the NL model could be collapsed into an MNL form

Table 3- Nested logit (NL) model resultValueParameter

0889IV (Private)1000IV (nPrivate)

-2668335L( )-4057684L(0)

0342sup2

After the calibration process the variables that were statistically significant were identified and are

presented in Table 4 Table 5 presents the final model of the study with a goodness of fit of 031 for

the six studied modes For a general review of the model calibration results the effective factors can

be grouped under the following three categories TDM policy characteristics commuting trip

characteristics and household socio-economic characteristics which are all treated as alternative-

specific variables

52 Model results

Car (C)

It is expected that push policies impel car-drivers to choose other modes Table 5 shows that cordon

pricing and increase in parking cost cause individuals to choose not to use their car This is in line

with other studies suggesting that these policies are effective to discourage car usage (Hensher amp

Rose 2007 OFallon et al 2004) In addition the interaction between the policies of fuel cost

increase and increase in parking cost shows similar car usage discourage effect Because fuel cost is

related to the distance between home and work locations and parking cost is related to work time

the time that an individual spends out of the home is negatively affected by hisher likelihood to use

a car

14

Table 4 - Definition of the significant variables

AbbreviationVariableTransportation demand management measures

Measures

ParkingParking cost increase Rials per hour

CordonCordon price Rials per entranceAccessTransit access time shortage percent

Interaction of push measures

ParkampFuelParking cost and fuel cost simultaneous effectsCordonampFuelCordon pricing and fuel cost simultaneous effects

Interaction of pull measures

PT_timeampaccessPT time reduction and access improvement simultaneous effectsCommuting trip characteristics

Trip distanceDistance between home and workplaceTrip timeTravel time between home and workplace

Exp FuelLikelihood of unsubsidized fuel use (self-reported on a Likert scale)NtripsNumber of daily tripsPattern2Commuting with 1+ stop(s) in go or return

Pattern3Commuting with 2 workplacesFirst trip timeStart time of first trip

PnocarwkLikelihood of going to work in absence of that car (self-reported)PTnwaccNon-walk access to transit (yes=1)First NaccoNumber of passengers in first trip

PassengerAny passenger on that day (yes=1)Park_paymentParking payment in last weekNhempfullNumber of full employees in HH

CardependencyBoardalight a passenger or move freight in the trip (yes=1)D car ownBe the owner of the used vehicle (yes=1)

Car accCar accessibility in household (number of cars to number of HH driving licenses ratio)NmotorcycleNumber of motorcycles owned by HHD home placeHome Location is in study area (yes=1)

PermissionPermission to enter to study area (yes=1)ComfortI use my car because it is comfortablePoor_PTI use my car because transit is not good

HH socio-economic characteristics

FemaleGender (Female=1)Age lt30Age younger than 30 (yes=1)Age 30_39Age between 30 to 39 (yes=1)Job_durationNumber of years that individual has been at hisher job

Emp_fullFull-time employee (yes=1)Edu BSDegree of education is BSc (yes=1)Edu BS+Degree of education is higher than BSc(yes=1)

D childlt=18Child younger than 18 in HH (yes=1)

15

Table 5 ndash The mode choice model

Tel-Taxi(T_T)

Motorcycle(MC)

Drive amp Ride(DampR)

Taxi (T)Walk amp Ride(WampR)

Car (C)Mode

Variable-471756-37067-147911Constant

Transportation demand management measure variables00019-00045Cordon

-000072Parking-004308Access

-28443D-05Parkampfuel-32475D-06Cordonampfuel

00029Pt_timeampaccess

Commuting trip characteristics-04709Trip distance

-02163-00831Trip time-96755163655Exp fuel-16253Ntrips

-114779Pattern2-71008Pattern3

00282-00270First trip time-02439-01549Pnocarwk

-11322992883-32765PTnwacc-133701First Nacco

-7778-73782Accompany-00049000010Park_payment

201646195554Nhempfull-160144ComfortCar1

-206142DependencyCar1-16101883385-121224DependencyCar2

42176Poor_PTCar1-24988Poor_PTCar2

- -27221D car own70960-39136Car acc

1 -71112-156123Nmotorcycle-1436322762D home place

2 78826Permission

HH socio-economic characteristics149490Female

297584-24548Agelt30-136490Age30_39

079430366303585Job_duration-108743Emp_full-203468-64900Edu BS

10932856687-4499984445Edu BS+102271D childlt=18

-2677366L( )-3849556L(0)0305sup2

112127178592580607N

Note = Positive significance at 1 5 10 level

As expected individuals with higher income are more likely to use their car This is indicated in the

model by the positive signs of individuals who use fuel with fixed (unsubsidized) cost and individuals

16

who pay more in parking charges in the previous week of study Negative sign of Pnocarwk variable

shows that the commuters who stated that their commute depends on car availability are more

likely to use their car Individuals in households with more full-time employees are more likely to use

their car which may be the result of higher household income Not surprisingly commuters who

have permission are more likely to maintain car usage Among the household socio-economic

parameters greater job experience (Job_duration) and higher graduate levels (EduBS+) increase the

probability of car usage

Public transit accessed by walking (WampR)

Access time to transit negatively impacts WampR choice which is expected This result is similar to

findings for the city of Sydney (Hensher amp Rose 2007) The negative coefficient of first trip time

indicates that individuals are more likely to use WampR in the early morning This result seems to

reflect the better weather for walking and faster speed of WampR mode early in the morning

Obviously individuals who are not able to access transit stations via walking (PTnwacc) are less likely

to consider this mode Furthermore serving passengers on daily trips is also a deterrent to using

WampR

Initially assessing the individuals who stated that their car usage is due to poor public transit service

(Poor_PT) led to an unexpected result in favor of considering WampR By introducing to this variable

the number of household cars as a proxy for household income (Poor_PTCar1) the model shows

that of the previously mentioned individuals those who have lower income are the ones who have

to consider WampR The result is understandable as these individuals may have no alternative when

they have to change their mode (they also are not likely to consider other modes) Individuals with

higher levels of income who have to use their car during before or after work (Dependencycar1+)

are not likely to use WampR

The greater the number of motorcycles in a household the less likely commuters is to consider

WampR There appears to be a competition between motorcycle and PT for access to the city center

17

Better PT services in the center of the city in terms of coverage and frequency increases the

likelihood that its residents will consider WampR This is verified by the positive sign of the

D_home_place variable Commuters with greater job experience (Job_duration) in their workplace

are more likely to use this mode Although individuals with higher levels of education are not likely

to use WampR as education level increases avoidance of WampR decreases

Taxi (T)

Table 5 shows that none of the studied policies are significant in considering taxi usage It seems that

taxi usage considering its function in Iran as a non-private and non-public mode of transport is not

affected by pull or push policies A negative sign for taxi travel time indicates that individuals are not

likely to use this mode for longer trips This seems reasonable given that longer trips are more

expensive Commuters who are more likely to use fuel with no subsidy are not likely to use taxis As

mentioned before they prefer to use their car A higher number of trips in a day are also a deterrent

to considering taxi usage which may be due to increased cost for more trips Results show that an

individual with more daily trips avoids using taxis Commuters who are employed in more than one

workplace (Pattern 3) are not likely to use taxis This may be due to the fact that they have a lower

level of income which forces them to dedicate more time on the job

Initial results showed that individuals who stated that their car usage is due to poor public transit

service (Poor_PT) are not likely to use taxis This result was far from our expectations By introducing

to this variable the number of household cars as a proxy for household income the model shows

that the previously mentioned individuals who have higher income (Poor_PTCar1+) are the ones

who are not likely to consider taxis Furthermore because such individuals are not considering any

other modes they may treat taxi usage as a kind of PT mode with poor service

As expected greater access to cars in a household (Car_acc) lessens the likelihood of considering

taxis as an alternative Furthermore individuals in households with more motorcycle ownership are

less likely to consider taxis It seems like there is a competition among motorcycles and taxis for

18

access to the city center Younger commuters are less likely to use taxis and individuals with at least

master degrees do consider this mode in addition to their car

Public transit accessed by Drive (DampR)

This mode is affected by the simultaneous interaction of transit time and transit access

(PT_TimeampAccess) which is reflected in the fact that individuals prefer to use this mode for longer

trips Comparing this mode and WampR the first trip start time affects the consideration of this mode

differently Later morning commuters prefer to use their car to access PT modes Such commuters

may have higher income levels or managerial jobs Obviously individuals who are not able to access

PT stations by walking (PTnwacc) are likely to use DampR Serving passengers in daily trips is also a

deterrent in considering this mode which is similar to WampR but with a lower coefficient

Commuters with higher income levels who depend on their car during before or after work

(Dependencycar1+) are likely to use DampR Individuals who use their own car are less likely to use

this mode which is unexpected As a city center develops better PT network coverage and residents

have smaller distances to their workplaces they are unlikely to use DampR This is proven in the model

by a negative sign for D_home_place

Motorcycle (MC)

Increasing fuel cost and cordon pricing simultaneously discourage motorcycle usages Although fuel

cost is expected to reduce motorcycle usage to some extent its combined effect with cordon pricing

also reduces motorcycle usage However this variable is not as strong as other policy variables

=10)

Of the studied modes motorcycle usage is affected by the most commuting variables This may be

due to the fact that this mode is not common Commuting distance has a negative effect on

motorcycle usage which is expected It is worth noting that trip distance appears only in this mode

which may be a reflection of the role of distance in regards to the safety risk in considering this

19

mode Commuters with more stops to serve passengers while commuting (Pattern 2) are not likely

to use this mode which may be due to the poor passenger service of this mode

Individuals who state that commuting is independent of the mode (Pnocarwk) are not likely to use

MC By looking at the (First_Nacco) negative sign this could stem from the fact that the more

passengers there are on the first trip the less likely individuals are to consider MC Regarding the

low capacity of MC and its safety concerns such commuters avoid using this mode Commuters who

pay more parking charges (Park_payment) are less likely to use MC which is expected Individuals

who are dependent on their car during before or after their work time are not likely to use MC

even if they have lower levels of income (DependencyCar1) Individuals who use their own car

(D_car_own) are less likely to use this mode As expected individuals who live in households with

more motorcycle ownership are more likely to use this mode The positive sign of (Permission)

indicates that commuters who have permission to enter the study area do consider MC Because

such commuters generally provide that permission just for car usage this result is unexpected

As with commute variables of all the studied modes MC is affected by the greatest number of

socio-economic variables As expected young commuters (Agelt30) are more likely to use this mode

Commuters with Bachelor of Science degree are less likely to use this mode among others Full time

employees (Emp_full) are less likely to consider MC whereas commuters with more experience in

their jobs prefer to use it Results show that individuals who live in a household with children

younger than 18 are more likely to consider using a car

Tel-Taxi (T_T)

Results show that cordon pricing causes higher probability of using T_T In fact individuals who use

T_T as a mode with similar level of service as cars9 are more willing to pay the cost and make use of

the mode It is worth noting that the effect of cordon pricing in pushing commuters from car usage

9 As this mode does not have driving stress and parking search time in some cases it may have more amount of utility thana car does

20

(000045) is greater than its effect on pulling them to Tel-taxi (000019) This is because of the

possibility of considering other non-car modes

Because consideration of this mode is a function of its operation travel time (Trip_time) appears as

a deterrent in this mode utility function Table 5 shows that individuals are more sensitive to the trip

time when using T_T mode versus taxi which is expected due to their relative costs

The greater the number of full time employees in a family (Nhempfull) the higher the probability of

considering T_T by its commuters which may be due to the higher income level of these

households This is verified by the greater likelihood of using T_T rather than taxis by such

commuters Individuals with higher levels of income who depend on their car during before or after

work time are less likely to use T_T Commuters with lower income levels who state that they use

their car for the sake of comfort (Comfortcar1) are less likely to use T_T which may be due to its

cost Although such individuals do not consider any other modes they specifically avoid T_T Greater

access to cars in a household leads to greater likelihood of T_T usage which could be due to the

higher income level of a household As mentioned before such individuals even avoid taxis

Females who drive to their workplace are more likely to use T_T It seems like this part of society

considers this mode when desiring to avoid the difficulties of driving Younger commuters are less

likely to use T_T and individuals between 30 and 39 years of age are specifically avoiding this mode

Results show that university graduated commuters are more likely to use this mode

6 Marginal effects

To explore the effects of each policy on mode choice and to answer the second issue raised at the

beginning of this paper the marginal effects approach can be adopted Although the coefficients of

the models utility functions show the drivers behavior when facing one or more policies the

marginal effects of policies or their interactions may appropriately show the results of their

implementation More specifically the marginal effect for this study is interpreted as the change in

21

probability given a unit change in a variable ceteris paribus In this section the variable is defined as

a specific policy or a specific interaction of two policies Table 6 presents the marginal effects of the

studied policies and their interactions with mode choice The results are shown in the form of trip

percentages transferred away from the car to the studied modes and the probability-weighted

sample enumeration approach is adopted to find the values It is worth noting that this table is fully

compatible with Table 5 but the marginal effects that were less significant than 90 percent have

been removed

Table 6 - Marginal effects of policies (percent)

Tel-Taxi(TT)

Motorcycle(MC)

Drive ampRide(DampR)

Taxi (T)Walk ampRide(WampR)

Car (C)Mode

Variable-000088Cordon-000140Parking

-09069Access-0000001ParkampFuel

00040PT_TimeampAccess

Table 5 shows that a 1 Rial increase in cordon pricing causes a 000088 percent decrease in car

usage Furthermore a 1 Rial per hour increase in parking cost decreases the car usage probability to

00014 percent By assuming 8 hours for the average parking duration the daily marginal value of

parking cost converts to 000018 percent These values show that cordon pricing is more effective in

forcing individuals not to use their car than increasing parking cost with the same value Results also

show that a 1 Rial increase in parking cost and fuel cost simultaneously decreases the probability of

choosing car usage by 0000001 percent Table 5 also shows that a 1 minute decrease in transit

access time would result in a 09 percent increase in probability of choosing this mode It also shows

that each 1 minute decrease in transit travel time and transit access leads to a 0004 decrease in the

probability of choosing the DampR mode

22

The above results show that just one of the policy interactions is not significantly different from

none (ie CordonampFuel) and push policies have a significant role in impelling car users to consider

other modes

7 Planning

To demonstrate the application of the model some of its planning aspects for the city of Tehran are

presented here In fact the model is able to assess the effect of all studied variables which were

grouped into three tiers on commuters mode choice In this section we assess the effect of TDM

policies by adopting their pair-wise combinations for the city Table 6 shows that pull policies in this

study did not impose a major effect on car usage Because the average access time of the sample

was 11 minutes improving it by 50 percent (about 5 minutes) increases the probability of WampR to

509069100 which is equal to a 048 percent increase in considering this mode (and decrease in all

other modes) On the other hand decreasing transit time about 50 percent (about 19 minutes based

on sample average) and similarly improving access time decreases the probability of considering

DampR to 0004195100 which is equal to a 038 percent decrease in considering this mode (and

increase in all other modes) Therefore in this section we focused on the push policies which

appear in the utility function of car mode It is worth noting that in assessing each combination of

policies all other variables were assumed to be fixed as the current state

71 Parking pricing and cordon pricing

The average hourly parking charge of individuals in the sample is 71 Rials This charge is 3000 Rials

per hour in the curb lane on some streets about 2000 Rials per hour in parking lots and free in

alleys and other streets The average amount shows that most of the commuters benefited from

free parking spaces in alleys and non-controlled streets Figure 2 depicts the result of implementing

this policy and the cordon pricing policy simultaneously In line with marginal effects it can be seen

that in implementing each policy separately cordon pricing is more effective than parking pricing in

the planning range It is worth noting that commuters in this study are less sensitive to parking

23

pricing than cordon pricing One reason may be that they thought in any case they would find free

parking space Figure 2 shows that the effect of each policy depends on the level of the other policy

In other words the effect of the simultaneous implementation of two policies is not equal to the

sum of their separate individual effects

Figure 2- Effect of implementing parking pricing and cordon pricing

72 Parking pricing and increasing fuel cost

Figure 3 shows the simultaneous effects of parking pricing and increasing fuel cost policies Although

both of these policies are not effective separately their simultaneous effect is significant It can be

seen that their synergy appears in the higher levels of both policies

Figure 3 - Effect of implementing parking pricing and increasing fuel cost

24

73 Cordon pricing and increasing fuel cost

The effect of implementing cordon pricing and increasing fuel cost is shown in Figure 4 It can be

seen that these two policies have no interaction effect in decreasing car usage This result is verified

by this interactions insignificant marginal value in Table 6

Figure 4- Effect of implementing cordon pricing and increasing fuel cost

8 Conclusions

This study examined the role of TDM policies in individual mode choice for work trips in the city of

Tehran Five policies including increasing parking cost increasing fuel cost cordon pricing in the odd-

even zone of the city transit time reduction and transit access improvement were investigated Of

the five policies the former three were push policies and the latter two were pull policies The

design of experiments approach was used to design the questionnaire that would capture the stated

preferences of car commuters

A number of nested structures based on recognizing differences in the variances associated

with unobserved influences were examined and finally the MNL as a superior model for this study

was developed

25

The results show that for car users who faced TDM policies all of the five studied TDM

policies were factors in the mode choice process The model shows that in considering car usage

push policies play a main role and pull policies do not

Results also show that the interactions of policies are also significant The model shows that

the interaction of parking cost policy and fuel cost policy is significant in prompting car users to

consider other modes It also shows that the interaction of cordon pricing and increase in fuel cost is

effective to discourage motorcycle usage Furthermore the interaction of transit time reduction

policy and access improvements policy is significant in considering DampR as a transportation mode to

workplace

Pull policies in the study were expected to attract individuals to transit modes (pull policies

of this study are related to transit modes) This was verified by the model

Results show that although taxi usage is usually treated as an alternative for urban trips it

was not directly affected by the studied policies This may be due to its special function in Tehran as

a non-public and non-private mode

The results also show that except for the interaction of cordon pricing and increase in fuel

cost other interaction effects have significant marginal effects on mode choice

Results of the model show that to implement a single policy cordon pricing is the most

effective in decreasing car usage

This study assessed workplace commuters Thus future studies should explore the

individuals mode choice with other trip aims Additionally including more policies especially pull

policies in favor of non-transit modes is reasonable to rigorously assess the model

It can be seen that the studied modes are affected differently by the TDM policies and their

interactions The impact of the interaction of policies appeared in the consideration of three out of

six studied modes and also in two out of three developed graphs This issue needs more attention

within the transportation demand management context and further studies should be performed

26

Design of this study is based on the consideration of all two-way interactions with some

degrees of deficiency in D-efficient policy Focusing on the significant interactions of this study is

recommended to perform other studies with more D-efficient values

Further understanding of the impacts of TDM policies might be provided in future research

by focusing on the amount of their synergy in prohibiting car usage

9 References

Arentze TA Hofman F amp Timmermans HJP (2004) Predicting multi-faceted activityndashtravel adjustmentstrategies in response to possible congestion pricing scenarios using an Internet-based stated adaptationexperiment Transport Policy 11 p31ndash41

Burris M amp Patil S (2009) Estimating the Benefits of Managed Lanes Technical Report Texas TexasTransportation Institute University Transportation Center for Mobility

Cao X amp Mokhtarian PL (2005) How do individuals adapt their personal travel A conceptual exploration ofthe consideation of travel-related strategies Transport Policy 12(3) pp199-206

de Palma A amp Lindsey R (2001) Transportation Supply and congestion In International Encyclopedia of theSocial and Behavioral Sciences 1st ed Elsevier p15882ndash15888

Eriksson L Garvill J amp Nordlund AM (2008) Acceptability of single and combined transport policy measuresThe importance of environmental and policy specific beliefs Transportation Research Part A 42 p1117ndash1128

Eriksson L Nordlund AM amp Garvill J (2010) Expected car use reduction in response to structural traveldemand management measures Transportation Research Part F 13 p329ndash342

Golob TF amp Hensher DA (2007) The Trip chaining Activity of Sydney Residents A Cross-Section Assessmentby Age Group with a Focus on Seniors Journal of Transport Geography 15 p298ndash312

Hensher DA amp King J (2001) Parking demand and responsiveness to supply pricing and location in theSydney central business district Transportation Research A 35 pp177-96

Hensher DA amp Rose JM (2007) Development of commuter and non-commuter mode choice models for theassessment of new public transport infrastructure projects A case study Transportation Research Part A41 p428ndash443

Hensher DA Rose JM amp Greene WH (2005) Applied Choice Analysis A Primer New York CambridgeUniversity Press

Iranian Center of Statistics (ICS) (2009) Information of Iranian States [Online] Available at HYPERLINK httpwwwamarorgirUploadModulesContentsasset16tehrantehparthtmlhttpwwwamarorgirUploadModulesContentsasset16tehrantehparthtml [Accessed 13 November2009]

Kingham S Dickinson J amp Copsey S (2001) Travelling to work Will people move out of their cars TransportPolicy 8 pp151-60

Kuhfeld WF (2009) Marketing Research Methods in SAS Experimental Design Choice Conjoint andGraphical Techniques SAS 92 Edition NC USA SAS Institute Inc Cary Available on the Webwwwsupportsascomtechsuptnotetnote_stathtmlmarket

Litman T (2003) The online TDM Encyclopedia mobility management information gateway Transport Policy10 pp245-49

Litman T (2010) Online TDM Encyclopedia [Online] Available at HYPERLINK httpwwwvtpiorghttpwwwvtpiorg [Accessed 21 Augest 2010]

4

or against their car usage To analyze the effects of such decisions on individuals travel patterns one

should be aware of the interactions in addition to the policies

The above discussions show that introducing more than one policy to manage the transportation

demand of a megalopolis is a common problem faced by their policymakers This paper is focused on

modeling the role of multi-TDM policies on commuters mode choice especially in regards to the

interactions of these policies The model provides a number of graphs that enhance transportation

planning for the city of Tehran After describing the research context this paper describes the stated

choice design and the stated preferences survey Then the developed mode choice model is

explored followed by presentation of the graphs The conclusion summarizes the findings and

discusses the implications of the results

2 Research context

Although there are many studies that look at the impact of a single TDM policy on a society such as

studies on congestion pricing (Arentze et al 2004) park and ride (Parkhurst 2000) and parking

pricing (Visser amp Van der Mede 1986 Hensher amp King 2001) only some studies focus on the impact

of multiple policies Among these a few look at the simultaneous implementation of TDM policies

In the context of multiple policies Stradling et al detected motorists who are ready to reduce their

car usage and how they should be helped to change (Stradling et al 2000) In fact individual

reasons for car usage and sensitivity to a number of pull and push policies were detected through a

postal questionnaire survey of English car drivers Those authors verified the difference between pull

and push policies through a factor analysis approach The researchers also found the ranking of the

TDM policies stated by the drivers Another study performed by Mackett focused on pull policies and

personal actions that might attract motorists out of their cars and into transportation alternatives

for short trips (Mackett 2001) He examined various events that could attract car drivers to use an

alternative mode and their associated actions He classified these actions as collective actions

(actions that may be undertaken by the government or other organizations) or non-collective

5

actions (other actions) and assessed the role of policies coerciveness on driver mode change by

assuming that the collective policies are more coercive than the non-collective policies By focusing

on a smaller community and introducing a push policy (fuel pricing) in addition to some pull policies

Kingham et al examined the travel behavior of two companies employees (Kingham et al 2001)

through a survey They studied each employees perception of hisher mode choice during the trip

to work and investigated the potential of transferring car trips to other modes in the presence of

studied policies They also examined the importance of policies that would encourage employees to

use public transit or car-share to travel to work

Although the above studies focused on individualsrsquo perceptions of car-use for daily trips by adopting

some assumptions on the mode choice model of Shiraz Iran Zareii examined the results of

implementing five push policies in the city in terms of total travel time and amount of CO2 emission

(Zareii 2003) Because this study determines car usage by transferring the imposed cost of each

policy to extra time (calculated via the individuals time-value) push policies were of interest

Loukopoulos et al attempted to obtain quantitative estimates of the size of car usage reduction

goals and frequency of implementation of adaptation alternatives (Loukopoulos et al 2004) They

assessed two push policies and a pull policy for different trip aims and examined the cost-

minimization principle in relation to five adaptation alternatives OFallon et al explored the

potential effect of 11 policies on the respondents decision to choose to drive a car to work or school

during the morning peak period in three cities of New Zealand through a stated preferences survey

(OFallon et al 2004) They also reported the marginal effect of each of the studied policies and

recommended a study with fewer policies to explore the possible impacts of combinations of

specific policies Washbrook et al examined the role of seven policies on mode choice (Washbrook

et al 2006) Although the design of this study focused on the policiesrsquo main effects the results were

used to estimate commuter response to various policy combinations of charges and incentives

6

Because the aims of the above studies were to look for the best studied policy by assessing its

impact on car usage they did not deal with the effects of simultaneous implementation of TDM

policies Pendyala et al assessed five TDM policies by adopting an activity-based micro-simulation

model system (AMOS) to simulate changes in individual travel patterns (Pendyala et al 1997) In

their survey they also assessed combinations of specific policies in four transportation control

management scenarios and determined the possible impacts in those scenarios Thorpe et al

presented the individuals attitudinal responses to three push and one pull TDM policies in two case-

study cities in the UK Cambridge and Newcastle (Thorpe et al 2000) They examined the

relationship between the perceived effectiveness and public acceptance of alternative TDM policies

and showed that the public acceptance order of generic TDM policies is improving public transit

road-user charging zone-access controls and increased parking charge This study concluded that

there was evidence of interaction effects between levels of public acceptance of TDM policies when

considered separately and in combination with other policies Further these effects could be

investigated more rigorously with a stated preferences experimental design of alternative TDM

packages which allow the investigation of both main and interaction effects1

Eriksson et al examined the acceptability of one push policy (raised tax on fuel) and two pull policies

(improved public transport and subsidized renewable energy) individually and as packages

combining one push and one pull policy (Eriksson et al 2008) By proposing a model of factors

predicting acceptability of TDM policies they concluded that while the pull policies are perceived to

be effective fair and acceptable the push policy and the packages are perceived to be ineffective

unfair and unacceptable By removing one of the pull policies (ie subsidizing renewable energy)

these authors further assessed the expected car usage reduction in response to other two policies

(Eriksson et al 2010) By focusing on improved public transport raised tax on fuel and their

combination as a package the results showed that the combination was more effective than the

1 In a few studies in choice modeling researchers also examined the second order interactions of attributes in the models(eg (Mogas et al 2006))

7

individual policies Vieira et al explored the concept of multi-instrumentality as a procedure of

policy integration and implementation whereby a systematic search for complementary policies was

sought when planning and designing one (or several) core policy(s) aiming to fulfill one particular

policy more effectively (Vieira et al 2007) They defined criteria to assess the TDM policies and

analyzed four improvement mechanisms in each pair of policies Based on meta-studies they

defined the potential improvement between different types of policies By defining synergy concept

as a benefit of integration May et al reviewed a number of examples to assess the concept and

found little evidence of synergy in outcome indicators (May et al 2006)

Based on the above discussion assessing individual behavioral response to more than one TDM

policy is an interesting issue within the TDM context The following three issues are addressed in this

paper developing a model to investigate the role of TDM policies in commuters mode choice

exploring the role of effective parameters on the consideration of each mode of travel and

suggesting a method to determine the results of implementing two TDM policies simultaneously In

this paper the stated preferences approach is used to model the car users mode choice using the

design of experiments principles

3 Stated preferences

The five policies selected for the city of Tehran consisted of three push and two pull policies The

policies were increasing parking cost increasing fuel cost cordon pricing into an odd-even zone2

transit (bus or subway) time reduction and transit access improvement The latter two were

described by setting measures in favor of the public transit vehicles in streets and intersections

decreasing the time of boarding and alighting at the stations and increasing the number of transit

lines and stops in the city

2 This zone explored in the next section

8

Parking costs fuel costs and public transit time policies are designated with three levels and cordon

price and public access time are designed with two levels Table 1 shows the policies and their levels

All push policies had fixed values for their levels for pull policies because there were variations in

the transit time and transit access time for individuals proportional values of the current state were

used which is different for each individual The term no change in Table 1 refers to the current value

of a policy that each individual already experiences The mean values are also presented in Table 1

for a better description of current state

In preparing a questionnaire for the stated preferences part the design of experiments approach

was adopted Full factorial design is the most general type of design in this approach which

introduces all combinations of all levels in the modeling process In other words full factorial design

produces 108 possible choice sets (33322) This design allows the investigation of all

interactions as well as the main effects in the model On the one hand fewer choice sets are

available when ignoring the effects of higher-order interactions and on the other hand these

interactions have a negligible role in the variance (Louviere et al 2000 Hensher et al 2005) thus

fractional factorial design methods have been proposed

Table 1- Policies and their levels

Measure Type Numberof levels

Description of levels Mean Value

Increasing parking cost Push 3 No change 4000 7000 Rials3 h 71 RialshNACordon pricing Push 2 25000 50000 Rialsday

Increasing fuel cost Push 3 No change 3000 5000 Rialsliter 1470 RialsLiterTransit time reduction Pull 3 No change 15 30 percent shortage 385 minTransit access improvement Pull 2 No change 25 percent shortage 11 min

Efficient design a type of fractional factorial design was used in the study and a design with 895

efficiency was adopted which allows assessing all two-way interactions of policies as well as the

3 10000 Rials are almost equal to 1 US dollar

9

main effects with only 36 choice sets4 (See (Rose amp Bliemer 2009) or (Kuhfeld 2009) for more

details on efficient design) To avoid a time-consuming questionnaire 36 choice sets (scenarios)

were randomly ordered and divided into six separate questionnaire types coded as 1 to 6 Each of

the questionnaires had six scenarios and each scenario consisted of five policies

4 Survey

Two push policies are currently being implemented in the city of Tehran The first is car-free

planning in the CBD area of the city and the second one is an odd-even scheme based on the last

digit of car plates that attempt to enter a zone which is about three times larger than and includes

the CBD area Based on their occupation a few people can drive to the CBD area with a license

called permission A stated preferences survey was assigned for the morning car commuters to the

odd-even zone but they were asked to ignore these two policies to find the accurate sensitivity of

individuals to the study policies The odd-even zone is selected as study area for the two following

reasons 1) because of odd-even control respondents are familiar with the fringes and they can

better imagine the entrance pricing area and 2) respondents are familiar with the limits that they

face half of the week and are thus aware of the alternative existing modes Compared to the CBD

area this zone covers more car commuters and the entrance restriction is more imaginable for this

zone than the former one Respondents were interviewed face-to-face in their workplaces midway

through the year 2009 The interviews were enhanced with a special card to better define the

scenarios

For this study 2196 scenario observations from 366 individuals were adopted The sample included

308 men (ie 841) and 58 women (ie 159) The figures are close to the employment

percentages in the city according to the Iranian Center of Statistics (ICS) This source indicates that

825 of Tehran employees are men and 175 are women (Iranian Center of Statistics (ICS) 2009)

Because this study focuses on car-using commuters comparisons between the sample and city data

4 Efficient design is also adopted in other studies such as managed lanes (Burris amp Patil 2009)

10

especially regarding educational distribution were impossible Table 2 presents demographics of the

sample

Table 2- Demographics gender marital household (HH) size employee type age HH employee(s)

Amount Percent

Gender Male 308 841Female 58 159

Marital Single 100 273Married 266 727

HH Size 1 4 112 86 2353 129 3524 90 2465 42 1156+ 15 41

Age 18~29 122 33330~39 146 39940~49 58 15950~59 32 8760+ 8 22

HH employee(s) 1 156 4262 159 4343 41 1124+ 10 27

The first part of the questionnaire is dedicated to gathering the occupation state home and job

locations the distance between these locations round-trip time (from home to workplace and then

workplace to home) and all car trip characteristics in the previous day or the day before it based on

plate number It was necessary that the respondents drive hisher car in the day studied to complete

the trip diary portion of the questionnaire5 The general reasons for car usage and the scenarios

formed the next portion In each scenario every respondent was asked the question How would

you travel to the workplace if all of these changes were in place on the day studied For example

one may have to pay 4000 Rialsh for parking 50000 Rials per entrance to the cordon the same

amount in transit access and fuel cost and a 15 percent decrease in transit time simultaneously

Depending on individual responses six main options were distinguished6 These choices were still

5 In designing the questionnaire the general form of questionnaire which has mentioned in OFallons study was adopted6 In the pre-test survey 14 modes is distinguished

11

drive a car (C) walk to the station and catch public transit (WampR) drive to a public station and catch

public transit (DampR) ride a motorcycle (MC) catch a taxi7 (T) and catch a taxi by phone (T_T) DampR is

somewhat different than the more familiar ldquoPark amp Riderdquo In fact in the fringes of the odd-even

zone there were no specialized parking lots dedicated to this purpose and commuters considered

Drive amp Ride because they were not allowed to pass the fringes

After each scenario if the respondent changed hisher mode the reason(s) for the change were

asked It could be a sole policy or a bundle of them Furthermore travel-related information was

sought These data were not part of the stated choice but they might have important influences on

individual choices These data consisted of car dependency (need to drive someone or move freight

in the trip) parking place type and average weekly parking costs car and motorcycle ownership and

number of household driving licenses

Depending on the individuals activity in that day three types of activity patterns were detected

Pattern 1 described individuals who had no stop in their commute Pattern 2 was for individuals who

had at least one stop on their way to or from work and pattern 3 was for the individuals who went

to another workplace in their daily activities

Finally for the sake of data generalization and the examination of household characteristics gender

age and household type employment status and education level were also asked

5 Mode choice model

In order to detect the policies that affect individual mode choice the logit modeling approach was

adopted In this model one can determine if the interaction of two policies affects the mode choice

In the calibration step 152 variables were defined and their effects on consideration of each mode

were examined

7 Taxis in Iran are somewhat different than taxis in other countries of the world In fact taxis in Iran are not hiring by oneperson or a group of people at a time Taxis allow passengers to board or alight along their path with respect to theircapacity In other word this mode is functioning similar to transit vehicles but the stops are not predefined

12

51 Model structure

Initially a multinomial logit (MNL) model is developed (Figure 1a) By selecting a number of tree

structures based on recognizing differences in the variances associated with unobserved influences

we find that the greatest similarity in variance profiles is associated with public transport modes as

opposed to non-public modes (Figure 1b) This structure has two nests one including Car (C) and

Motorcycle (MC) as private modes and the other including Walk and Ride (WampR) Drive and Ride

(DampR) Taxi (T) and Tel-taxi (T_T) as non-private modes The result of this nested logit (NL) model is

shown in Table 3

Although it is not a statistically significant improvement overall on the MNL model the statistically

significant inclusive value8 (IV) of 0889 for non-public modes relative to the fixed parameter value of

10 for public modes suggests that there is a structural advantage in selecting the NL specification

The normal test of a statistically significant difference between NL and MNL is an IV parameter

relative to 10 calculated using a Wald-test via equation 1

)1(Wald-test = (IVparameter ndash 1)std error

a The MNL structure

b Final nested structure

Figure 1- Model structure

8 Also called scale parameter

Alternatives

MCCar T_TT WampR DampR

Alternatives

MCCar T_TT

Public

WampR DampR

Private

13

We have (0889-1)2508 =-075 which would be rejected at the usual acceptable significance levels

This suggests that the NL model could be collapsed into an MNL form

Table 3- Nested logit (NL) model resultValueParameter

0889IV (Private)1000IV (nPrivate)

-2668335L( )-4057684L(0)

0342sup2

After the calibration process the variables that were statistically significant were identified and are

presented in Table 4 Table 5 presents the final model of the study with a goodness of fit of 031 for

the six studied modes For a general review of the model calibration results the effective factors can

be grouped under the following three categories TDM policy characteristics commuting trip

characteristics and household socio-economic characteristics which are all treated as alternative-

specific variables

52 Model results

Car (C)

It is expected that push policies impel car-drivers to choose other modes Table 5 shows that cordon

pricing and increase in parking cost cause individuals to choose not to use their car This is in line

with other studies suggesting that these policies are effective to discourage car usage (Hensher amp

Rose 2007 OFallon et al 2004) In addition the interaction between the policies of fuel cost

increase and increase in parking cost shows similar car usage discourage effect Because fuel cost is

related to the distance between home and work locations and parking cost is related to work time

the time that an individual spends out of the home is negatively affected by hisher likelihood to use

a car

14

Table 4 - Definition of the significant variables

AbbreviationVariableTransportation demand management measures

Measures

ParkingParking cost increase Rials per hour

CordonCordon price Rials per entranceAccessTransit access time shortage percent

Interaction of push measures

ParkampFuelParking cost and fuel cost simultaneous effectsCordonampFuelCordon pricing and fuel cost simultaneous effects

Interaction of pull measures

PT_timeampaccessPT time reduction and access improvement simultaneous effectsCommuting trip characteristics

Trip distanceDistance between home and workplaceTrip timeTravel time between home and workplace

Exp FuelLikelihood of unsubsidized fuel use (self-reported on a Likert scale)NtripsNumber of daily tripsPattern2Commuting with 1+ stop(s) in go or return

Pattern3Commuting with 2 workplacesFirst trip timeStart time of first trip

PnocarwkLikelihood of going to work in absence of that car (self-reported)PTnwaccNon-walk access to transit (yes=1)First NaccoNumber of passengers in first trip

PassengerAny passenger on that day (yes=1)Park_paymentParking payment in last weekNhempfullNumber of full employees in HH

CardependencyBoardalight a passenger or move freight in the trip (yes=1)D car ownBe the owner of the used vehicle (yes=1)

Car accCar accessibility in household (number of cars to number of HH driving licenses ratio)NmotorcycleNumber of motorcycles owned by HHD home placeHome Location is in study area (yes=1)

PermissionPermission to enter to study area (yes=1)ComfortI use my car because it is comfortablePoor_PTI use my car because transit is not good

HH socio-economic characteristics

FemaleGender (Female=1)Age lt30Age younger than 30 (yes=1)Age 30_39Age between 30 to 39 (yes=1)Job_durationNumber of years that individual has been at hisher job

Emp_fullFull-time employee (yes=1)Edu BSDegree of education is BSc (yes=1)Edu BS+Degree of education is higher than BSc(yes=1)

D childlt=18Child younger than 18 in HH (yes=1)

15

Table 5 ndash The mode choice model

Tel-Taxi(T_T)

Motorcycle(MC)

Drive amp Ride(DampR)

Taxi (T)Walk amp Ride(WampR)

Car (C)Mode

Variable-471756-37067-147911Constant

Transportation demand management measure variables00019-00045Cordon

-000072Parking-004308Access

-28443D-05Parkampfuel-32475D-06Cordonampfuel

00029Pt_timeampaccess

Commuting trip characteristics-04709Trip distance

-02163-00831Trip time-96755163655Exp fuel-16253Ntrips

-114779Pattern2-71008Pattern3

00282-00270First trip time-02439-01549Pnocarwk

-11322992883-32765PTnwacc-133701First Nacco

-7778-73782Accompany-00049000010Park_payment

201646195554Nhempfull-160144ComfortCar1

-206142DependencyCar1-16101883385-121224DependencyCar2

42176Poor_PTCar1-24988Poor_PTCar2

- -27221D car own70960-39136Car acc

1 -71112-156123Nmotorcycle-1436322762D home place

2 78826Permission

HH socio-economic characteristics149490Female

297584-24548Agelt30-136490Age30_39

079430366303585Job_duration-108743Emp_full-203468-64900Edu BS

10932856687-4499984445Edu BS+102271D childlt=18

-2677366L( )-3849556L(0)0305sup2

112127178592580607N

Note = Positive significance at 1 5 10 level

As expected individuals with higher income are more likely to use their car This is indicated in the

model by the positive signs of individuals who use fuel with fixed (unsubsidized) cost and individuals

16

who pay more in parking charges in the previous week of study Negative sign of Pnocarwk variable

shows that the commuters who stated that their commute depends on car availability are more

likely to use their car Individuals in households with more full-time employees are more likely to use

their car which may be the result of higher household income Not surprisingly commuters who

have permission are more likely to maintain car usage Among the household socio-economic

parameters greater job experience (Job_duration) and higher graduate levels (EduBS+) increase the

probability of car usage

Public transit accessed by walking (WampR)

Access time to transit negatively impacts WampR choice which is expected This result is similar to

findings for the city of Sydney (Hensher amp Rose 2007) The negative coefficient of first trip time

indicates that individuals are more likely to use WampR in the early morning This result seems to

reflect the better weather for walking and faster speed of WampR mode early in the morning

Obviously individuals who are not able to access transit stations via walking (PTnwacc) are less likely

to consider this mode Furthermore serving passengers on daily trips is also a deterrent to using

WampR

Initially assessing the individuals who stated that their car usage is due to poor public transit service

(Poor_PT) led to an unexpected result in favor of considering WampR By introducing to this variable

the number of household cars as a proxy for household income (Poor_PTCar1) the model shows

that of the previously mentioned individuals those who have lower income are the ones who have

to consider WampR The result is understandable as these individuals may have no alternative when

they have to change their mode (they also are not likely to consider other modes) Individuals with

higher levels of income who have to use their car during before or after work (Dependencycar1+)

are not likely to use WampR

The greater the number of motorcycles in a household the less likely commuters is to consider

WampR There appears to be a competition between motorcycle and PT for access to the city center

17

Better PT services in the center of the city in terms of coverage and frequency increases the

likelihood that its residents will consider WampR This is verified by the positive sign of the

D_home_place variable Commuters with greater job experience (Job_duration) in their workplace

are more likely to use this mode Although individuals with higher levels of education are not likely

to use WampR as education level increases avoidance of WampR decreases

Taxi (T)

Table 5 shows that none of the studied policies are significant in considering taxi usage It seems that

taxi usage considering its function in Iran as a non-private and non-public mode of transport is not

affected by pull or push policies A negative sign for taxi travel time indicates that individuals are not

likely to use this mode for longer trips This seems reasonable given that longer trips are more

expensive Commuters who are more likely to use fuel with no subsidy are not likely to use taxis As

mentioned before they prefer to use their car A higher number of trips in a day are also a deterrent

to considering taxi usage which may be due to increased cost for more trips Results show that an

individual with more daily trips avoids using taxis Commuters who are employed in more than one

workplace (Pattern 3) are not likely to use taxis This may be due to the fact that they have a lower

level of income which forces them to dedicate more time on the job

Initial results showed that individuals who stated that their car usage is due to poor public transit

service (Poor_PT) are not likely to use taxis This result was far from our expectations By introducing

to this variable the number of household cars as a proxy for household income the model shows

that the previously mentioned individuals who have higher income (Poor_PTCar1+) are the ones

who are not likely to consider taxis Furthermore because such individuals are not considering any

other modes they may treat taxi usage as a kind of PT mode with poor service

As expected greater access to cars in a household (Car_acc) lessens the likelihood of considering

taxis as an alternative Furthermore individuals in households with more motorcycle ownership are

less likely to consider taxis It seems like there is a competition among motorcycles and taxis for

18

access to the city center Younger commuters are less likely to use taxis and individuals with at least

master degrees do consider this mode in addition to their car

Public transit accessed by Drive (DampR)

This mode is affected by the simultaneous interaction of transit time and transit access

(PT_TimeampAccess) which is reflected in the fact that individuals prefer to use this mode for longer

trips Comparing this mode and WampR the first trip start time affects the consideration of this mode

differently Later morning commuters prefer to use their car to access PT modes Such commuters

may have higher income levels or managerial jobs Obviously individuals who are not able to access

PT stations by walking (PTnwacc) are likely to use DampR Serving passengers in daily trips is also a

deterrent in considering this mode which is similar to WampR but with a lower coefficient

Commuters with higher income levels who depend on their car during before or after work

(Dependencycar1+) are likely to use DampR Individuals who use their own car are less likely to use

this mode which is unexpected As a city center develops better PT network coverage and residents

have smaller distances to their workplaces they are unlikely to use DampR This is proven in the model

by a negative sign for D_home_place

Motorcycle (MC)

Increasing fuel cost and cordon pricing simultaneously discourage motorcycle usages Although fuel

cost is expected to reduce motorcycle usage to some extent its combined effect with cordon pricing

also reduces motorcycle usage However this variable is not as strong as other policy variables

=10)

Of the studied modes motorcycle usage is affected by the most commuting variables This may be

due to the fact that this mode is not common Commuting distance has a negative effect on

motorcycle usage which is expected It is worth noting that trip distance appears only in this mode

which may be a reflection of the role of distance in regards to the safety risk in considering this

19

mode Commuters with more stops to serve passengers while commuting (Pattern 2) are not likely

to use this mode which may be due to the poor passenger service of this mode

Individuals who state that commuting is independent of the mode (Pnocarwk) are not likely to use

MC By looking at the (First_Nacco) negative sign this could stem from the fact that the more

passengers there are on the first trip the less likely individuals are to consider MC Regarding the

low capacity of MC and its safety concerns such commuters avoid using this mode Commuters who

pay more parking charges (Park_payment) are less likely to use MC which is expected Individuals

who are dependent on their car during before or after their work time are not likely to use MC

even if they have lower levels of income (DependencyCar1) Individuals who use their own car

(D_car_own) are less likely to use this mode As expected individuals who live in households with

more motorcycle ownership are more likely to use this mode The positive sign of (Permission)

indicates that commuters who have permission to enter the study area do consider MC Because

such commuters generally provide that permission just for car usage this result is unexpected

As with commute variables of all the studied modes MC is affected by the greatest number of

socio-economic variables As expected young commuters (Agelt30) are more likely to use this mode

Commuters with Bachelor of Science degree are less likely to use this mode among others Full time

employees (Emp_full) are less likely to consider MC whereas commuters with more experience in

their jobs prefer to use it Results show that individuals who live in a household with children

younger than 18 are more likely to consider using a car

Tel-Taxi (T_T)

Results show that cordon pricing causes higher probability of using T_T In fact individuals who use

T_T as a mode with similar level of service as cars9 are more willing to pay the cost and make use of

the mode It is worth noting that the effect of cordon pricing in pushing commuters from car usage

9 As this mode does not have driving stress and parking search time in some cases it may have more amount of utility thana car does

20

(000045) is greater than its effect on pulling them to Tel-taxi (000019) This is because of the

possibility of considering other non-car modes

Because consideration of this mode is a function of its operation travel time (Trip_time) appears as

a deterrent in this mode utility function Table 5 shows that individuals are more sensitive to the trip

time when using T_T mode versus taxi which is expected due to their relative costs

The greater the number of full time employees in a family (Nhempfull) the higher the probability of

considering T_T by its commuters which may be due to the higher income level of these

households This is verified by the greater likelihood of using T_T rather than taxis by such

commuters Individuals with higher levels of income who depend on their car during before or after

work time are less likely to use T_T Commuters with lower income levels who state that they use

their car for the sake of comfort (Comfortcar1) are less likely to use T_T which may be due to its

cost Although such individuals do not consider any other modes they specifically avoid T_T Greater

access to cars in a household leads to greater likelihood of T_T usage which could be due to the

higher income level of a household As mentioned before such individuals even avoid taxis

Females who drive to their workplace are more likely to use T_T It seems like this part of society

considers this mode when desiring to avoid the difficulties of driving Younger commuters are less

likely to use T_T and individuals between 30 and 39 years of age are specifically avoiding this mode

Results show that university graduated commuters are more likely to use this mode

6 Marginal effects

To explore the effects of each policy on mode choice and to answer the second issue raised at the

beginning of this paper the marginal effects approach can be adopted Although the coefficients of

the models utility functions show the drivers behavior when facing one or more policies the

marginal effects of policies or their interactions may appropriately show the results of their

implementation More specifically the marginal effect for this study is interpreted as the change in

21

probability given a unit change in a variable ceteris paribus In this section the variable is defined as

a specific policy or a specific interaction of two policies Table 6 presents the marginal effects of the

studied policies and their interactions with mode choice The results are shown in the form of trip

percentages transferred away from the car to the studied modes and the probability-weighted

sample enumeration approach is adopted to find the values It is worth noting that this table is fully

compatible with Table 5 but the marginal effects that were less significant than 90 percent have

been removed

Table 6 - Marginal effects of policies (percent)

Tel-Taxi(TT)

Motorcycle(MC)

Drive ampRide(DampR)

Taxi (T)Walk ampRide(WampR)

Car (C)Mode

Variable-000088Cordon-000140Parking

-09069Access-0000001ParkampFuel

00040PT_TimeampAccess

Table 5 shows that a 1 Rial increase in cordon pricing causes a 000088 percent decrease in car

usage Furthermore a 1 Rial per hour increase in parking cost decreases the car usage probability to

00014 percent By assuming 8 hours for the average parking duration the daily marginal value of

parking cost converts to 000018 percent These values show that cordon pricing is more effective in

forcing individuals not to use their car than increasing parking cost with the same value Results also

show that a 1 Rial increase in parking cost and fuel cost simultaneously decreases the probability of

choosing car usage by 0000001 percent Table 5 also shows that a 1 minute decrease in transit

access time would result in a 09 percent increase in probability of choosing this mode It also shows

that each 1 minute decrease in transit travel time and transit access leads to a 0004 decrease in the

probability of choosing the DampR mode

22

The above results show that just one of the policy interactions is not significantly different from

none (ie CordonampFuel) and push policies have a significant role in impelling car users to consider

other modes

7 Planning

To demonstrate the application of the model some of its planning aspects for the city of Tehran are

presented here In fact the model is able to assess the effect of all studied variables which were

grouped into three tiers on commuters mode choice In this section we assess the effect of TDM

policies by adopting their pair-wise combinations for the city Table 6 shows that pull policies in this

study did not impose a major effect on car usage Because the average access time of the sample

was 11 minutes improving it by 50 percent (about 5 minutes) increases the probability of WampR to

509069100 which is equal to a 048 percent increase in considering this mode (and decrease in all

other modes) On the other hand decreasing transit time about 50 percent (about 19 minutes based

on sample average) and similarly improving access time decreases the probability of considering

DampR to 0004195100 which is equal to a 038 percent decrease in considering this mode (and

increase in all other modes) Therefore in this section we focused on the push policies which

appear in the utility function of car mode It is worth noting that in assessing each combination of

policies all other variables were assumed to be fixed as the current state

71 Parking pricing and cordon pricing

The average hourly parking charge of individuals in the sample is 71 Rials This charge is 3000 Rials

per hour in the curb lane on some streets about 2000 Rials per hour in parking lots and free in

alleys and other streets The average amount shows that most of the commuters benefited from

free parking spaces in alleys and non-controlled streets Figure 2 depicts the result of implementing

this policy and the cordon pricing policy simultaneously In line with marginal effects it can be seen

that in implementing each policy separately cordon pricing is more effective than parking pricing in

the planning range It is worth noting that commuters in this study are less sensitive to parking

23

pricing than cordon pricing One reason may be that they thought in any case they would find free

parking space Figure 2 shows that the effect of each policy depends on the level of the other policy

In other words the effect of the simultaneous implementation of two policies is not equal to the

sum of their separate individual effects

Figure 2- Effect of implementing parking pricing and cordon pricing

72 Parking pricing and increasing fuel cost

Figure 3 shows the simultaneous effects of parking pricing and increasing fuel cost policies Although

both of these policies are not effective separately their simultaneous effect is significant It can be

seen that their synergy appears in the higher levels of both policies

Figure 3 - Effect of implementing parking pricing and increasing fuel cost

24

73 Cordon pricing and increasing fuel cost

The effect of implementing cordon pricing and increasing fuel cost is shown in Figure 4 It can be

seen that these two policies have no interaction effect in decreasing car usage This result is verified

by this interactions insignificant marginal value in Table 6

Figure 4- Effect of implementing cordon pricing and increasing fuel cost

8 Conclusions

This study examined the role of TDM policies in individual mode choice for work trips in the city of

Tehran Five policies including increasing parking cost increasing fuel cost cordon pricing in the odd-

even zone of the city transit time reduction and transit access improvement were investigated Of

the five policies the former three were push policies and the latter two were pull policies The

design of experiments approach was used to design the questionnaire that would capture the stated

preferences of car commuters

A number of nested structures based on recognizing differences in the variances associated

with unobserved influences were examined and finally the MNL as a superior model for this study

was developed

25

The results show that for car users who faced TDM policies all of the five studied TDM

policies were factors in the mode choice process The model shows that in considering car usage

push policies play a main role and pull policies do not

Results also show that the interactions of policies are also significant The model shows that

the interaction of parking cost policy and fuel cost policy is significant in prompting car users to

consider other modes It also shows that the interaction of cordon pricing and increase in fuel cost is

effective to discourage motorcycle usage Furthermore the interaction of transit time reduction

policy and access improvements policy is significant in considering DampR as a transportation mode to

workplace

Pull policies in the study were expected to attract individuals to transit modes (pull policies

of this study are related to transit modes) This was verified by the model

Results show that although taxi usage is usually treated as an alternative for urban trips it

was not directly affected by the studied policies This may be due to its special function in Tehran as

a non-public and non-private mode

The results also show that except for the interaction of cordon pricing and increase in fuel

cost other interaction effects have significant marginal effects on mode choice

Results of the model show that to implement a single policy cordon pricing is the most

effective in decreasing car usage

This study assessed workplace commuters Thus future studies should explore the

individuals mode choice with other trip aims Additionally including more policies especially pull

policies in favor of non-transit modes is reasonable to rigorously assess the model

It can be seen that the studied modes are affected differently by the TDM policies and their

interactions The impact of the interaction of policies appeared in the consideration of three out of

six studied modes and also in two out of three developed graphs This issue needs more attention

within the transportation demand management context and further studies should be performed

26

Design of this study is based on the consideration of all two-way interactions with some

degrees of deficiency in D-efficient policy Focusing on the significant interactions of this study is

recommended to perform other studies with more D-efficient values

Further understanding of the impacts of TDM policies might be provided in future research

by focusing on the amount of their synergy in prohibiting car usage

9 References

Arentze TA Hofman F amp Timmermans HJP (2004) Predicting multi-faceted activityndashtravel adjustmentstrategies in response to possible congestion pricing scenarios using an Internet-based stated adaptationexperiment Transport Policy 11 p31ndash41

Burris M amp Patil S (2009) Estimating the Benefits of Managed Lanes Technical Report Texas TexasTransportation Institute University Transportation Center for Mobility

Cao X amp Mokhtarian PL (2005) How do individuals adapt their personal travel A conceptual exploration ofthe consideation of travel-related strategies Transport Policy 12(3) pp199-206

de Palma A amp Lindsey R (2001) Transportation Supply and congestion In International Encyclopedia of theSocial and Behavioral Sciences 1st ed Elsevier p15882ndash15888

Eriksson L Garvill J amp Nordlund AM (2008) Acceptability of single and combined transport policy measuresThe importance of environmental and policy specific beliefs Transportation Research Part A 42 p1117ndash1128

Eriksson L Nordlund AM amp Garvill J (2010) Expected car use reduction in response to structural traveldemand management measures Transportation Research Part F 13 p329ndash342

Golob TF amp Hensher DA (2007) The Trip chaining Activity of Sydney Residents A Cross-Section Assessmentby Age Group with a Focus on Seniors Journal of Transport Geography 15 p298ndash312

Hensher DA amp King J (2001) Parking demand and responsiveness to supply pricing and location in theSydney central business district Transportation Research A 35 pp177-96

Hensher DA amp Rose JM (2007) Development of commuter and non-commuter mode choice models for theassessment of new public transport infrastructure projects A case study Transportation Research Part A41 p428ndash443

Hensher DA Rose JM amp Greene WH (2005) Applied Choice Analysis A Primer New York CambridgeUniversity Press

Iranian Center of Statistics (ICS) (2009) Information of Iranian States [Online] Available at HYPERLINK httpwwwamarorgirUploadModulesContentsasset16tehrantehparthtmlhttpwwwamarorgirUploadModulesContentsasset16tehrantehparthtml [Accessed 13 November2009]

Kingham S Dickinson J amp Copsey S (2001) Travelling to work Will people move out of their cars TransportPolicy 8 pp151-60

Kuhfeld WF (2009) Marketing Research Methods in SAS Experimental Design Choice Conjoint andGraphical Techniques SAS 92 Edition NC USA SAS Institute Inc Cary Available on the Webwwwsupportsascomtechsuptnotetnote_stathtmlmarket

Litman T (2003) The online TDM Encyclopedia mobility management information gateway Transport Policy10 pp245-49

Litman T (2010) Online TDM Encyclopedia [Online] Available at HYPERLINK httpwwwvtpiorghttpwwwvtpiorg [Accessed 21 Augest 2010]

5

actions (other actions) and assessed the role of policies coerciveness on driver mode change by

assuming that the collective policies are more coercive than the non-collective policies By focusing

on a smaller community and introducing a push policy (fuel pricing) in addition to some pull policies

Kingham et al examined the travel behavior of two companies employees (Kingham et al 2001)

through a survey They studied each employees perception of hisher mode choice during the trip

to work and investigated the potential of transferring car trips to other modes in the presence of

studied policies They also examined the importance of policies that would encourage employees to

use public transit or car-share to travel to work

Although the above studies focused on individualsrsquo perceptions of car-use for daily trips by adopting

some assumptions on the mode choice model of Shiraz Iran Zareii examined the results of

implementing five push policies in the city in terms of total travel time and amount of CO2 emission

(Zareii 2003) Because this study determines car usage by transferring the imposed cost of each

policy to extra time (calculated via the individuals time-value) push policies were of interest

Loukopoulos et al attempted to obtain quantitative estimates of the size of car usage reduction

goals and frequency of implementation of adaptation alternatives (Loukopoulos et al 2004) They

assessed two push policies and a pull policy for different trip aims and examined the cost-

minimization principle in relation to five adaptation alternatives OFallon et al explored the

potential effect of 11 policies on the respondents decision to choose to drive a car to work or school

during the morning peak period in three cities of New Zealand through a stated preferences survey

(OFallon et al 2004) They also reported the marginal effect of each of the studied policies and

recommended a study with fewer policies to explore the possible impacts of combinations of

specific policies Washbrook et al examined the role of seven policies on mode choice (Washbrook

et al 2006) Although the design of this study focused on the policiesrsquo main effects the results were

used to estimate commuter response to various policy combinations of charges and incentives

6

Because the aims of the above studies were to look for the best studied policy by assessing its

impact on car usage they did not deal with the effects of simultaneous implementation of TDM

policies Pendyala et al assessed five TDM policies by adopting an activity-based micro-simulation

model system (AMOS) to simulate changes in individual travel patterns (Pendyala et al 1997) In

their survey they also assessed combinations of specific policies in four transportation control

management scenarios and determined the possible impacts in those scenarios Thorpe et al

presented the individuals attitudinal responses to three push and one pull TDM policies in two case-

study cities in the UK Cambridge and Newcastle (Thorpe et al 2000) They examined the

relationship between the perceived effectiveness and public acceptance of alternative TDM policies

and showed that the public acceptance order of generic TDM policies is improving public transit

road-user charging zone-access controls and increased parking charge This study concluded that

there was evidence of interaction effects between levels of public acceptance of TDM policies when

considered separately and in combination with other policies Further these effects could be

investigated more rigorously with a stated preferences experimental design of alternative TDM

packages which allow the investigation of both main and interaction effects1

Eriksson et al examined the acceptability of one push policy (raised tax on fuel) and two pull policies

(improved public transport and subsidized renewable energy) individually and as packages

combining one push and one pull policy (Eriksson et al 2008) By proposing a model of factors

predicting acceptability of TDM policies they concluded that while the pull policies are perceived to

be effective fair and acceptable the push policy and the packages are perceived to be ineffective

unfair and unacceptable By removing one of the pull policies (ie subsidizing renewable energy)

these authors further assessed the expected car usage reduction in response to other two policies

(Eriksson et al 2010) By focusing on improved public transport raised tax on fuel and their

combination as a package the results showed that the combination was more effective than the

1 In a few studies in choice modeling researchers also examined the second order interactions of attributes in the models(eg (Mogas et al 2006))

7

individual policies Vieira et al explored the concept of multi-instrumentality as a procedure of

policy integration and implementation whereby a systematic search for complementary policies was

sought when planning and designing one (or several) core policy(s) aiming to fulfill one particular

policy more effectively (Vieira et al 2007) They defined criteria to assess the TDM policies and

analyzed four improvement mechanisms in each pair of policies Based on meta-studies they

defined the potential improvement between different types of policies By defining synergy concept

as a benefit of integration May et al reviewed a number of examples to assess the concept and

found little evidence of synergy in outcome indicators (May et al 2006)

Based on the above discussion assessing individual behavioral response to more than one TDM

policy is an interesting issue within the TDM context The following three issues are addressed in this

paper developing a model to investigate the role of TDM policies in commuters mode choice

exploring the role of effective parameters on the consideration of each mode of travel and

suggesting a method to determine the results of implementing two TDM policies simultaneously In

this paper the stated preferences approach is used to model the car users mode choice using the

design of experiments principles

3 Stated preferences

The five policies selected for the city of Tehran consisted of three push and two pull policies The

policies were increasing parking cost increasing fuel cost cordon pricing into an odd-even zone2

transit (bus or subway) time reduction and transit access improvement The latter two were

described by setting measures in favor of the public transit vehicles in streets and intersections

decreasing the time of boarding and alighting at the stations and increasing the number of transit

lines and stops in the city

2 This zone explored in the next section

8

Parking costs fuel costs and public transit time policies are designated with three levels and cordon

price and public access time are designed with two levels Table 1 shows the policies and their levels

All push policies had fixed values for their levels for pull policies because there were variations in

the transit time and transit access time for individuals proportional values of the current state were

used which is different for each individual The term no change in Table 1 refers to the current value

of a policy that each individual already experiences The mean values are also presented in Table 1

for a better description of current state

In preparing a questionnaire for the stated preferences part the design of experiments approach

was adopted Full factorial design is the most general type of design in this approach which

introduces all combinations of all levels in the modeling process In other words full factorial design

produces 108 possible choice sets (33322) This design allows the investigation of all

interactions as well as the main effects in the model On the one hand fewer choice sets are

available when ignoring the effects of higher-order interactions and on the other hand these

interactions have a negligible role in the variance (Louviere et al 2000 Hensher et al 2005) thus

fractional factorial design methods have been proposed

Table 1- Policies and their levels

Measure Type Numberof levels

Description of levels Mean Value

Increasing parking cost Push 3 No change 4000 7000 Rials3 h 71 RialshNACordon pricing Push 2 25000 50000 Rialsday

Increasing fuel cost Push 3 No change 3000 5000 Rialsliter 1470 RialsLiterTransit time reduction Pull 3 No change 15 30 percent shortage 385 minTransit access improvement Pull 2 No change 25 percent shortage 11 min

Efficient design a type of fractional factorial design was used in the study and a design with 895

efficiency was adopted which allows assessing all two-way interactions of policies as well as the

3 10000 Rials are almost equal to 1 US dollar

9

main effects with only 36 choice sets4 (See (Rose amp Bliemer 2009) or (Kuhfeld 2009) for more

details on efficient design) To avoid a time-consuming questionnaire 36 choice sets (scenarios)

were randomly ordered and divided into six separate questionnaire types coded as 1 to 6 Each of

the questionnaires had six scenarios and each scenario consisted of five policies

4 Survey

Two push policies are currently being implemented in the city of Tehran The first is car-free

planning in the CBD area of the city and the second one is an odd-even scheme based on the last

digit of car plates that attempt to enter a zone which is about three times larger than and includes

the CBD area Based on their occupation a few people can drive to the CBD area with a license

called permission A stated preferences survey was assigned for the morning car commuters to the

odd-even zone but they were asked to ignore these two policies to find the accurate sensitivity of

individuals to the study policies The odd-even zone is selected as study area for the two following

reasons 1) because of odd-even control respondents are familiar with the fringes and they can

better imagine the entrance pricing area and 2) respondents are familiar with the limits that they

face half of the week and are thus aware of the alternative existing modes Compared to the CBD

area this zone covers more car commuters and the entrance restriction is more imaginable for this

zone than the former one Respondents were interviewed face-to-face in their workplaces midway

through the year 2009 The interviews were enhanced with a special card to better define the

scenarios

For this study 2196 scenario observations from 366 individuals were adopted The sample included

308 men (ie 841) and 58 women (ie 159) The figures are close to the employment

percentages in the city according to the Iranian Center of Statistics (ICS) This source indicates that

825 of Tehran employees are men and 175 are women (Iranian Center of Statistics (ICS) 2009)

Because this study focuses on car-using commuters comparisons between the sample and city data

4 Efficient design is also adopted in other studies such as managed lanes (Burris amp Patil 2009)

10

especially regarding educational distribution were impossible Table 2 presents demographics of the

sample

Table 2- Demographics gender marital household (HH) size employee type age HH employee(s)

Amount Percent

Gender Male 308 841Female 58 159

Marital Single 100 273Married 266 727

HH Size 1 4 112 86 2353 129 3524 90 2465 42 1156+ 15 41

Age 18~29 122 33330~39 146 39940~49 58 15950~59 32 8760+ 8 22

HH employee(s) 1 156 4262 159 4343 41 1124+ 10 27

The first part of the questionnaire is dedicated to gathering the occupation state home and job

locations the distance between these locations round-trip time (from home to workplace and then

workplace to home) and all car trip characteristics in the previous day or the day before it based on

plate number It was necessary that the respondents drive hisher car in the day studied to complete

the trip diary portion of the questionnaire5 The general reasons for car usage and the scenarios

formed the next portion In each scenario every respondent was asked the question How would

you travel to the workplace if all of these changes were in place on the day studied For example

one may have to pay 4000 Rialsh for parking 50000 Rials per entrance to the cordon the same

amount in transit access and fuel cost and a 15 percent decrease in transit time simultaneously

Depending on individual responses six main options were distinguished6 These choices were still

5 In designing the questionnaire the general form of questionnaire which has mentioned in OFallons study was adopted6 In the pre-test survey 14 modes is distinguished

11

drive a car (C) walk to the station and catch public transit (WampR) drive to a public station and catch

public transit (DampR) ride a motorcycle (MC) catch a taxi7 (T) and catch a taxi by phone (T_T) DampR is

somewhat different than the more familiar ldquoPark amp Riderdquo In fact in the fringes of the odd-even

zone there were no specialized parking lots dedicated to this purpose and commuters considered

Drive amp Ride because they were not allowed to pass the fringes

After each scenario if the respondent changed hisher mode the reason(s) for the change were

asked It could be a sole policy or a bundle of them Furthermore travel-related information was

sought These data were not part of the stated choice but they might have important influences on

individual choices These data consisted of car dependency (need to drive someone or move freight

in the trip) parking place type and average weekly parking costs car and motorcycle ownership and

number of household driving licenses

Depending on the individuals activity in that day three types of activity patterns were detected

Pattern 1 described individuals who had no stop in their commute Pattern 2 was for individuals who

had at least one stop on their way to or from work and pattern 3 was for the individuals who went

to another workplace in their daily activities

Finally for the sake of data generalization and the examination of household characteristics gender

age and household type employment status and education level were also asked

5 Mode choice model

In order to detect the policies that affect individual mode choice the logit modeling approach was

adopted In this model one can determine if the interaction of two policies affects the mode choice

In the calibration step 152 variables were defined and their effects on consideration of each mode

were examined

7 Taxis in Iran are somewhat different than taxis in other countries of the world In fact taxis in Iran are not hiring by oneperson or a group of people at a time Taxis allow passengers to board or alight along their path with respect to theircapacity In other word this mode is functioning similar to transit vehicles but the stops are not predefined

12

51 Model structure

Initially a multinomial logit (MNL) model is developed (Figure 1a) By selecting a number of tree

structures based on recognizing differences in the variances associated with unobserved influences

we find that the greatest similarity in variance profiles is associated with public transport modes as

opposed to non-public modes (Figure 1b) This structure has two nests one including Car (C) and

Motorcycle (MC) as private modes and the other including Walk and Ride (WampR) Drive and Ride

(DampR) Taxi (T) and Tel-taxi (T_T) as non-private modes The result of this nested logit (NL) model is

shown in Table 3

Although it is not a statistically significant improvement overall on the MNL model the statistically

significant inclusive value8 (IV) of 0889 for non-public modes relative to the fixed parameter value of

10 for public modes suggests that there is a structural advantage in selecting the NL specification

The normal test of a statistically significant difference between NL and MNL is an IV parameter

relative to 10 calculated using a Wald-test via equation 1

)1(Wald-test = (IVparameter ndash 1)std error

a The MNL structure

b Final nested structure

Figure 1- Model structure

8 Also called scale parameter

Alternatives

MCCar T_TT WampR DampR

Alternatives

MCCar T_TT

Public

WampR DampR

Private

13

We have (0889-1)2508 =-075 which would be rejected at the usual acceptable significance levels

This suggests that the NL model could be collapsed into an MNL form

Table 3- Nested logit (NL) model resultValueParameter

0889IV (Private)1000IV (nPrivate)

-2668335L( )-4057684L(0)

0342sup2

After the calibration process the variables that were statistically significant were identified and are

presented in Table 4 Table 5 presents the final model of the study with a goodness of fit of 031 for

the six studied modes For a general review of the model calibration results the effective factors can

be grouped under the following three categories TDM policy characteristics commuting trip

characteristics and household socio-economic characteristics which are all treated as alternative-

specific variables

52 Model results

Car (C)

It is expected that push policies impel car-drivers to choose other modes Table 5 shows that cordon

pricing and increase in parking cost cause individuals to choose not to use their car This is in line

with other studies suggesting that these policies are effective to discourage car usage (Hensher amp

Rose 2007 OFallon et al 2004) In addition the interaction between the policies of fuel cost

increase and increase in parking cost shows similar car usage discourage effect Because fuel cost is

related to the distance between home and work locations and parking cost is related to work time

the time that an individual spends out of the home is negatively affected by hisher likelihood to use

a car

14

Table 4 - Definition of the significant variables

AbbreviationVariableTransportation demand management measures

Measures

ParkingParking cost increase Rials per hour

CordonCordon price Rials per entranceAccessTransit access time shortage percent

Interaction of push measures

ParkampFuelParking cost and fuel cost simultaneous effectsCordonampFuelCordon pricing and fuel cost simultaneous effects

Interaction of pull measures

PT_timeampaccessPT time reduction and access improvement simultaneous effectsCommuting trip characteristics

Trip distanceDistance between home and workplaceTrip timeTravel time between home and workplace

Exp FuelLikelihood of unsubsidized fuel use (self-reported on a Likert scale)NtripsNumber of daily tripsPattern2Commuting with 1+ stop(s) in go or return

Pattern3Commuting with 2 workplacesFirst trip timeStart time of first trip

PnocarwkLikelihood of going to work in absence of that car (self-reported)PTnwaccNon-walk access to transit (yes=1)First NaccoNumber of passengers in first trip

PassengerAny passenger on that day (yes=1)Park_paymentParking payment in last weekNhempfullNumber of full employees in HH

CardependencyBoardalight a passenger or move freight in the trip (yes=1)D car ownBe the owner of the used vehicle (yes=1)

Car accCar accessibility in household (number of cars to number of HH driving licenses ratio)NmotorcycleNumber of motorcycles owned by HHD home placeHome Location is in study area (yes=1)

PermissionPermission to enter to study area (yes=1)ComfortI use my car because it is comfortablePoor_PTI use my car because transit is not good

HH socio-economic characteristics

FemaleGender (Female=1)Age lt30Age younger than 30 (yes=1)Age 30_39Age between 30 to 39 (yes=1)Job_durationNumber of years that individual has been at hisher job

Emp_fullFull-time employee (yes=1)Edu BSDegree of education is BSc (yes=1)Edu BS+Degree of education is higher than BSc(yes=1)

D childlt=18Child younger than 18 in HH (yes=1)

15

Table 5 ndash The mode choice model

Tel-Taxi(T_T)

Motorcycle(MC)

Drive amp Ride(DampR)

Taxi (T)Walk amp Ride(WampR)

Car (C)Mode

Variable-471756-37067-147911Constant

Transportation demand management measure variables00019-00045Cordon

-000072Parking-004308Access

-28443D-05Parkampfuel-32475D-06Cordonampfuel

00029Pt_timeampaccess

Commuting trip characteristics-04709Trip distance

-02163-00831Trip time-96755163655Exp fuel-16253Ntrips

-114779Pattern2-71008Pattern3

00282-00270First trip time-02439-01549Pnocarwk

-11322992883-32765PTnwacc-133701First Nacco

-7778-73782Accompany-00049000010Park_payment

201646195554Nhempfull-160144ComfortCar1

-206142DependencyCar1-16101883385-121224DependencyCar2

42176Poor_PTCar1-24988Poor_PTCar2

- -27221D car own70960-39136Car acc

1 -71112-156123Nmotorcycle-1436322762D home place

2 78826Permission

HH socio-economic characteristics149490Female

297584-24548Agelt30-136490Age30_39

079430366303585Job_duration-108743Emp_full-203468-64900Edu BS

10932856687-4499984445Edu BS+102271D childlt=18

-2677366L( )-3849556L(0)0305sup2

112127178592580607N

Note = Positive significance at 1 5 10 level

As expected individuals with higher income are more likely to use their car This is indicated in the

model by the positive signs of individuals who use fuel with fixed (unsubsidized) cost and individuals

16

who pay more in parking charges in the previous week of study Negative sign of Pnocarwk variable

shows that the commuters who stated that their commute depends on car availability are more

likely to use their car Individuals in households with more full-time employees are more likely to use

their car which may be the result of higher household income Not surprisingly commuters who

have permission are more likely to maintain car usage Among the household socio-economic

parameters greater job experience (Job_duration) and higher graduate levels (EduBS+) increase the

probability of car usage

Public transit accessed by walking (WampR)

Access time to transit negatively impacts WampR choice which is expected This result is similar to

findings for the city of Sydney (Hensher amp Rose 2007) The negative coefficient of first trip time

indicates that individuals are more likely to use WampR in the early morning This result seems to

reflect the better weather for walking and faster speed of WampR mode early in the morning

Obviously individuals who are not able to access transit stations via walking (PTnwacc) are less likely

to consider this mode Furthermore serving passengers on daily trips is also a deterrent to using

WampR

Initially assessing the individuals who stated that their car usage is due to poor public transit service

(Poor_PT) led to an unexpected result in favor of considering WampR By introducing to this variable

the number of household cars as a proxy for household income (Poor_PTCar1) the model shows

that of the previously mentioned individuals those who have lower income are the ones who have

to consider WampR The result is understandable as these individuals may have no alternative when

they have to change their mode (they also are not likely to consider other modes) Individuals with

higher levels of income who have to use their car during before or after work (Dependencycar1+)

are not likely to use WampR

The greater the number of motorcycles in a household the less likely commuters is to consider

WampR There appears to be a competition between motorcycle and PT for access to the city center

17

Better PT services in the center of the city in terms of coverage and frequency increases the

likelihood that its residents will consider WampR This is verified by the positive sign of the

D_home_place variable Commuters with greater job experience (Job_duration) in their workplace

are more likely to use this mode Although individuals with higher levels of education are not likely

to use WampR as education level increases avoidance of WampR decreases

Taxi (T)

Table 5 shows that none of the studied policies are significant in considering taxi usage It seems that

taxi usage considering its function in Iran as a non-private and non-public mode of transport is not

affected by pull or push policies A negative sign for taxi travel time indicates that individuals are not

likely to use this mode for longer trips This seems reasonable given that longer trips are more

expensive Commuters who are more likely to use fuel with no subsidy are not likely to use taxis As

mentioned before they prefer to use their car A higher number of trips in a day are also a deterrent

to considering taxi usage which may be due to increased cost for more trips Results show that an

individual with more daily trips avoids using taxis Commuters who are employed in more than one

workplace (Pattern 3) are not likely to use taxis This may be due to the fact that they have a lower

level of income which forces them to dedicate more time on the job

Initial results showed that individuals who stated that their car usage is due to poor public transit

service (Poor_PT) are not likely to use taxis This result was far from our expectations By introducing

to this variable the number of household cars as a proxy for household income the model shows

that the previously mentioned individuals who have higher income (Poor_PTCar1+) are the ones

who are not likely to consider taxis Furthermore because such individuals are not considering any

other modes they may treat taxi usage as a kind of PT mode with poor service

As expected greater access to cars in a household (Car_acc) lessens the likelihood of considering

taxis as an alternative Furthermore individuals in households with more motorcycle ownership are

less likely to consider taxis It seems like there is a competition among motorcycles and taxis for

18

access to the city center Younger commuters are less likely to use taxis and individuals with at least

master degrees do consider this mode in addition to their car

Public transit accessed by Drive (DampR)

This mode is affected by the simultaneous interaction of transit time and transit access

(PT_TimeampAccess) which is reflected in the fact that individuals prefer to use this mode for longer

trips Comparing this mode and WampR the first trip start time affects the consideration of this mode

differently Later morning commuters prefer to use their car to access PT modes Such commuters

may have higher income levels or managerial jobs Obviously individuals who are not able to access

PT stations by walking (PTnwacc) are likely to use DampR Serving passengers in daily trips is also a

deterrent in considering this mode which is similar to WampR but with a lower coefficient

Commuters with higher income levels who depend on their car during before or after work

(Dependencycar1+) are likely to use DampR Individuals who use their own car are less likely to use

this mode which is unexpected As a city center develops better PT network coverage and residents

have smaller distances to their workplaces they are unlikely to use DampR This is proven in the model

by a negative sign for D_home_place

Motorcycle (MC)

Increasing fuel cost and cordon pricing simultaneously discourage motorcycle usages Although fuel

cost is expected to reduce motorcycle usage to some extent its combined effect with cordon pricing

also reduces motorcycle usage However this variable is not as strong as other policy variables

=10)

Of the studied modes motorcycle usage is affected by the most commuting variables This may be

due to the fact that this mode is not common Commuting distance has a negative effect on

motorcycle usage which is expected It is worth noting that trip distance appears only in this mode

which may be a reflection of the role of distance in regards to the safety risk in considering this

19

mode Commuters with more stops to serve passengers while commuting (Pattern 2) are not likely

to use this mode which may be due to the poor passenger service of this mode

Individuals who state that commuting is independent of the mode (Pnocarwk) are not likely to use

MC By looking at the (First_Nacco) negative sign this could stem from the fact that the more

passengers there are on the first trip the less likely individuals are to consider MC Regarding the

low capacity of MC and its safety concerns such commuters avoid using this mode Commuters who

pay more parking charges (Park_payment) are less likely to use MC which is expected Individuals

who are dependent on their car during before or after their work time are not likely to use MC

even if they have lower levels of income (DependencyCar1) Individuals who use their own car

(D_car_own) are less likely to use this mode As expected individuals who live in households with

more motorcycle ownership are more likely to use this mode The positive sign of (Permission)

indicates that commuters who have permission to enter the study area do consider MC Because

such commuters generally provide that permission just for car usage this result is unexpected

As with commute variables of all the studied modes MC is affected by the greatest number of

socio-economic variables As expected young commuters (Agelt30) are more likely to use this mode

Commuters with Bachelor of Science degree are less likely to use this mode among others Full time

employees (Emp_full) are less likely to consider MC whereas commuters with more experience in

their jobs prefer to use it Results show that individuals who live in a household with children

younger than 18 are more likely to consider using a car

Tel-Taxi (T_T)

Results show that cordon pricing causes higher probability of using T_T In fact individuals who use

T_T as a mode with similar level of service as cars9 are more willing to pay the cost and make use of

the mode It is worth noting that the effect of cordon pricing in pushing commuters from car usage

9 As this mode does not have driving stress and parking search time in some cases it may have more amount of utility thana car does

20

(000045) is greater than its effect on pulling them to Tel-taxi (000019) This is because of the

possibility of considering other non-car modes

Because consideration of this mode is a function of its operation travel time (Trip_time) appears as

a deterrent in this mode utility function Table 5 shows that individuals are more sensitive to the trip

time when using T_T mode versus taxi which is expected due to their relative costs

The greater the number of full time employees in a family (Nhempfull) the higher the probability of

considering T_T by its commuters which may be due to the higher income level of these

households This is verified by the greater likelihood of using T_T rather than taxis by such

commuters Individuals with higher levels of income who depend on their car during before or after

work time are less likely to use T_T Commuters with lower income levels who state that they use

their car for the sake of comfort (Comfortcar1) are less likely to use T_T which may be due to its

cost Although such individuals do not consider any other modes they specifically avoid T_T Greater

access to cars in a household leads to greater likelihood of T_T usage which could be due to the

higher income level of a household As mentioned before such individuals even avoid taxis

Females who drive to their workplace are more likely to use T_T It seems like this part of society

considers this mode when desiring to avoid the difficulties of driving Younger commuters are less

likely to use T_T and individuals between 30 and 39 years of age are specifically avoiding this mode

Results show that university graduated commuters are more likely to use this mode

6 Marginal effects

To explore the effects of each policy on mode choice and to answer the second issue raised at the

beginning of this paper the marginal effects approach can be adopted Although the coefficients of

the models utility functions show the drivers behavior when facing one or more policies the

marginal effects of policies or their interactions may appropriately show the results of their

implementation More specifically the marginal effect for this study is interpreted as the change in

21

probability given a unit change in a variable ceteris paribus In this section the variable is defined as

a specific policy or a specific interaction of two policies Table 6 presents the marginal effects of the

studied policies and their interactions with mode choice The results are shown in the form of trip

percentages transferred away from the car to the studied modes and the probability-weighted

sample enumeration approach is adopted to find the values It is worth noting that this table is fully

compatible with Table 5 but the marginal effects that were less significant than 90 percent have

been removed

Table 6 - Marginal effects of policies (percent)

Tel-Taxi(TT)

Motorcycle(MC)

Drive ampRide(DampR)

Taxi (T)Walk ampRide(WampR)

Car (C)Mode

Variable-000088Cordon-000140Parking

-09069Access-0000001ParkampFuel

00040PT_TimeampAccess

Table 5 shows that a 1 Rial increase in cordon pricing causes a 000088 percent decrease in car

usage Furthermore a 1 Rial per hour increase in parking cost decreases the car usage probability to

00014 percent By assuming 8 hours for the average parking duration the daily marginal value of

parking cost converts to 000018 percent These values show that cordon pricing is more effective in

forcing individuals not to use their car than increasing parking cost with the same value Results also

show that a 1 Rial increase in parking cost and fuel cost simultaneously decreases the probability of

choosing car usage by 0000001 percent Table 5 also shows that a 1 minute decrease in transit

access time would result in a 09 percent increase in probability of choosing this mode It also shows

that each 1 minute decrease in transit travel time and transit access leads to a 0004 decrease in the

probability of choosing the DampR mode

22

The above results show that just one of the policy interactions is not significantly different from

none (ie CordonampFuel) and push policies have a significant role in impelling car users to consider

other modes

7 Planning

To demonstrate the application of the model some of its planning aspects for the city of Tehran are

presented here In fact the model is able to assess the effect of all studied variables which were

grouped into three tiers on commuters mode choice In this section we assess the effect of TDM

policies by adopting their pair-wise combinations for the city Table 6 shows that pull policies in this

study did not impose a major effect on car usage Because the average access time of the sample

was 11 minutes improving it by 50 percent (about 5 minutes) increases the probability of WampR to

509069100 which is equal to a 048 percent increase in considering this mode (and decrease in all

other modes) On the other hand decreasing transit time about 50 percent (about 19 minutes based

on sample average) and similarly improving access time decreases the probability of considering

DampR to 0004195100 which is equal to a 038 percent decrease in considering this mode (and

increase in all other modes) Therefore in this section we focused on the push policies which

appear in the utility function of car mode It is worth noting that in assessing each combination of

policies all other variables were assumed to be fixed as the current state

71 Parking pricing and cordon pricing

The average hourly parking charge of individuals in the sample is 71 Rials This charge is 3000 Rials

per hour in the curb lane on some streets about 2000 Rials per hour in parking lots and free in

alleys and other streets The average amount shows that most of the commuters benefited from

free parking spaces in alleys and non-controlled streets Figure 2 depicts the result of implementing

this policy and the cordon pricing policy simultaneously In line with marginal effects it can be seen

that in implementing each policy separately cordon pricing is more effective than parking pricing in

the planning range It is worth noting that commuters in this study are less sensitive to parking

23

pricing than cordon pricing One reason may be that they thought in any case they would find free

parking space Figure 2 shows that the effect of each policy depends on the level of the other policy

In other words the effect of the simultaneous implementation of two policies is not equal to the

sum of their separate individual effects

Figure 2- Effect of implementing parking pricing and cordon pricing

72 Parking pricing and increasing fuel cost

Figure 3 shows the simultaneous effects of parking pricing and increasing fuel cost policies Although

both of these policies are not effective separately their simultaneous effect is significant It can be

seen that their synergy appears in the higher levels of both policies

Figure 3 - Effect of implementing parking pricing and increasing fuel cost

24

73 Cordon pricing and increasing fuel cost

The effect of implementing cordon pricing and increasing fuel cost is shown in Figure 4 It can be

seen that these two policies have no interaction effect in decreasing car usage This result is verified

by this interactions insignificant marginal value in Table 6

Figure 4- Effect of implementing cordon pricing and increasing fuel cost

8 Conclusions

This study examined the role of TDM policies in individual mode choice for work trips in the city of

Tehran Five policies including increasing parking cost increasing fuel cost cordon pricing in the odd-

even zone of the city transit time reduction and transit access improvement were investigated Of

the five policies the former three were push policies and the latter two were pull policies The

design of experiments approach was used to design the questionnaire that would capture the stated

preferences of car commuters

A number of nested structures based on recognizing differences in the variances associated

with unobserved influences were examined and finally the MNL as a superior model for this study

was developed

25

The results show that for car users who faced TDM policies all of the five studied TDM

policies were factors in the mode choice process The model shows that in considering car usage

push policies play a main role and pull policies do not

Results also show that the interactions of policies are also significant The model shows that

the interaction of parking cost policy and fuel cost policy is significant in prompting car users to

consider other modes It also shows that the interaction of cordon pricing and increase in fuel cost is

effective to discourage motorcycle usage Furthermore the interaction of transit time reduction

policy and access improvements policy is significant in considering DampR as a transportation mode to

workplace

Pull policies in the study were expected to attract individuals to transit modes (pull policies

of this study are related to transit modes) This was verified by the model

Results show that although taxi usage is usually treated as an alternative for urban trips it

was not directly affected by the studied policies This may be due to its special function in Tehran as

a non-public and non-private mode

The results also show that except for the interaction of cordon pricing and increase in fuel

cost other interaction effects have significant marginal effects on mode choice

Results of the model show that to implement a single policy cordon pricing is the most

effective in decreasing car usage

This study assessed workplace commuters Thus future studies should explore the

individuals mode choice with other trip aims Additionally including more policies especially pull

policies in favor of non-transit modes is reasonable to rigorously assess the model

It can be seen that the studied modes are affected differently by the TDM policies and their

interactions The impact of the interaction of policies appeared in the consideration of three out of

six studied modes and also in two out of three developed graphs This issue needs more attention

within the transportation demand management context and further studies should be performed

26

Design of this study is based on the consideration of all two-way interactions with some

degrees of deficiency in D-efficient policy Focusing on the significant interactions of this study is

recommended to perform other studies with more D-efficient values

Further understanding of the impacts of TDM policies might be provided in future research

by focusing on the amount of their synergy in prohibiting car usage

9 References

Arentze TA Hofman F amp Timmermans HJP (2004) Predicting multi-faceted activityndashtravel adjustmentstrategies in response to possible congestion pricing scenarios using an Internet-based stated adaptationexperiment Transport Policy 11 p31ndash41

Burris M amp Patil S (2009) Estimating the Benefits of Managed Lanes Technical Report Texas TexasTransportation Institute University Transportation Center for Mobility

Cao X amp Mokhtarian PL (2005) How do individuals adapt their personal travel A conceptual exploration ofthe consideation of travel-related strategies Transport Policy 12(3) pp199-206

de Palma A amp Lindsey R (2001) Transportation Supply and congestion In International Encyclopedia of theSocial and Behavioral Sciences 1st ed Elsevier p15882ndash15888

Eriksson L Garvill J amp Nordlund AM (2008) Acceptability of single and combined transport policy measuresThe importance of environmental and policy specific beliefs Transportation Research Part A 42 p1117ndash1128

Eriksson L Nordlund AM amp Garvill J (2010) Expected car use reduction in response to structural traveldemand management measures Transportation Research Part F 13 p329ndash342

Golob TF amp Hensher DA (2007) The Trip chaining Activity of Sydney Residents A Cross-Section Assessmentby Age Group with a Focus on Seniors Journal of Transport Geography 15 p298ndash312

Hensher DA amp King J (2001) Parking demand and responsiveness to supply pricing and location in theSydney central business district Transportation Research A 35 pp177-96

Hensher DA amp Rose JM (2007) Development of commuter and non-commuter mode choice models for theassessment of new public transport infrastructure projects A case study Transportation Research Part A41 p428ndash443

Hensher DA Rose JM amp Greene WH (2005) Applied Choice Analysis A Primer New York CambridgeUniversity Press

Iranian Center of Statistics (ICS) (2009) Information of Iranian States [Online] Available at HYPERLINK httpwwwamarorgirUploadModulesContentsasset16tehrantehparthtmlhttpwwwamarorgirUploadModulesContentsasset16tehrantehparthtml [Accessed 13 November2009]

Kingham S Dickinson J amp Copsey S (2001) Travelling to work Will people move out of their cars TransportPolicy 8 pp151-60

Kuhfeld WF (2009) Marketing Research Methods in SAS Experimental Design Choice Conjoint andGraphical Techniques SAS 92 Edition NC USA SAS Institute Inc Cary Available on the Webwwwsupportsascomtechsuptnotetnote_stathtmlmarket

Litman T (2003) The online TDM Encyclopedia mobility management information gateway Transport Policy10 pp245-49

Litman T (2010) Online TDM Encyclopedia [Online] Available at HYPERLINK httpwwwvtpiorghttpwwwvtpiorg [Accessed 21 Augest 2010]

6

Because the aims of the above studies were to look for the best studied policy by assessing its

impact on car usage they did not deal with the effects of simultaneous implementation of TDM

policies Pendyala et al assessed five TDM policies by adopting an activity-based micro-simulation

model system (AMOS) to simulate changes in individual travel patterns (Pendyala et al 1997) In

their survey they also assessed combinations of specific policies in four transportation control

management scenarios and determined the possible impacts in those scenarios Thorpe et al

presented the individuals attitudinal responses to three push and one pull TDM policies in two case-

study cities in the UK Cambridge and Newcastle (Thorpe et al 2000) They examined the

relationship between the perceived effectiveness and public acceptance of alternative TDM policies

and showed that the public acceptance order of generic TDM policies is improving public transit

road-user charging zone-access controls and increased parking charge This study concluded that

there was evidence of interaction effects between levels of public acceptance of TDM policies when

considered separately and in combination with other policies Further these effects could be

investigated more rigorously with a stated preferences experimental design of alternative TDM

packages which allow the investigation of both main and interaction effects1

Eriksson et al examined the acceptability of one push policy (raised tax on fuel) and two pull policies

(improved public transport and subsidized renewable energy) individually and as packages

combining one push and one pull policy (Eriksson et al 2008) By proposing a model of factors

predicting acceptability of TDM policies they concluded that while the pull policies are perceived to

be effective fair and acceptable the push policy and the packages are perceived to be ineffective

unfair and unacceptable By removing one of the pull policies (ie subsidizing renewable energy)

these authors further assessed the expected car usage reduction in response to other two policies

(Eriksson et al 2010) By focusing on improved public transport raised tax on fuel and their

combination as a package the results showed that the combination was more effective than the

1 In a few studies in choice modeling researchers also examined the second order interactions of attributes in the models(eg (Mogas et al 2006))

7

individual policies Vieira et al explored the concept of multi-instrumentality as a procedure of

policy integration and implementation whereby a systematic search for complementary policies was

sought when planning and designing one (or several) core policy(s) aiming to fulfill one particular

policy more effectively (Vieira et al 2007) They defined criteria to assess the TDM policies and

analyzed four improvement mechanisms in each pair of policies Based on meta-studies they

defined the potential improvement between different types of policies By defining synergy concept

as a benefit of integration May et al reviewed a number of examples to assess the concept and

found little evidence of synergy in outcome indicators (May et al 2006)

Based on the above discussion assessing individual behavioral response to more than one TDM

policy is an interesting issue within the TDM context The following three issues are addressed in this

paper developing a model to investigate the role of TDM policies in commuters mode choice

exploring the role of effective parameters on the consideration of each mode of travel and

suggesting a method to determine the results of implementing two TDM policies simultaneously In

this paper the stated preferences approach is used to model the car users mode choice using the

design of experiments principles

3 Stated preferences

The five policies selected for the city of Tehran consisted of three push and two pull policies The

policies were increasing parking cost increasing fuel cost cordon pricing into an odd-even zone2

transit (bus or subway) time reduction and transit access improvement The latter two were

described by setting measures in favor of the public transit vehicles in streets and intersections

decreasing the time of boarding and alighting at the stations and increasing the number of transit

lines and stops in the city

2 This zone explored in the next section

8

Parking costs fuel costs and public transit time policies are designated with three levels and cordon

price and public access time are designed with two levels Table 1 shows the policies and their levels

All push policies had fixed values for their levels for pull policies because there were variations in

the transit time and transit access time for individuals proportional values of the current state were

used which is different for each individual The term no change in Table 1 refers to the current value

of a policy that each individual already experiences The mean values are also presented in Table 1

for a better description of current state

In preparing a questionnaire for the stated preferences part the design of experiments approach

was adopted Full factorial design is the most general type of design in this approach which

introduces all combinations of all levels in the modeling process In other words full factorial design

produces 108 possible choice sets (33322) This design allows the investigation of all

interactions as well as the main effects in the model On the one hand fewer choice sets are

available when ignoring the effects of higher-order interactions and on the other hand these

interactions have a negligible role in the variance (Louviere et al 2000 Hensher et al 2005) thus

fractional factorial design methods have been proposed

Table 1- Policies and their levels

Measure Type Numberof levels

Description of levels Mean Value

Increasing parking cost Push 3 No change 4000 7000 Rials3 h 71 RialshNACordon pricing Push 2 25000 50000 Rialsday

Increasing fuel cost Push 3 No change 3000 5000 Rialsliter 1470 RialsLiterTransit time reduction Pull 3 No change 15 30 percent shortage 385 minTransit access improvement Pull 2 No change 25 percent shortage 11 min

Efficient design a type of fractional factorial design was used in the study and a design with 895

efficiency was adopted which allows assessing all two-way interactions of policies as well as the

3 10000 Rials are almost equal to 1 US dollar

9

main effects with only 36 choice sets4 (See (Rose amp Bliemer 2009) or (Kuhfeld 2009) for more

details on efficient design) To avoid a time-consuming questionnaire 36 choice sets (scenarios)

were randomly ordered and divided into six separate questionnaire types coded as 1 to 6 Each of

the questionnaires had six scenarios and each scenario consisted of five policies

4 Survey

Two push policies are currently being implemented in the city of Tehran The first is car-free

planning in the CBD area of the city and the second one is an odd-even scheme based on the last

digit of car plates that attempt to enter a zone which is about three times larger than and includes

the CBD area Based on their occupation a few people can drive to the CBD area with a license

called permission A stated preferences survey was assigned for the morning car commuters to the

odd-even zone but they were asked to ignore these two policies to find the accurate sensitivity of

individuals to the study policies The odd-even zone is selected as study area for the two following

reasons 1) because of odd-even control respondents are familiar with the fringes and they can

better imagine the entrance pricing area and 2) respondents are familiar with the limits that they

face half of the week and are thus aware of the alternative existing modes Compared to the CBD

area this zone covers more car commuters and the entrance restriction is more imaginable for this

zone than the former one Respondents were interviewed face-to-face in their workplaces midway

through the year 2009 The interviews were enhanced with a special card to better define the

scenarios

For this study 2196 scenario observations from 366 individuals were adopted The sample included

308 men (ie 841) and 58 women (ie 159) The figures are close to the employment

percentages in the city according to the Iranian Center of Statistics (ICS) This source indicates that

825 of Tehran employees are men and 175 are women (Iranian Center of Statistics (ICS) 2009)

Because this study focuses on car-using commuters comparisons between the sample and city data

4 Efficient design is also adopted in other studies such as managed lanes (Burris amp Patil 2009)

10

especially regarding educational distribution were impossible Table 2 presents demographics of the

sample

Table 2- Demographics gender marital household (HH) size employee type age HH employee(s)

Amount Percent

Gender Male 308 841Female 58 159

Marital Single 100 273Married 266 727

HH Size 1 4 112 86 2353 129 3524 90 2465 42 1156+ 15 41

Age 18~29 122 33330~39 146 39940~49 58 15950~59 32 8760+ 8 22

HH employee(s) 1 156 4262 159 4343 41 1124+ 10 27

The first part of the questionnaire is dedicated to gathering the occupation state home and job

locations the distance between these locations round-trip time (from home to workplace and then

workplace to home) and all car trip characteristics in the previous day or the day before it based on

plate number It was necessary that the respondents drive hisher car in the day studied to complete

the trip diary portion of the questionnaire5 The general reasons for car usage and the scenarios

formed the next portion In each scenario every respondent was asked the question How would

you travel to the workplace if all of these changes were in place on the day studied For example

one may have to pay 4000 Rialsh for parking 50000 Rials per entrance to the cordon the same

amount in transit access and fuel cost and a 15 percent decrease in transit time simultaneously

Depending on individual responses six main options were distinguished6 These choices were still

5 In designing the questionnaire the general form of questionnaire which has mentioned in OFallons study was adopted6 In the pre-test survey 14 modes is distinguished

11

drive a car (C) walk to the station and catch public transit (WampR) drive to a public station and catch

public transit (DampR) ride a motorcycle (MC) catch a taxi7 (T) and catch a taxi by phone (T_T) DampR is

somewhat different than the more familiar ldquoPark amp Riderdquo In fact in the fringes of the odd-even

zone there were no specialized parking lots dedicated to this purpose and commuters considered

Drive amp Ride because they were not allowed to pass the fringes

After each scenario if the respondent changed hisher mode the reason(s) for the change were

asked It could be a sole policy or a bundle of them Furthermore travel-related information was

sought These data were not part of the stated choice but they might have important influences on

individual choices These data consisted of car dependency (need to drive someone or move freight

in the trip) parking place type and average weekly parking costs car and motorcycle ownership and

number of household driving licenses

Depending on the individuals activity in that day three types of activity patterns were detected

Pattern 1 described individuals who had no stop in their commute Pattern 2 was for individuals who

had at least one stop on their way to or from work and pattern 3 was for the individuals who went

to another workplace in their daily activities

Finally for the sake of data generalization and the examination of household characteristics gender

age and household type employment status and education level were also asked

5 Mode choice model

In order to detect the policies that affect individual mode choice the logit modeling approach was

adopted In this model one can determine if the interaction of two policies affects the mode choice

In the calibration step 152 variables were defined and their effects on consideration of each mode

were examined

7 Taxis in Iran are somewhat different than taxis in other countries of the world In fact taxis in Iran are not hiring by oneperson or a group of people at a time Taxis allow passengers to board or alight along their path with respect to theircapacity In other word this mode is functioning similar to transit vehicles but the stops are not predefined

12

51 Model structure

Initially a multinomial logit (MNL) model is developed (Figure 1a) By selecting a number of tree

structures based on recognizing differences in the variances associated with unobserved influences

we find that the greatest similarity in variance profiles is associated with public transport modes as

opposed to non-public modes (Figure 1b) This structure has two nests one including Car (C) and

Motorcycle (MC) as private modes and the other including Walk and Ride (WampR) Drive and Ride

(DampR) Taxi (T) and Tel-taxi (T_T) as non-private modes The result of this nested logit (NL) model is

shown in Table 3

Although it is not a statistically significant improvement overall on the MNL model the statistically

significant inclusive value8 (IV) of 0889 for non-public modes relative to the fixed parameter value of

10 for public modes suggests that there is a structural advantage in selecting the NL specification

The normal test of a statistically significant difference between NL and MNL is an IV parameter

relative to 10 calculated using a Wald-test via equation 1

)1(Wald-test = (IVparameter ndash 1)std error

a The MNL structure

b Final nested structure

Figure 1- Model structure

8 Also called scale parameter

Alternatives

MCCar T_TT WampR DampR

Alternatives

MCCar T_TT

Public

WampR DampR

Private

13

We have (0889-1)2508 =-075 which would be rejected at the usual acceptable significance levels

This suggests that the NL model could be collapsed into an MNL form

Table 3- Nested logit (NL) model resultValueParameter

0889IV (Private)1000IV (nPrivate)

-2668335L( )-4057684L(0)

0342sup2

After the calibration process the variables that were statistically significant were identified and are

presented in Table 4 Table 5 presents the final model of the study with a goodness of fit of 031 for

the six studied modes For a general review of the model calibration results the effective factors can

be grouped under the following three categories TDM policy characteristics commuting trip

characteristics and household socio-economic characteristics which are all treated as alternative-

specific variables

52 Model results

Car (C)

It is expected that push policies impel car-drivers to choose other modes Table 5 shows that cordon

pricing and increase in parking cost cause individuals to choose not to use their car This is in line

with other studies suggesting that these policies are effective to discourage car usage (Hensher amp

Rose 2007 OFallon et al 2004) In addition the interaction between the policies of fuel cost

increase and increase in parking cost shows similar car usage discourage effect Because fuel cost is

related to the distance between home and work locations and parking cost is related to work time

the time that an individual spends out of the home is negatively affected by hisher likelihood to use

a car

14

Table 4 - Definition of the significant variables

AbbreviationVariableTransportation demand management measures

Measures

ParkingParking cost increase Rials per hour

CordonCordon price Rials per entranceAccessTransit access time shortage percent

Interaction of push measures

ParkampFuelParking cost and fuel cost simultaneous effectsCordonampFuelCordon pricing and fuel cost simultaneous effects

Interaction of pull measures

PT_timeampaccessPT time reduction and access improvement simultaneous effectsCommuting trip characteristics

Trip distanceDistance between home and workplaceTrip timeTravel time between home and workplace

Exp FuelLikelihood of unsubsidized fuel use (self-reported on a Likert scale)NtripsNumber of daily tripsPattern2Commuting with 1+ stop(s) in go or return

Pattern3Commuting with 2 workplacesFirst trip timeStart time of first trip

PnocarwkLikelihood of going to work in absence of that car (self-reported)PTnwaccNon-walk access to transit (yes=1)First NaccoNumber of passengers in first trip

PassengerAny passenger on that day (yes=1)Park_paymentParking payment in last weekNhempfullNumber of full employees in HH

CardependencyBoardalight a passenger or move freight in the trip (yes=1)D car ownBe the owner of the used vehicle (yes=1)

Car accCar accessibility in household (number of cars to number of HH driving licenses ratio)NmotorcycleNumber of motorcycles owned by HHD home placeHome Location is in study area (yes=1)

PermissionPermission to enter to study area (yes=1)ComfortI use my car because it is comfortablePoor_PTI use my car because transit is not good

HH socio-economic characteristics

FemaleGender (Female=1)Age lt30Age younger than 30 (yes=1)Age 30_39Age between 30 to 39 (yes=1)Job_durationNumber of years that individual has been at hisher job

Emp_fullFull-time employee (yes=1)Edu BSDegree of education is BSc (yes=1)Edu BS+Degree of education is higher than BSc(yes=1)

D childlt=18Child younger than 18 in HH (yes=1)

15

Table 5 ndash The mode choice model

Tel-Taxi(T_T)

Motorcycle(MC)

Drive amp Ride(DampR)

Taxi (T)Walk amp Ride(WampR)

Car (C)Mode

Variable-471756-37067-147911Constant

Transportation demand management measure variables00019-00045Cordon

-000072Parking-004308Access

-28443D-05Parkampfuel-32475D-06Cordonampfuel

00029Pt_timeampaccess

Commuting trip characteristics-04709Trip distance

-02163-00831Trip time-96755163655Exp fuel-16253Ntrips

-114779Pattern2-71008Pattern3

00282-00270First trip time-02439-01549Pnocarwk

-11322992883-32765PTnwacc-133701First Nacco

-7778-73782Accompany-00049000010Park_payment

201646195554Nhempfull-160144ComfortCar1

-206142DependencyCar1-16101883385-121224DependencyCar2

42176Poor_PTCar1-24988Poor_PTCar2

- -27221D car own70960-39136Car acc

1 -71112-156123Nmotorcycle-1436322762D home place

2 78826Permission

HH socio-economic characteristics149490Female

297584-24548Agelt30-136490Age30_39

079430366303585Job_duration-108743Emp_full-203468-64900Edu BS

10932856687-4499984445Edu BS+102271D childlt=18

-2677366L( )-3849556L(0)0305sup2

112127178592580607N

Note = Positive significance at 1 5 10 level

As expected individuals with higher income are more likely to use their car This is indicated in the

model by the positive signs of individuals who use fuel with fixed (unsubsidized) cost and individuals

16

who pay more in parking charges in the previous week of study Negative sign of Pnocarwk variable

shows that the commuters who stated that their commute depends on car availability are more

likely to use their car Individuals in households with more full-time employees are more likely to use

their car which may be the result of higher household income Not surprisingly commuters who

have permission are more likely to maintain car usage Among the household socio-economic

parameters greater job experience (Job_duration) and higher graduate levels (EduBS+) increase the

probability of car usage

Public transit accessed by walking (WampR)

Access time to transit negatively impacts WampR choice which is expected This result is similar to

findings for the city of Sydney (Hensher amp Rose 2007) The negative coefficient of first trip time

indicates that individuals are more likely to use WampR in the early morning This result seems to

reflect the better weather for walking and faster speed of WampR mode early in the morning

Obviously individuals who are not able to access transit stations via walking (PTnwacc) are less likely

to consider this mode Furthermore serving passengers on daily trips is also a deterrent to using

WampR

Initially assessing the individuals who stated that their car usage is due to poor public transit service

(Poor_PT) led to an unexpected result in favor of considering WampR By introducing to this variable

the number of household cars as a proxy for household income (Poor_PTCar1) the model shows

that of the previously mentioned individuals those who have lower income are the ones who have

to consider WampR The result is understandable as these individuals may have no alternative when

they have to change their mode (they also are not likely to consider other modes) Individuals with

higher levels of income who have to use their car during before or after work (Dependencycar1+)

are not likely to use WampR

The greater the number of motorcycles in a household the less likely commuters is to consider

WampR There appears to be a competition between motorcycle and PT for access to the city center

17

Better PT services in the center of the city in terms of coverage and frequency increases the

likelihood that its residents will consider WampR This is verified by the positive sign of the

D_home_place variable Commuters with greater job experience (Job_duration) in their workplace

are more likely to use this mode Although individuals with higher levels of education are not likely

to use WampR as education level increases avoidance of WampR decreases

Taxi (T)

Table 5 shows that none of the studied policies are significant in considering taxi usage It seems that

taxi usage considering its function in Iran as a non-private and non-public mode of transport is not

affected by pull or push policies A negative sign for taxi travel time indicates that individuals are not

likely to use this mode for longer trips This seems reasonable given that longer trips are more

expensive Commuters who are more likely to use fuel with no subsidy are not likely to use taxis As

mentioned before they prefer to use their car A higher number of trips in a day are also a deterrent

to considering taxi usage which may be due to increased cost for more trips Results show that an

individual with more daily trips avoids using taxis Commuters who are employed in more than one

workplace (Pattern 3) are not likely to use taxis This may be due to the fact that they have a lower

level of income which forces them to dedicate more time on the job

Initial results showed that individuals who stated that their car usage is due to poor public transit

service (Poor_PT) are not likely to use taxis This result was far from our expectations By introducing

to this variable the number of household cars as a proxy for household income the model shows

that the previously mentioned individuals who have higher income (Poor_PTCar1+) are the ones

who are not likely to consider taxis Furthermore because such individuals are not considering any

other modes they may treat taxi usage as a kind of PT mode with poor service

As expected greater access to cars in a household (Car_acc) lessens the likelihood of considering

taxis as an alternative Furthermore individuals in households with more motorcycle ownership are

less likely to consider taxis It seems like there is a competition among motorcycles and taxis for

18

access to the city center Younger commuters are less likely to use taxis and individuals with at least

master degrees do consider this mode in addition to their car

Public transit accessed by Drive (DampR)

This mode is affected by the simultaneous interaction of transit time and transit access

(PT_TimeampAccess) which is reflected in the fact that individuals prefer to use this mode for longer

trips Comparing this mode and WampR the first trip start time affects the consideration of this mode

differently Later morning commuters prefer to use their car to access PT modes Such commuters

may have higher income levels or managerial jobs Obviously individuals who are not able to access

PT stations by walking (PTnwacc) are likely to use DampR Serving passengers in daily trips is also a

deterrent in considering this mode which is similar to WampR but with a lower coefficient

Commuters with higher income levels who depend on their car during before or after work

(Dependencycar1+) are likely to use DampR Individuals who use their own car are less likely to use

this mode which is unexpected As a city center develops better PT network coverage and residents

have smaller distances to their workplaces they are unlikely to use DampR This is proven in the model

by a negative sign for D_home_place

Motorcycle (MC)

Increasing fuel cost and cordon pricing simultaneously discourage motorcycle usages Although fuel

cost is expected to reduce motorcycle usage to some extent its combined effect with cordon pricing

also reduces motorcycle usage However this variable is not as strong as other policy variables

=10)

Of the studied modes motorcycle usage is affected by the most commuting variables This may be

due to the fact that this mode is not common Commuting distance has a negative effect on

motorcycle usage which is expected It is worth noting that trip distance appears only in this mode

which may be a reflection of the role of distance in regards to the safety risk in considering this

19

mode Commuters with more stops to serve passengers while commuting (Pattern 2) are not likely

to use this mode which may be due to the poor passenger service of this mode

Individuals who state that commuting is independent of the mode (Pnocarwk) are not likely to use

MC By looking at the (First_Nacco) negative sign this could stem from the fact that the more

passengers there are on the first trip the less likely individuals are to consider MC Regarding the

low capacity of MC and its safety concerns such commuters avoid using this mode Commuters who

pay more parking charges (Park_payment) are less likely to use MC which is expected Individuals

who are dependent on their car during before or after their work time are not likely to use MC

even if they have lower levels of income (DependencyCar1) Individuals who use their own car

(D_car_own) are less likely to use this mode As expected individuals who live in households with

more motorcycle ownership are more likely to use this mode The positive sign of (Permission)

indicates that commuters who have permission to enter the study area do consider MC Because

such commuters generally provide that permission just for car usage this result is unexpected

As with commute variables of all the studied modes MC is affected by the greatest number of

socio-economic variables As expected young commuters (Agelt30) are more likely to use this mode

Commuters with Bachelor of Science degree are less likely to use this mode among others Full time

employees (Emp_full) are less likely to consider MC whereas commuters with more experience in

their jobs prefer to use it Results show that individuals who live in a household with children

younger than 18 are more likely to consider using a car

Tel-Taxi (T_T)

Results show that cordon pricing causes higher probability of using T_T In fact individuals who use

T_T as a mode with similar level of service as cars9 are more willing to pay the cost and make use of

the mode It is worth noting that the effect of cordon pricing in pushing commuters from car usage

9 As this mode does not have driving stress and parking search time in some cases it may have more amount of utility thana car does

20

(000045) is greater than its effect on pulling them to Tel-taxi (000019) This is because of the

possibility of considering other non-car modes

Because consideration of this mode is a function of its operation travel time (Trip_time) appears as

a deterrent in this mode utility function Table 5 shows that individuals are more sensitive to the trip

time when using T_T mode versus taxi which is expected due to their relative costs

The greater the number of full time employees in a family (Nhempfull) the higher the probability of

considering T_T by its commuters which may be due to the higher income level of these

households This is verified by the greater likelihood of using T_T rather than taxis by such

commuters Individuals with higher levels of income who depend on their car during before or after

work time are less likely to use T_T Commuters with lower income levels who state that they use

their car for the sake of comfort (Comfortcar1) are less likely to use T_T which may be due to its

cost Although such individuals do not consider any other modes they specifically avoid T_T Greater

access to cars in a household leads to greater likelihood of T_T usage which could be due to the

higher income level of a household As mentioned before such individuals even avoid taxis

Females who drive to their workplace are more likely to use T_T It seems like this part of society

considers this mode when desiring to avoid the difficulties of driving Younger commuters are less

likely to use T_T and individuals between 30 and 39 years of age are specifically avoiding this mode

Results show that university graduated commuters are more likely to use this mode

6 Marginal effects

To explore the effects of each policy on mode choice and to answer the second issue raised at the

beginning of this paper the marginal effects approach can be adopted Although the coefficients of

the models utility functions show the drivers behavior when facing one or more policies the

marginal effects of policies or their interactions may appropriately show the results of their

implementation More specifically the marginal effect for this study is interpreted as the change in

21

probability given a unit change in a variable ceteris paribus In this section the variable is defined as

a specific policy or a specific interaction of two policies Table 6 presents the marginal effects of the

studied policies and their interactions with mode choice The results are shown in the form of trip

percentages transferred away from the car to the studied modes and the probability-weighted

sample enumeration approach is adopted to find the values It is worth noting that this table is fully

compatible with Table 5 but the marginal effects that were less significant than 90 percent have

been removed

Table 6 - Marginal effects of policies (percent)

Tel-Taxi(TT)

Motorcycle(MC)

Drive ampRide(DampR)

Taxi (T)Walk ampRide(WampR)

Car (C)Mode

Variable-000088Cordon-000140Parking

-09069Access-0000001ParkampFuel

00040PT_TimeampAccess

Table 5 shows that a 1 Rial increase in cordon pricing causes a 000088 percent decrease in car

usage Furthermore a 1 Rial per hour increase in parking cost decreases the car usage probability to

00014 percent By assuming 8 hours for the average parking duration the daily marginal value of

parking cost converts to 000018 percent These values show that cordon pricing is more effective in

forcing individuals not to use their car than increasing parking cost with the same value Results also

show that a 1 Rial increase in parking cost and fuel cost simultaneously decreases the probability of

choosing car usage by 0000001 percent Table 5 also shows that a 1 minute decrease in transit

access time would result in a 09 percent increase in probability of choosing this mode It also shows

that each 1 minute decrease in transit travel time and transit access leads to a 0004 decrease in the

probability of choosing the DampR mode

22

The above results show that just one of the policy interactions is not significantly different from

none (ie CordonampFuel) and push policies have a significant role in impelling car users to consider

other modes

7 Planning

To demonstrate the application of the model some of its planning aspects for the city of Tehran are

presented here In fact the model is able to assess the effect of all studied variables which were

grouped into three tiers on commuters mode choice In this section we assess the effect of TDM

policies by adopting their pair-wise combinations for the city Table 6 shows that pull policies in this

study did not impose a major effect on car usage Because the average access time of the sample

was 11 minutes improving it by 50 percent (about 5 minutes) increases the probability of WampR to

509069100 which is equal to a 048 percent increase in considering this mode (and decrease in all

other modes) On the other hand decreasing transit time about 50 percent (about 19 minutes based

on sample average) and similarly improving access time decreases the probability of considering

DampR to 0004195100 which is equal to a 038 percent decrease in considering this mode (and

increase in all other modes) Therefore in this section we focused on the push policies which

appear in the utility function of car mode It is worth noting that in assessing each combination of

policies all other variables were assumed to be fixed as the current state

71 Parking pricing and cordon pricing

The average hourly parking charge of individuals in the sample is 71 Rials This charge is 3000 Rials

per hour in the curb lane on some streets about 2000 Rials per hour in parking lots and free in

alleys and other streets The average amount shows that most of the commuters benefited from

free parking spaces in alleys and non-controlled streets Figure 2 depicts the result of implementing

this policy and the cordon pricing policy simultaneously In line with marginal effects it can be seen

that in implementing each policy separately cordon pricing is more effective than parking pricing in

the planning range It is worth noting that commuters in this study are less sensitive to parking

23

pricing than cordon pricing One reason may be that they thought in any case they would find free

parking space Figure 2 shows that the effect of each policy depends on the level of the other policy

In other words the effect of the simultaneous implementation of two policies is not equal to the

sum of their separate individual effects

Figure 2- Effect of implementing parking pricing and cordon pricing

72 Parking pricing and increasing fuel cost

Figure 3 shows the simultaneous effects of parking pricing and increasing fuel cost policies Although

both of these policies are not effective separately their simultaneous effect is significant It can be

seen that their synergy appears in the higher levels of both policies

Figure 3 - Effect of implementing parking pricing and increasing fuel cost

24

73 Cordon pricing and increasing fuel cost

The effect of implementing cordon pricing and increasing fuel cost is shown in Figure 4 It can be

seen that these two policies have no interaction effect in decreasing car usage This result is verified

by this interactions insignificant marginal value in Table 6

Figure 4- Effect of implementing cordon pricing and increasing fuel cost

8 Conclusions

This study examined the role of TDM policies in individual mode choice for work trips in the city of

Tehran Five policies including increasing parking cost increasing fuel cost cordon pricing in the odd-

even zone of the city transit time reduction and transit access improvement were investigated Of

the five policies the former three were push policies and the latter two were pull policies The

design of experiments approach was used to design the questionnaire that would capture the stated

preferences of car commuters

A number of nested structures based on recognizing differences in the variances associated

with unobserved influences were examined and finally the MNL as a superior model for this study

was developed

25

The results show that for car users who faced TDM policies all of the five studied TDM

policies were factors in the mode choice process The model shows that in considering car usage

push policies play a main role and pull policies do not

Results also show that the interactions of policies are also significant The model shows that

the interaction of parking cost policy and fuel cost policy is significant in prompting car users to

consider other modes It also shows that the interaction of cordon pricing and increase in fuel cost is

effective to discourage motorcycle usage Furthermore the interaction of transit time reduction

policy and access improvements policy is significant in considering DampR as a transportation mode to

workplace

Pull policies in the study were expected to attract individuals to transit modes (pull policies

of this study are related to transit modes) This was verified by the model

Results show that although taxi usage is usually treated as an alternative for urban trips it

was not directly affected by the studied policies This may be due to its special function in Tehran as

a non-public and non-private mode

The results also show that except for the interaction of cordon pricing and increase in fuel

cost other interaction effects have significant marginal effects on mode choice

Results of the model show that to implement a single policy cordon pricing is the most

effective in decreasing car usage

This study assessed workplace commuters Thus future studies should explore the

individuals mode choice with other trip aims Additionally including more policies especially pull

policies in favor of non-transit modes is reasonable to rigorously assess the model

It can be seen that the studied modes are affected differently by the TDM policies and their

interactions The impact of the interaction of policies appeared in the consideration of three out of

six studied modes and also in two out of three developed graphs This issue needs more attention

within the transportation demand management context and further studies should be performed

26

Design of this study is based on the consideration of all two-way interactions with some

degrees of deficiency in D-efficient policy Focusing on the significant interactions of this study is

recommended to perform other studies with more D-efficient values

Further understanding of the impacts of TDM policies might be provided in future research

by focusing on the amount of their synergy in prohibiting car usage

9 References

Arentze TA Hofman F amp Timmermans HJP (2004) Predicting multi-faceted activityndashtravel adjustmentstrategies in response to possible congestion pricing scenarios using an Internet-based stated adaptationexperiment Transport Policy 11 p31ndash41

Burris M amp Patil S (2009) Estimating the Benefits of Managed Lanes Technical Report Texas TexasTransportation Institute University Transportation Center for Mobility

Cao X amp Mokhtarian PL (2005) How do individuals adapt their personal travel A conceptual exploration ofthe consideation of travel-related strategies Transport Policy 12(3) pp199-206

de Palma A amp Lindsey R (2001) Transportation Supply and congestion In International Encyclopedia of theSocial and Behavioral Sciences 1st ed Elsevier p15882ndash15888

Eriksson L Garvill J amp Nordlund AM (2008) Acceptability of single and combined transport policy measuresThe importance of environmental and policy specific beliefs Transportation Research Part A 42 p1117ndash1128

Eriksson L Nordlund AM amp Garvill J (2010) Expected car use reduction in response to structural traveldemand management measures Transportation Research Part F 13 p329ndash342

Golob TF amp Hensher DA (2007) The Trip chaining Activity of Sydney Residents A Cross-Section Assessmentby Age Group with a Focus on Seniors Journal of Transport Geography 15 p298ndash312

Hensher DA amp King J (2001) Parking demand and responsiveness to supply pricing and location in theSydney central business district Transportation Research A 35 pp177-96

Hensher DA amp Rose JM (2007) Development of commuter and non-commuter mode choice models for theassessment of new public transport infrastructure projects A case study Transportation Research Part A41 p428ndash443

Hensher DA Rose JM amp Greene WH (2005) Applied Choice Analysis A Primer New York CambridgeUniversity Press

Iranian Center of Statistics (ICS) (2009) Information of Iranian States [Online] Available at HYPERLINK httpwwwamarorgirUploadModulesContentsasset16tehrantehparthtmlhttpwwwamarorgirUploadModulesContentsasset16tehrantehparthtml [Accessed 13 November2009]

Kingham S Dickinson J amp Copsey S (2001) Travelling to work Will people move out of their cars TransportPolicy 8 pp151-60

Kuhfeld WF (2009) Marketing Research Methods in SAS Experimental Design Choice Conjoint andGraphical Techniques SAS 92 Edition NC USA SAS Institute Inc Cary Available on the Webwwwsupportsascomtechsuptnotetnote_stathtmlmarket

Litman T (2003) The online TDM Encyclopedia mobility management information gateway Transport Policy10 pp245-49

Litman T (2010) Online TDM Encyclopedia [Online] Available at HYPERLINK httpwwwvtpiorghttpwwwvtpiorg [Accessed 21 Augest 2010]

7

individual policies Vieira et al explored the concept of multi-instrumentality as a procedure of

policy integration and implementation whereby a systematic search for complementary policies was

sought when planning and designing one (or several) core policy(s) aiming to fulfill one particular

policy more effectively (Vieira et al 2007) They defined criteria to assess the TDM policies and

analyzed four improvement mechanisms in each pair of policies Based on meta-studies they

defined the potential improvement between different types of policies By defining synergy concept

as a benefit of integration May et al reviewed a number of examples to assess the concept and

found little evidence of synergy in outcome indicators (May et al 2006)

Based on the above discussion assessing individual behavioral response to more than one TDM

policy is an interesting issue within the TDM context The following three issues are addressed in this

paper developing a model to investigate the role of TDM policies in commuters mode choice

exploring the role of effective parameters on the consideration of each mode of travel and

suggesting a method to determine the results of implementing two TDM policies simultaneously In

this paper the stated preferences approach is used to model the car users mode choice using the

design of experiments principles

3 Stated preferences

The five policies selected for the city of Tehran consisted of three push and two pull policies The

policies were increasing parking cost increasing fuel cost cordon pricing into an odd-even zone2

transit (bus or subway) time reduction and transit access improvement The latter two were

described by setting measures in favor of the public transit vehicles in streets and intersections

decreasing the time of boarding and alighting at the stations and increasing the number of transit

lines and stops in the city

2 This zone explored in the next section

8

Parking costs fuel costs and public transit time policies are designated with three levels and cordon

price and public access time are designed with two levels Table 1 shows the policies and their levels

All push policies had fixed values for their levels for pull policies because there were variations in

the transit time and transit access time for individuals proportional values of the current state were

used which is different for each individual The term no change in Table 1 refers to the current value

of a policy that each individual already experiences The mean values are also presented in Table 1

for a better description of current state

In preparing a questionnaire for the stated preferences part the design of experiments approach

was adopted Full factorial design is the most general type of design in this approach which

introduces all combinations of all levels in the modeling process In other words full factorial design

produces 108 possible choice sets (33322) This design allows the investigation of all

interactions as well as the main effects in the model On the one hand fewer choice sets are

available when ignoring the effects of higher-order interactions and on the other hand these

interactions have a negligible role in the variance (Louviere et al 2000 Hensher et al 2005) thus

fractional factorial design methods have been proposed

Table 1- Policies and their levels

Measure Type Numberof levels

Description of levels Mean Value

Increasing parking cost Push 3 No change 4000 7000 Rials3 h 71 RialshNACordon pricing Push 2 25000 50000 Rialsday

Increasing fuel cost Push 3 No change 3000 5000 Rialsliter 1470 RialsLiterTransit time reduction Pull 3 No change 15 30 percent shortage 385 minTransit access improvement Pull 2 No change 25 percent shortage 11 min

Efficient design a type of fractional factorial design was used in the study and a design with 895

efficiency was adopted which allows assessing all two-way interactions of policies as well as the

3 10000 Rials are almost equal to 1 US dollar

9

main effects with only 36 choice sets4 (See (Rose amp Bliemer 2009) or (Kuhfeld 2009) for more

details on efficient design) To avoid a time-consuming questionnaire 36 choice sets (scenarios)

were randomly ordered and divided into six separate questionnaire types coded as 1 to 6 Each of

the questionnaires had six scenarios and each scenario consisted of five policies

4 Survey

Two push policies are currently being implemented in the city of Tehran The first is car-free

planning in the CBD area of the city and the second one is an odd-even scheme based on the last

digit of car plates that attempt to enter a zone which is about three times larger than and includes

the CBD area Based on their occupation a few people can drive to the CBD area with a license

called permission A stated preferences survey was assigned for the morning car commuters to the

odd-even zone but they were asked to ignore these two policies to find the accurate sensitivity of

individuals to the study policies The odd-even zone is selected as study area for the two following

reasons 1) because of odd-even control respondents are familiar with the fringes and they can

better imagine the entrance pricing area and 2) respondents are familiar with the limits that they

face half of the week and are thus aware of the alternative existing modes Compared to the CBD

area this zone covers more car commuters and the entrance restriction is more imaginable for this

zone than the former one Respondents were interviewed face-to-face in their workplaces midway

through the year 2009 The interviews were enhanced with a special card to better define the

scenarios

For this study 2196 scenario observations from 366 individuals were adopted The sample included

308 men (ie 841) and 58 women (ie 159) The figures are close to the employment

percentages in the city according to the Iranian Center of Statistics (ICS) This source indicates that

825 of Tehran employees are men and 175 are women (Iranian Center of Statistics (ICS) 2009)

Because this study focuses on car-using commuters comparisons between the sample and city data

4 Efficient design is also adopted in other studies such as managed lanes (Burris amp Patil 2009)

10

especially regarding educational distribution were impossible Table 2 presents demographics of the

sample

Table 2- Demographics gender marital household (HH) size employee type age HH employee(s)

Amount Percent

Gender Male 308 841Female 58 159

Marital Single 100 273Married 266 727

HH Size 1 4 112 86 2353 129 3524 90 2465 42 1156+ 15 41

Age 18~29 122 33330~39 146 39940~49 58 15950~59 32 8760+ 8 22

HH employee(s) 1 156 4262 159 4343 41 1124+ 10 27

The first part of the questionnaire is dedicated to gathering the occupation state home and job

locations the distance between these locations round-trip time (from home to workplace and then

workplace to home) and all car trip characteristics in the previous day or the day before it based on

plate number It was necessary that the respondents drive hisher car in the day studied to complete

the trip diary portion of the questionnaire5 The general reasons for car usage and the scenarios

formed the next portion In each scenario every respondent was asked the question How would

you travel to the workplace if all of these changes were in place on the day studied For example

one may have to pay 4000 Rialsh for parking 50000 Rials per entrance to the cordon the same

amount in transit access and fuel cost and a 15 percent decrease in transit time simultaneously

Depending on individual responses six main options were distinguished6 These choices were still

5 In designing the questionnaire the general form of questionnaire which has mentioned in OFallons study was adopted6 In the pre-test survey 14 modes is distinguished

11

drive a car (C) walk to the station and catch public transit (WampR) drive to a public station and catch

public transit (DampR) ride a motorcycle (MC) catch a taxi7 (T) and catch a taxi by phone (T_T) DampR is

somewhat different than the more familiar ldquoPark amp Riderdquo In fact in the fringes of the odd-even

zone there were no specialized parking lots dedicated to this purpose and commuters considered

Drive amp Ride because they were not allowed to pass the fringes

After each scenario if the respondent changed hisher mode the reason(s) for the change were

asked It could be a sole policy or a bundle of them Furthermore travel-related information was

sought These data were not part of the stated choice but they might have important influences on

individual choices These data consisted of car dependency (need to drive someone or move freight

in the trip) parking place type and average weekly parking costs car and motorcycle ownership and

number of household driving licenses

Depending on the individuals activity in that day three types of activity patterns were detected

Pattern 1 described individuals who had no stop in their commute Pattern 2 was for individuals who

had at least one stop on their way to or from work and pattern 3 was for the individuals who went

to another workplace in their daily activities

Finally for the sake of data generalization and the examination of household characteristics gender

age and household type employment status and education level were also asked

5 Mode choice model

In order to detect the policies that affect individual mode choice the logit modeling approach was

adopted In this model one can determine if the interaction of two policies affects the mode choice

In the calibration step 152 variables were defined and their effects on consideration of each mode

were examined

7 Taxis in Iran are somewhat different than taxis in other countries of the world In fact taxis in Iran are not hiring by oneperson or a group of people at a time Taxis allow passengers to board or alight along their path with respect to theircapacity In other word this mode is functioning similar to transit vehicles but the stops are not predefined

12

51 Model structure

Initially a multinomial logit (MNL) model is developed (Figure 1a) By selecting a number of tree

structures based on recognizing differences in the variances associated with unobserved influences

we find that the greatest similarity in variance profiles is associated with public transport modes as

opposed to non-public modes (Figure 1b) This structure has two nests one including Car (C) and

Motorcycle (MC) as private modes and the other including Walk and Ride (WampR) Drive and Ride

(DampR) Taxi (T) and Tel-taxi (T_T) as non-private modes The result of this nested logit (NL) model is

shown in Table 3

Although it is not a statistically significant improvement overall on the MNL model the statistically

significant inclusive value8 (IV) of 0889 for non-public modes relative to the fixed parameter value of

10 for public modes suggests that there is a structural advantage in selecting the NL specification

The normal test of a statistically significant difference between NL and MNL is an IV parameter

relative to 10 calculated using a Wald-test via equation 1

)1(Wald-test = (IVparameter ndash 1)std error

a The MNL structure

b Final nested structure

Figure 1- Model structure

8 Also called scale parameter

Alternatives

MCCar T_TT WampR DampR

Alternatives

MCCar T_TT

Public

WampR DampR

Private

13

We have (0889-1)2508 =-075 which would be rejected at the usual acceptable significance levels

This suggests that the NL model could be collapsed into an MNL form

Table 3- Nested logit (NL) model resultValueParameter

0889IV (Private)1000IV (nPrivate)

-2668335L( )-4057684L(0)

0342sup2

After the calibration process the variables that were statistically significant were identified and are

presented in Table 4 Table 5 presents the final model of the study with a goodness of fit of 031 for

the six studied modes For a general review of the model calibration results the effective factors can

be grouped under the following three categories TDM policy characteristics commuting trip

characteristics and household socio-economic characteristics which are all treated as alternative-

specific variables

52 Model results

Car (C)

It is expected that push policies impel car-drivers to choose other modes Table 5 shows that cordon

pricing and increase in parking cost cause individuals to choose not to use their car This is in line

with other studies suggesting that these policies are effective to discourage car usage (Hensher amp

Rose 2007 OFallon et al 2004) In addition the interaction between the policies of fuel cost

increase and increase in parking cost shows similar car usage discourage effect Because fuel cost is

related to the distance between home and work locations and parking cost is related to work time

the time that an individual spends out of the home is negatively affected by hisher likelihood to use

a car

14

Table 4 - Definition of the significant variables

AbbreviationVariableTransportation demand management measures

Measures

ParkingParking cost increase Rials per hour

CordonCordon price Rials per entranceAccessTransit access time shortage percent

Interaction of push measures

ParkampFuelParking cost and fuel cost simultaneous effectsCordonampFuelCordon pricing and fuel cost simultaneous effects

Interaction of pull measures

PT_timeampaccessPT time reduction and access improvement simultaneous effectsCommuting trip characteristics

Trip distanceDistance between home and workplaceTrip timeTravel time between home and workplace

Exp FuelLikelihood of unsubsidized fuel use (self-reported on a Likert scale)NtripsNumber of daily tripsPattern2Commuting with 1+ stop(s) in go or return

Pattern3Commuting with 2 workplacesFirst trip timeStart time of first trip

PnocarwkLikelihood of going to work in absence of that car (self-reported)PTnwaccNon-walk access to transit (yes=1)First NaccoNumber of passengers in first trip

PassengerAny passenger on that day (yes=1)Park_paymentParking payment in last weekNhempfullNumber of full employees in HH

CardependencyBoardalight a passenger or move freight in the trip (yes=1)D car ownBe the owner of the used vehicle (yes=1)

Car accCar accessibility in household (number of cars to number of HH driving licenses ratio)NmotorcycleNumber of motorcycles owned by HHD home placeHome Location is in study area (yes=1)

PermissionPermission to enter to study area (yes=1)ComfortI use my car because it is comfortablePoor_PTI use my car because transit is not good

HH socio-economic characteristics

FemaleGender (Female=1)Age lt30Age younger than 30 (yes=1)Age 30_39Age between 30 to 39 (yes=1)Job_durationNumber of years that individual has been at hisher job

Emp_fullFull-time employee (yes=1)Edu BSDegree of education is BSc (yes=1)Edu BS+Degree of education is higher than BSc(yes=1)

D childlt=18Child younger than 18 in HH (yes=1)

15

Table 5 ndash The mode choice model

Tel-Taxi(T_T)

Motorcycle(MC)

Drive amp Ride(DampR)

Taxi (T)Walk amp Ride(WampR)

Car (C)Mode

Variable-471756-37067-147911Constant

Transportation demand management measure variables00019-00045Cordon

-000072Parking-004308Access

-28443D-05Parkampfuel-32475D-06Cordonampfuel

00029Pt_timeampaccess

Commuting trip characteristics-04709Trip distance

-02163-00831Trip time-96755163655Exp fuel-16253Ntrips

-114779Pattern2-71008Pattern3

00282-00270First trip time-02439-01549Pnocarwk

-11322992883-32765PTnwacc-133701First Nacco

-7778-73782Accompany-00049000010Park_payment

201646195554Nhempfull-160144ComfortCar1

-206142DependencyCar1-16101883385-121224DependencyCar2

42176Poor_PTCar1-24988Poor_PTCar2

- -27221D car own70960-39136Car acc

1 -71112-156123Nmotorcycle-1436322762D home place

2 78826Permission

HH socio-economic characteristics149490Female

297584-24548Agelt30-136490Age30_39

079430366303585Job_duration-108743Emp_full-203468-64900Edu BS

10932856687-4499984445Edu BS+102271D childlt=18

-2677366L( )-3849556L(0)0305sup2

112127178592580607N

Note = Positive significance at 1 5 10 level

As expected individuals with higher income are more likely to use their car This is indicated in the

model by the positive signs of individuals who use fuel with fixed (unsubsidized) cost and individuals

16

who pay more in parking charges in the previous week of study Negative sign of Pnocarwk variable

shows that the commuters who stated that their commute depends on car availability are more

likely to use their car Individuals in households with more full-time employees are more likely to use

their car which may be the result of higher household income Not surprisingly commuters who

have permission are more likely to maintain car usage Among the household socio-economic

parameters greater job experience (Job_duration) and higher graduate levels (EduBS+) increase the

probability of car usage

Public transit accessed by walking (WampR)

Access time to transit negatively impacts WampR choice which is expected This result is similar to

findings for the city of Sydney (Hensher amp Rose 2007) The negative coefficient of first trip time

indicates that individuals are more likely to use WampR in the early morning This result seems to

reflect the better weather for walking and faster speed of WampR mode early in the morning

Obviously individuals who are not able to access transit stations via walking (PTnwacc) are less likely

to consider this mode Furthermore serving passengers on daily trips is also a deterrent to using

WampR

Initially assessing the individuals who stated that their car usage is due to poor public transit service

(Poor_PT) led to an unexpected result in favor of considering WampR By introducing to this variable

the number of household cars as a proxy for household income (Poor_PTCar1) the model shows

that of the previously mentioned individuals those who have lower income are the ones who have

to consider WampR The result is understandable as these individuals may have no alternative when

they have to change their mode (they also are not likely to consider other modes) Individuals with

higher levels of income who have to use their car during before or after work (Dependencycar1+)

are not likely to use WampR

The greater the number of motorcycles in a household the less likely commuters is to consider

WampR There appears to be a competition between motorcycle and PT for access to the city center

17

Better PT services in the center of the city in terms of coverage and frequency increases the

likelihood that its residents will consider WampR This is verified by the positive sign of the

D_home_place variable Commuters with greater job experience (Job_duration) in their workplace

are more likely to use this mode Although individuals with higher levels of education are not likely

to use WampR as education level increases avoidance of WampR decreases

Taxi (T)

Table 5 shows that none of the studied policies are significant in considering taxi usage It seems that

taxi usage considering its function in Iran as a non-private and non-public mode of transport is not

affected by pull or push policies A negative sign for taxi travel time indicates that individuals are not

likely to use this mode for longer trips This seems reasonable given that longer trips are more

expensive Commuters who are more likely to use fuel with no subsidy are not likely to use taxis As

mentioned before they prefer to use their car A higher number of trips in a day are also a deterrent

to considering taxi usage which may be due to increased cost for more trips Results show that an

individual with more daily trips avoids using taxis Commuters who are employed in more than one

workplace (Pattern 3) are not likely to use taxis This may be due to the fact that they have a lower

level of income which forces them to dedicate more time on the job

Initial results showed that individuals who stated that their car usage is due to poor public transit

service (Poor_PT) are not likely to use taxis This result was far from our expectations By introducing

to this variable the number of household cars as a proxy for household income the model shows

that the previously mentioned individuals who have higher income (Poor_PTCar1+) are the ones

who are not likely to consider taxis Furthermore because such individuals are not considering any

other modes they may treat taxi usage as a kind of PT mode with poor service

As expected greater access to cars in a household (Car_acc) lessens the likelihood of considering

taxis as an alternative Furthermore individuals in households with more motorcycle ownership are

less likely to consider taxis It seems like there is a competition among motorcycles and taxis for

18

access to the city center Younger commuters are less likely to use taxis and individuals with at least

master degrees do consider this mode in addition to their car

Public transit accessed by Drive (DampR)

This mode is affected by the simultaneous interaction of transit time and transit access

(PT_TimeampAccess) which is reflected in the fact that individuals prefer to use this mode for longer

trips Comparing this mode and WampR the first trip start time affects the consideration of this mode

differently Later morning commuters prefer to use their car to access PT modes Such commuters

may have higher income levels or managerial jobs Obviously individuals who are not able to access

PT stations by walking (PTnwacc) are likely to use DampR Serving passengers in daily trips is also a

deterrent in considering this mode which is similar to WampR but with a lower coefficient

Commuters with higher income levels who depend on their car during before or after work

(Dependencycar1+) are likely to use DampR Individuals who use their own car are less likely to use

this mode which is unexpected As a city center develops better PT network coverage and residents

have smaller distances to their workplaces they are unlikely to use DampR This is proven in the model

by a negative sign for D_home_place

Motorcycle (MC)

Increasing fuel cost and cordon pricing simultaneously discourage motorcycle usages Although fuel

cost is expected to reduce motorcycle usage to some extent its combined effect with cordon pricing

also reduces motorcycle usage However this variable is not as strong as other policy variables

=10)

Of the studied modes motorcycle usage is affected by the most commuting variables This may be

due to the fact that this mode is not common Commuting distance has a negative effect on

motorcycle usage which is expected It is worth noting that trip distance appears only in this mode

which may be a reflection of the role of distance in regards to the safety risk in considering this

19

mode Commuters with more stops to serve passengers while commuting (Pattern 2) are not likely

to use this mode which may be due to the poor passenger service of this mode

Individuals who state that commuting is independent of the mode (Pnocarwk) are not likely to use

MC By looking at the (First_Nacco) negative sign this could stem from the fact that the more

passengers there are on the first trip the less likely individuals are to consider MC Regarding the

low capacity of MC and its safety concerns such commuters avoid using this mode Commuters who

pay more parking charges (Park_payment) are less likely to use MC which is expected Individuals

who are dependent on their car during before or after their work time are not likely to use MC

even if they have lower levels of income (DependencyCar1) Individuals who use their own car

(D_car_own) are less likely to use this mode As expected individuals who live in households with

more motorcycle ownership are more likely to use this mode The positive sign of (Permission)

indicates that commuters who have permission to enter the study area do consider MC Because

such commuters generally provide that permission just for car usage this result is unexpected

As with commute variables of all the studied modes MC is affected by the greatest number of

socio-economic variables As expected young commuters (Agelt30) are more likely to use this mode

Commuters with Bachelor of Science degree are less likely to use this mode among others Full time

employees (Emp_full) are less likely to consider MC whereas commuters with more experience in

their jobs prefer to use it Results show that individuals who live in a household with children

younger than 18 are more likely to consider using a car

Tel-Taxi (T_T)

Results show that cordon pricing causes higher probability of using T_T In fact individuals who use

T_T as a mode with similar level of service as cars9 are more willing to pay the cost and make use of

the mode It is worth noting that the effect of cordon pricing in pushing commuters from car usage

9 As this mode does not have driving stress and parking search time in some cases it may have more amount of utility thana car does

20

(000045) is greater than its effect on pulling them to Tel-taxi (000019) This is because of the

possibility of considering other non-car modes

Because consideration of this mode is a function of its operation travel time (Trip_time) appears as

a deterrent in this mode utility function Table 5 shows that individuals are more sensitive to the trip

time when using T_T mode versus taxi which is expected due to their relative costs

The greater the number of full time employees in a family (Nhempfull) the higher the probability of

considering T_T by its commuters which may be due to the higher income level of these

households This is verified by the greater likelihood of using T_T rather than taxis by such

commuters Individuals with higher levels of income who depend on their car during before or after

work time are less likely to use T_T Commuters with lower income levels who state that they use

their car for the sake of comfort (Comfortcar1) are less likely to use T_T which may be due to its

cost Although such individuals do not consider any other modes they specifically avoid T_T Greater

access to cars in a household leads to greater likelihood of T_T usage which could be due to the

higher income level of a household As mentioned before such individuals even avoid taxis

Females who drive to their workplace are more likely to use T_T It seems like this part of society

considers this mode when desiring to avoid the difficulties of driving Younger commuters are less

likely to use T_T and individuals between 30 and 39 years of age are specifically avoiding this mode

Results show that university graduated commuters are more likely to use this mode

6 Marginal effects

To explore the effects of each policy on mode choice and to answer the second issue raised at the

beginning of this paper the marginal effects approach can be adopted Although the coefficients of

the models utility functions show the drivers behavior when facing one or more policies the

marginal effects of policies or their interactions may appropriately show the results of their

implementation More specifically the marginal effect for this study is interpreted as the change in

21

probability given a unit change in a variable ceteris paribus In this section the variable is defined as

a specific policy or a specific interaction of two policies Table 6 presents the marginal effects of the

studied policies and their interactions with mode choice The results are shown in the form of trip

percentages transferred away from the car to the studied modes and the probability-weighted

sample enumeration approach is adopted to find the values It is worth noting that this table is fully

compatible with Table 5 but the marginal effects that were less significant than 90 percent have

been removed

Table 6 - Marginal effects of policies (percent)

Tel-Taxi(TT)

Motorcycle(MC)

Drive ampRide(DampR)

Taxi (T)Walk ampRide(WampR)

Car (C)Mode

Variable-000088Cordon-000140Parking

-09069Access-0000001ParkampFuel

00040PT_TimeampAccess

Table 5 shows that a 1 Rial increase in cordon pricing causes a 000088 percent decrease in car

usage Furthermore a 1 Rial per hour increase in parking cost decreases the car usage probability to

00014 percent By assuming 8 hours for the average parking duration the daily marginal value of

parking cost converts to 000018 percent These values show that cordon pricing is more effective in

forcing individuals not to use their car than increasing parking cost with the same value Results also

show that a 1 Rial increase in parking cost and fuel cost simultaneously decreases the probability of

choosing car usage by 0000001 percent Table 5 also shows that a 1 minute decrease in transit

access time would result in a 09 percent increase in probability of choosing this mode It also shows

that each 1 minute decrease in transit travel time and transit access leads to a 0004 decrease in the

probability of choosing the DampR mode

22

The above results show that just one of the policy interactions is not significantly different from

none (ie CordonampFuel) and push policies have a significant role in impelling car users to consider

other modes

7 Planning

To demonstrate the application of the model some of its planning aspects for the city of Tehran are

presented here In fact the model is able to assess the effect of all studied variables which were

grouped into three tiers on commuters mode choice In this section we assess the effect of TDM

policies by adopting their pair-wise combinations for the city Table 6 shows that pull policies in this

study did not impose a major effect on car usage Because the average access time of the sample

was 11 minutes improving it by 50 percent (about 5 minutes) increases the probability of WampR to

509069100 which is equal to a 048 percent increase in considering this mode (and decrease in all

other modes) On the other hand decreasing transit time about 50 percent (about 19 minutes based

on sample average) and similarly improving access time decreases the probability of considering

DampR to 0004195100 which is equal to a 038 percent decrease in considering this mode (and

increase in all other modes) Therefore in this section we focused on the push policies which

appear in the utility function of car mode It is worth noting that in assessing each combination of

policies all other variables were assumed to be fixed as the current state

71 Parking pricing and cordon pricing

The average hourly parking charge of individuals in the sample is 71 Rials This charge is 3000 Rials

per hour in the curb lane on some streets about 2000 Rials per hour in parking lots and free in

alleys and other streets The average amount shows that most of the commuters benefited from

free parking spaces in alleys and non-controlled streets Figure 2 depicts the result of implementing

this policy and the cordon pricing policy simultaneously In line with marginal effects it can be seen

that in implementing each policy separately cordon pricing is more effective than parking pricing in

the planning range It is worth noting that commuters in this study are less sensitive to parking

23

pricing than cordon pricing One reason may be that they thought in any case they would find free

parking space Figure 2 shows that the effect of each policy depends on the level of the other policy

In other words the effect of the simultaneous implementation of two policies is not equal to the

sum of their separate individual effects

Figure 2- Effect of implementing parking pricing and cordon pricing

72 Parking pricing and increasing fuel cost

Figure 3 shows the simultaneous effects of parking pricing and increasing fuel cost policies Although

both of these policies are not effective separately their simultaneous effect is significant It can be

seen that their synergy appears in the higher levels of both policies

Figure 3 - Effect of implementing parking pricing and increasing fuel cost

24

73 Cordon pricing and increasing fuel cost

The effect of implementing cordon pricing and increasing fuel cost is shown in Figure 4 It can be

seen that these two policies have no interaction effect in decreasing car usage This result is verified

by this interactions insignificant marginal value in Table 6

Figure 4- Effect of implementing cordon pricing and increasing fuel cost

8 Conclusions

This study examined the role of TDM policies in individual mode choice for work trips in the city of

Tehran Five policies including increasing parking cost increasing fuel cost cordon pricing in the odd-

even zone of the city transit time reduction and transit access improvement were investigated Of

the five policies the former three were push policies and the latter two were pull policies The

design of experiments approach was used to design the questionnaire that would capture the stated

preferences of car commuters

A number of nested structures based on recognizing differences in the variances associated

with unobserved influences were examined and finally the MNL as a superior model for this study

was developed

25

The results show that for car users who faced TDM policies all of the five studied TDM

policies were factors in the mode choice process The model shows that in considering car usage

push policies play a main role and pull policies do not

Results also show that the interactions of policies are also significant The model shows that

the interaction of parking cost policy and fuel cost policy is significant in prompting car users to

consider other modes It also shows that the interaction of cordon pricing and increase in fuel cost is

effective to discourage motorcycle usage Furthermore the interaction of transit time reduction

policy and access improvements policy is significant in considering DampR as a transportation mode to

workplace

Pull policies in the study were expected to attract individuals to transit modes (pull policies

of this study are related to transit modes) This was verified by the model

Results show that although taxi usage is usually treated as an alternative for urban trips it

was not directly affected by the studied policies This may be due to its special function in Tehran as

a non-public and non-private mode

The results also show that except for the interaction of cordon pricing and increase in fuel

cost other interaction effects have significant marginal effects on mode choice

Results of the model show that to implement a single policy cordon pricing is the most

effective in decreasing car usage

This study assessed workplace commuters Thus future studies should explore the

individuals mode choice with other trip aims Additionally including more policies especially pull

policies in favor of non-transit modes is reasonable to rigorously assess the model

It can be seen that the studied modes are affected differently by the TDM policies and their

interactions The impact of the interaction of policies appeared in the consideration of three out of

six studied modes and also in two out of three developed graphs This issue needs more attention

within the transportation demand management context and further studies should be performed

26

Design of this study is based on the consideration of all two-way interactions with some

degrees of deficiency in D-efficient policy Focusing on the significant interactions of this study is

recommended to perform other studies with more D-efficient values

Further understanding of the impacts of TDM policies might be provided in future research

by focusing on the amount of their synergy in prohibiting car usage

9 References

Arentze TA Hofman F amp Timmermans HJP (2004) Predicting multi-faceted activityndashtravel adjustmentstrategies in response to possible congestion pricing scenarios using an Internet-based stated adaptationexperiment Transport Policy 11 p31ndash41

Burris M amp Patil S (2009) Estimating the Benefits of Managed Lanes Technical Report Texas TexasTransportation Institute University Transportation Center for Mobility

Cao X amp Mokhtarian PL (2005) How do individuals adapt their personal travel A conceptual exploration ofthe consideation of travel-related strategies Transport Policy 12(3) pp199-206

de Palma A amp Lindsey R (2001) Transportation Supply and congestion In International Encyclopedia of theSocial and Behavioral Sciences 1st ed Elsevier p15882ndash15888

Eriksson L Garvill J amp Nordlund AM (2008) Acceptability of single and combined transport policy measuresThe importance of environmental and policy specific beliefs Transportation Research Part A 42 p1117ndash1128

Eriksson L Nordlund AM amp Garvill J (2010) Expected car use reduction in response to structural traveldemand management measures Transportation Research Part F 13 p329ndash342

Golob TF amp Hensher DA (2007) The Trip chaining Activity of Sydney Residents A Cross-Section Assessmentby Age Group with a Focus on Seniors Journal of Transport Geography 15 p298ndash312

Hensher DA amp King J (2001) Parking demand and responsiveness to supply pricing and location in theSydney central business district Transportation Research A 35 pp177-96

Hensher DA amp Rose JM (2007) Development of commuter and non-commuter mode choice models for theassessment of new public transport infrastructure projects A case study Transportation Research Part A41 p428ndash443

Hensher DA Rose JM amp Greene WH (2005) Applied Choice Analysis A Primer New York CambridgeUniversity Press

Iranian Center of Statistics (ICS) (2009) Information of Iranian States [Online] Available at HYPERLINK httpwwwamarorgirUploadModulesContentsasset16tehrantehparthtmlhttpwwwamarorgirUploadModulesContentsasset16tehrantehparthtml [Accessed 13 November2009]

Kingham S Dickinson J amp Copsey S (2001) Travelling to work Will people move out of their cars TransportPolicy 8 pp151-60

Kuhfeld WF (2009) Marketing Research Methods in SAS Experimental Design Choice Conjoint andGraphical Techniques SAS 92 Edition NC USA SAS Institute Inc Cary Available on the Webwwwsupportsascomtechsuptnotetnote_stathtmlmarket

Litman T (2003) The online TDM Encyclopedia mobility management information gateway Transport Policy10 pp245-49

Litman T (2010) Online TDM Encyclopedia [Online] Available at HYPERLINK httpwwwvtpiorghttpwwwvtpiorg [Accessed 21 Augest 2010]

8

Parking costs fuel costs and public transit time policies are designated with three levels and cordon

price and public access time are designed with two levels Table 1 shows the policies and their levels

All push policies had fixed values for their levels for pull policies because there were variations in

the transit time and transit access time for individuals proportional values of the current state were

used which is different for each individual The term no change in Table 1 refers to the current value

of a policy that each individual already experiences The mean values are also presented in Table 1

for a better description of current state

In preparing a questionnaire for the stated preferences part the design of experiments approach

was adopted Full factorial design is the most general type of design in this approach which

introduces all combinations of all levels in the modeling process In other words full factorial design

produces 108 possible choice sets (33322) This design allows the investigation of all

interactions as well as the main effects in the model On the one hand fewer choice sets are

available when ignoring the effects of higher-order interactions and on the other hand these

interactions have a negligible role in the variance (Louviere et al 2000 Hensher et al 2005) thus

fractional factorial design methods have been proposed

Table 1- Policies and their levels

Measure Type Numberof levels

Description of levels Mean Value

Increasing parking cost Push 3 No change 4000 7000 Rials3 h 71 RialshNACordon pricing Push 2 25000 50000 Rialsday

Increasing fuel cost Push 3 No change 3000 5000 Rialsliter 1470 RialsLiterTransit time reduction Pull 3 No change 15 30 percent shortage 385 minTransit access improvement Pull 2 No change 25 percent shortage 11 min

Efficient design a type of fractional factorial design was used in the study and a design with 895

efficiency was adopted which allows assessing all two-way interactions of policies as well as the

3 10000 Rials are almost equal to 1 US dollar

9

main effects with only 36 choice sets4 (See (Rose amp Bliemer 2009) or (Kuhfeld 2009) for more

details on efficient design) To avoid a time-consuming questionnaire 36 choice sets (scenarios)

were randomly ordered and divided into six separate questionnaire types coded as 1 to 6 Each of

the questionnaires had six scenarios and each scenario consisted of five policies

4 Survey

Two push policies are currently being implemented in the city of Tehran The first is car-free

planning in the CBD area of the city and the second one is an odd-even scheme based on the last

digit of car plates that attempt to enter a zone which is about three times larger than and includes

the CBD area Based on their occupation a few people can drive to the CBD area with a license

called permission A stated preferences survey was assigned for the morning car commuters to the

odd-even zone but they were asked to ignore these two policies to find the accurate sensitivity of

individuals to the study policies The odd-even zone is selected as study area for the two following

reasons 1) because of odd-even control respondents are familiar with the fringes and they can

better imagine the entrance pricing area and 2) respondents are familiar with the limits that they

face half of the week and are thus aware of the alternative existing modes Compared to the CBD

area this zone covers more car commuters and the entrance restriction is more imaginable for this

zone than the former one Respondents were interviewed face-to-face in their workplaces midway

through the year 2009 The interviews were enhanced with a special card to better define the

scenarios

For this study 2196 scenario observations from 366 individuals were adopted The sample included

308 men (ie 841) and 58 women (ie 159) The figures are close to the employment

percentages in the city according to the Iranian Center of Statistics (ICS) This source indicates that

825 of Tehran employees are men and 175 are women (Iranian Center of Statistics (ICS) 2009)

Because this study focuses on car-using commuters comparisons between the sample and city data

4 Efficient design is also adopted in other studies such as managed lanes (Burris amp Patil 2009)

10

especially regarding educational distribution were impossible Table 2 presents demographics of the

sample

Table 2- Demographics gender marital household (HH) size employee type age HH employee(s)

Amount Percent

Gender Male 308 841Female 58 159

Marital Single 100 273Married 266 727

HH Size 1 4 112 86 2353 129 3524 90 2465 42 1156+ 15 41

Age 18~29 122 33330~39 146 39940~49 58 15950~59 32 8760+ 8 22

HH employee(s) 1 156 4262 159 4343 41 1124+ 10 27

The first part of the questionnaire is dedicated to gathering the occupation state home and job

locations the distance between these locations round-trip time (from home to workplace and then

workplace to home) and all car trip characteristics in the previous day or the day before it based on

plate number It was necessary that the respondents drive hisher car in the day studied to complete

the trip diary portion of the questionnaire5 The general reasons for car usage and the scenarios

formed the next portion In each scenario every respondent was asked the question How would

you travel to the workplace if all of these changes were in place on the day studied For example

one may have to pay 4000 Rialsh for parking 50000 Rials per entrance to the cordon the same

amount in transit access and fuel cost and a 15 percent decrease in transit time simultaneously

Depending on individual responses six main options were distinguished6 These choices were still

5 In designing the questionnaire the general form of questionnaire which has mentioned in OFallons study was adopted6 In the pre-test survey 14 modes is distinguished

11

drive a car (C) walk to the station and catch public transit (WampR) drive to a public station and catch

public transit (DampR) ride a motorcycle (MC) catch a taxi7 (T) and catch a taxi by phone (T_T) DampR is

somewhat different than the more familiar ldquoPark amp Riderdquo In fact in the fringes of the odd-even

zone there were no specialized parking lots dedicated to this purpose and commuters considered

Drive amp Ride because they were not allowed to pass the fringes

After each scenario if the respondent changed hisher mode the reason(s) for the change were

asked It could be a sole policy or a bundle of them Furthermore travel-related information was

sought These data were not part of the stated choice but they might have important influences on

individual choices These data consisted of car dependency (need to drive someone or move freight

in the trip) parking place type and average weekly parking costs car and motorcycle ownership and

number of household driving licenses

Depending on the individuals activity in that day three types of activity patterns were detected

Pattern 1 described individuals who had no stop in their commute Pattern 2 was for individuals who

had at least one stop on their way to or from work and pattern 3 was for the individuals who went

to another workplace in their daily activities

Finally for the sake of data generalization and the examination of household characteristics gender

age and household type employment status and education level were also asked

5 Mode choice model

In order to detect the policies that affect individual mode choice the logit modeling approach was

adopted In this model one can determine if the interaction of two policies affects the mode choice

In the calibration step 152 variables were defined and their effects on consideration of each mode

were examined

7 Taxis in Iran are somewhat different than taxis in other countries of the world In fact taxis in Iran are not hiring by oneperson or a group of people at a time Taxis allow passengers to board or alight along their path with respect to theircapacity In other word this mode is functioning similar to transit vehicles but the stops are not predefined

12

51 Model structure

Initially a multinomial logit (MNL) model is developed (Figure 1a) By selecting a number of tree

structures based on recognizing differences in the variances associated with unobserved influences

we find that the greatest similarity in variance profiles is associated with public transport modes as

opposed to non-public modes (Figure 1b) This structure has two nests one including Car (C) and

Motorcycle (MC) as private modes and the other including Walk and Ride (WampR) Drive and Ride

(DampR) Taxi (T) and Tel-taxi (T_T) as non-private modes The result of this nested logit (NL) model is

shown in Table 3

Although it is not a statistically significant improvement overall on the MNL model the statistically

significant inclusive value8 (IV) of 0889 for non-public modes relative to the fixed parameter value of

10 for public modes suggests that there is a structural advantage in selecting the NL specification

The normal test of a statistically significant difference between NL and MNL is an IV parameter

relative to 10 calculated using a Wald-test via equation 1

)1(Wald-test = (IVparameter ndash 1)std error

a The MNL structure

b Final nested structure

Figure 1- Model structure

8 Also called scale parameter

Alternatives

MCCar T_TT WampR DampR

Alternatives

MCCar T_TT

Public

WampR DampR

Private

13

We have (0889-1)2508 =-075 which would be rejected at the usual acceptable significance levels

This suggests that the NL model could be collapsed into an MNL form

Table 3- Nested logit (NL) model resultValueParameter

0889IV (Private)1000IV (nPrivate)

-2668335L( )-4057684L(0)

0342sup2

After the calibration process the variables that were statistically significant were identified and are

presented in Table 4 Table 5 presents the final model of the study with a goodness of fit of 031 for

the six studied modes For a general review of the model calibration results the effective factors can

be grouped under the following three categories TDM policy characteristics commuting trip

characteristics and household socio-economic characteristics which are all treated as alternative-

specific variables

52 Model results

Car (C)

It is expected that push policies impel car-drivers to choose other modes Table 5 shows that cordon

pricing and increase in parking cost cause individuals to choose not to use their car This is in line

with other studies suggesting that these policies are effective to discourage car usage (Hensher amp

Rose 2007 OFallon et al 2004) In addition the interaction between the policies of fuel cost

increase and increase in parking cost shows similar car usage discourage effect Because fuel cost is

related to the distance between home and work locations and parking cost is related to work time

the time that an individual spends out of the home is negatively affected by hisher likelihood to use

a car

14

Table 4 - Definition of the significant variables

AbbreviationVariableTransportation demand management measures

Measures

ParkingParking cost increase Rials per hour

CordonCordon price Rials per entranceAccessTransit access time shortage percent

Interaction of push measures

ParkampFuelParking cost and fuel cost simultaneous effectsCordonampFuelCordon pricing and fuel cost simultaneous effects

Interaction of pull measures

PT_timeampaccessPT time reduction and access improvement simultaneous effectsCommuting trip characteristics

Trip distanceDistance between home and workplaceTrip timeTravel time between home and workplace

Exp FuelLikelihood of unsubsidized fuel use (self-reported on a Likert scale)NtripsNumber of daily tripsPattern2Commuting with 1+ stop(s) in go or return

Pattern3Commuting with 2 workplacesFirst trip timeStart time of first trip

PnocarwkLikelihood of going to work in absence of that car (self-reported)PTnwaccNon-walk access to transit (yes=1)First NaccoNumber of passengers in first trip

PassengerAny passenger on that day (yes=1)Park_paymentParking payment in last weekNhempfullNumber of full employees in HH

CardependencyBoardalight a passenger or move freight in the trip (yes=1)D car ownBe the owner of the used vehicle (yes=1)

Car accCar accessibility in household (number of cars to number of HH driving licenses ratio)NmotorcycleNumber of motorcycles owned by HHD home placeHome Location is in study area (yes=1)

PermissionPermission to enter to study area (yes=1)ComfortI use my car because it is comfortablePoor_PTI use my car because transit is not good

HH socio-economic characteristics

FemaleGender (Female=1)Age lt30Age younger than 30 (yes=1)Age 30_39Age between 30 to 39 (yes=1)Job_durationNumber of years that individual has been at hisher job

Emp_fullFull-time employee (yes=1)Edu BSDegree of education is BSc (yes=1)Edu BS+Degree of education is higher than BSc(yes=1)

D childlt=18Child younger than 18 in HH (yes=1)

15

Table 5 ndash The mode choice model

Tel-Taxi(T_T)

Motorcycle(MC)

Drive amp Ride(DampR)

Taxi (T)Walk amp Ride(WampR)

Car (C)Mode

Variable-471756-37067-147911Constant

Transportation demand management measure variables00019-00045Cordon

-000072Parking-004308Access

-28443D-05Parkampfuel-32475D-06Cordonampfuel

00029Pt_timeampaccess

Commuting trip characteristics-04709Trip distance

-02163-00831Trip time-96755163655Exp fuel-16253Ntrips

-114779Pattern2-71008Pattern3

00282-00270First trip time-02439-01549Pnocarwk

-11322992883-32765PTnwacc-133701First Nacco

-7778-73782Accompany-00049000010Park_payment

201646195554Nhempfull-160144ComfortCar1

-206142DependencyCar1-16101883385-121224DependencyCar2

42176Poor_PTCar1-24988Poor_PTCar2

- -27221D car own70960-39136Car acc

1 -71112-156123Nmotorcycle-1436322762D home place

2 78826Permission

HH socio-economic characteristics149490Female

297584-24548Agelt30-136490Age30_39

079430366303585Job_duration-108743Emp_full-203468-64900Edu BS

10932856687-4499984445Edu BS+102271D childlt=18

-2677366L( )-3849556L(0)0305sup2

112127178592580607N

Note = Positive significance at 1 5 10 level

As expected individuals with higher income are more likely to use their car This is indicated in the

model by the positive signs of individuals who use fuel with fixed (unsubsidized) cost and individuals

16

who pay more in parking charges in the previous week of study Negative sign of Pnocarwk variable

shows that the commuters who stated that their commute depends on car availability are more

likely to use their car Individuals in households with more full-time employees are more likely to use

their car which may be the result of higher household income Not surprisingly commuters who

have permission are more likely to maintain car usage Among the household socio-economic

parameters greater job experience (Job_duration) and higher graduate levels (EduBS+) increase the

probability of car usage

Public transit accessed by walking (WampR)

Access time to transit negatively impacts WampR choice which is expected This result is similar to

findings for the city of Sydney (Hensher amp Rose 2007) The negative coefficient of first trip time

indicates that individuals are more likely to use WampR in the early morning This result seems to

reflect the better weather for walking and faster speed of WampR mode early in the morning

Obviously individuals who are not able to access transit stations via walking (PTnwacc) are less likely

to consider this mode Furthermore serving passengers on daily trips is also a deterrent to using

WampR

Initially assessing the individuals who stated that their car usage is due to poor public transit service

(Poor_PT) led to an unexpected result in favor of considering WampR By introducing to this variable

the number of household cars as a proxy for household income (Poor_PTCar1) the model shows

that of the previously mentioned individuals those who have lower income are the ones who have

to consider WampR The result is understandable as these individuals may have no alternative when

they have to change their mode (they also are not likely to consider other modes) Individuals with

higher levels of income who have to use their car during before or after work (Dependencycar1+)

are not likely to use WampR

The greater the number of motorcycles in a household the less likely commuters is to consider

WampR There appears to be a competition between motorcycle and PT for access to the city center

17

Better PT services in the center of the city in terms of coverage and frequency increases the

likelihood that its residents will consider WampR This is verified by the positive sign of the

D_home_place variable Commuters with greater job experience (Job_duration) in their workplace

are more likely to use this mode Although individuals with higher levels of education are not likely

to use WampR as education level increases avoidance of WampR decreases

Taxi (T)

Table 5 shows that none of the studied policies are significant in considering taxi usage It seems that

taxi usage considering its function in Iran as a non-private and non-public mode of transport is not

affected by pull or push policies A negative sign for taxi travel time indicates that individuals are not

likely to use this mode for longer trips This seems reasonable given that longer trips are more

expensive Commuters who are more likely to use fuel with no subsidy are not likely to use taxis As

mentioned before they prefer to use their car A higher number of trips in a day are also a deterrent

to considering taxi usage which may be due to increased cost for more trips Results show that an

individual with more daily trips avoids using taxis Commuters who are employed in more than one

workplace (Pattern 3) are not likely to use taxis This may be due to the fact that they have a lower

level of income which forces them to dedicate more time on the job

Initial results showed that individuals who stated that their car usage is due to poor public transit

service (Poor_PT) are not likely to use taxis This result was far from our expectations By introducing

to this variable the number of household cars as a proxy for household income the model shows

that the previously mentioned individuals who have higher income (Poor_PTCar1+) are the ones

who are not likely to consider taxis Furthermore because such individuals are not considering any

other modes they may treat taxi usage as a kind of PT mode with poor service

As expected greater access to cars in a household (Car_acc) lessens the likelihood of considering

taxis as an alternative Furthermore individuals in households with more motorcycle ownership are

less likely to consider taxis It seems like there is a competition among motorcycles and taxis for

18

access to the city center Younger commuters are less likely to use taxis and individuals with at least

master degrees do consider this mode in addition to their car

Public transit accessed by Drive (DampR)

This mode is affected by the simultaneous interaction of transit time and transit access

(PT_TimeampAccess) which is reflected in the fact that individuals prefer to use this mode for longer

trips Comparing this mode and WampR the first trip start time affects the consideration of this mode

differently Later morning commuters prefer to use their car to access PT modes Such commuters

may have higher income levels or managerial jobs Obviously individuals who are not able to access

PT stations by walking (PTnwacc) are likely to use DampR Serving passengers in daily trips is also a

deterrent in considering this mode which is similar to WampR but with a lower coefficient

Commuters with higher income levels who depend on their car during before or after work

(Dependencycar1+) are likely to use DampR Individuals who use their own car are less likely to use

this mode which is unexpected As a city center develops better PT network coverage and residents

have smaller distances to their workplaces they are unlikely to use DampR This is proven in the model

by a negative sign for D_home_place

Motorcycle (MC)

Increasing fuel cost and cordon pricing simultaneously discourage motorcycle usages Although fuel

cost is expected to reduce motorcycle usage to some extent its combined effect with cordon pricing

also reduces motorcycle usage However this variable is not as strong as other policy variables

=10)

Of the studied modes motorcycle usage is affected by the most commuting variables This may be

due to the fact that this mode is not common Commuting distance has a negative effect on

motorcycle usage which is expected It is worth noting that trip distance appears only in this mode

which may be a reflection of the role of distance in regards to the safety risk in considering this

19

mode Commuters with more stops to serve passengers while commuting (Pattern 2) are not likely

to use this mode which may be due to the poor passenger service of this mode

Individuals who state that commuting is independent of the mode (Pnocarwk) are not likely to use

MC By looking at the (First_Nacco) negative sign this could stem from the fact that the more

passengers there are on the first trip the less likely individuals are to consider MC Regarding the

low capacity of MC and its safety concerns such commuters avoid using this mode Commuters who

pay more parking charges (Park_payment) are less likely to use MC which is expected Individuals

who are dependent on their car during before or after their work time are not likely to use MC

even if they have lower levels of income (DependencyCar1) Individuals who use their own car

(D_car_own) are less likely to use this mode As expected individuals who live in households with

more motorcycle ownership are more likely to use this mode The positive sign of (Permission)

indicates that commuters who have permission to enter the study area do consider MC Because

such commuters generally provide that permission just for car usage this result is unexpected

As with commute variables of all the studied modes MC is affected by the greatest number of

socio-economic variables As expected young commuters (Agelt30) are more likely to use this mode

Commuters with Bachelor of Science degree are less likely to use this mode among others Full time

employees (Emp_full) are less likely to consider MC whereas commuters with more experience in

their jobs prefer to use it Results show that individuals who live in a household with children

younger than 18 are more likely to consider using a car

Tel-Taxi (T_T)

Results show that cordon pricing causes higher probability of using T_T In fact individuals who use

T_T as a mode with similar level of service as cars9 are more willing to pay the cost and make use of

the mode It is worth noting that the effect of cordon pricing in pushing commuters from car usage

9 As this mode does not have driving stress and parking search time in some cases it may have more amount of utility thana car does

20

(000045) is greater than its effect on pulling them to Tel-taxi (000019) This is because of the

possibility of considering other non-car modes

Because consideration of this mode is a function of its operation travel time (Trip_time) appears as

a deterrent in this mode utility function Table 5 shows that individuals are more sensitive to the trip

time when using T_T mode versus taxi which is expected due to their relative costs

The greater the number of full time employees in a family (Nhempfull) the higher the probability of

considering T_T by its commuters which may be due to the higher income level of these

households This is verified by the greater likelihood of using T_T rather than taxis by such

commuters Individuals with higher levels of income who depend on their car during before or after

work time are less likely to use T_T Commuters with lower income levels who state that they use

their car for the sake of comfort (Comfortcar1) are less likely to use T_T which may be due to its

cost Although such individuals do not consider any other modes they specifically avoid T_T Greater

access to cars in a household leads to greater likelihood of T_T usage which could be due to the

higher income level of a household As mentioned before such individuals even avoid taxis

Females who drive to their workplace are more likely to use T_T It seems like this part of society

considers this mode when desiring to avoid the difficulties of driving Younger commuters are less

likely to use T_T and individuals between 30 and 39 years of age are specifically avoiding this mode

Results show that university graduated commuters are more likely to use this mode

6 Marginal effects

To explore the effects of each policy on mode choice and to answer the second issue raised at the

beginning of this paper the marginal effects approach can be adopted Although the coefficients of

the models utility functions show the drivers behavior when facing one or more policies the

marginal effects of policies or their interactions may appropriately show the results of their

implementation More specifically the marginal effect for this study is interpreted as the change in

21

probability given a unit change in a variable ceteris paribus In this section the variable is defined as

a specific policy or a specific interaction of two policies Table 6 presents the marginal effects of the

studied policies and their interactions with mode choice The results are shown in the form of trip

percentages transferred away from the car to the studied modes and the probability-weighted

sample enumeration approach is adopted to find the values It is worth noting that this table is fully

compatible with Table 5 but the marginal effects that were less significant than 90 percent have

been removed

Table 6 - Marginal effects of policies (percent)

Tel-Taxi(TT)

Motorcycle(MC)

Drive ampRide(DampR)

Taxi (T)Walk ampRide(WampR)

Car (C)Mode

Variable-000088Cordon-000140Parking

-09069Access-0000001ParkampFuel

00040PT_TimeampAccess

Table 5 shows that a 1 Rial increase in cordon pricing causes a 000088 percent decrease in car

usage Furthermore a 1 Rial per hour increase in parking cost decreases the car usage probability to

00014 percent By assuming 8 hours for the average parking duration the daily marginal value of

parking cost converts to 000018 percent These values show that cordon pricing is more effective in

forcing individuals not to use their car than increasing parking cost with the same value Results also

show that a 1 Rial increase in parking cost and fuel cost simultaneously decreases the probability of

choosing car usage by 0000001 percent Table 5 also shows that a 1 minute decrease in transit

access time would result in a 09 percent increase in probability of choosing this mode It also shows

that each 1 minute decrease in transit travel time and transit access leads to a 0004 decrease in the

probability of choosing the DampR mode

22

The above results show that just one of the policy interactions is not significantly different from

none (ie CordonampFuel) and push policies have a significant role in impelling car users to consider

other modes

7 Planning

To demonstrate the application of the model some of its planning aspects for the city of Tehran are

presented here In fact the model is able to assess the effect of all studied variables which were

grouped into three tiers on commuters mode choice In this section we assess the effect of TDM

policies by adopting their pair-wise combinations for the city Table 6 shows that pull policies in this

study did not impose a major effect on car usage Because the average access time of the sample

was 11 minutes improving it by 50 percent (about 5 minutes) increases the probability of WampR to

509069100 which is equal to a 048 percent increase in considering this mode (and decrease in all

other modes) On the other hand decreasing transit time about 50 percent (about 19 minutes based

on sample average) and similarly improving access time decreases the probability of considering

DampR to 0004195100 which is equal to a 038 percent decrease in considering this mode (and

increase in all other modes) Therefore in this section we focused on the push policies which

appear in the utility function of car mode It is worth noting that in assessing each combination of

policies all other variables were assumed to be fixed as the current state

71 Parking pricing and cordon pricing

The average hourly parking charge of individuals in the sample is 71 Rials This charge is 3000 Rials

per hour in the curb lane on some streets about 2000 Rials per hour in parking lots and free in

alleys and other streets The average amount shows that most of the commuters benefited from

free parking spaces in alleys and non-controlled streets Figure 2 depicts the result of implementing

this policy and the cordon pricing policy simultaneously In line with marginal effects it can be seen

that in implementing each policy separately cordon pricing is more effective than parking pricing in

the planning range It is worth noting that commuters in this study are less sensitive to parking

23

pricing than cordon pricing One reason may be that they thought in any case they would find free

parking space Figure 2 shows that the effect of each policy depends on the level of the other policy

In other words the effect of the simultaneous implementation of two policies is not equal to the

sum of their separate individual effects

Figure 2- Effect of implementing parking pricing and cordon pricing

72 Parking pricing and increasing fuel cost

Figure 3 shows the simultaneous effects of parking pricing and increasing fuel cost policies Although

both of these policies are not effective separately their simultaneous effect is significant It can be

seen that their synergy appears in the higher levels of both policies

Figure 3 - Effect of implementing parking pricing and increasing fuel cost

24

73 Cordon pricing and increasing fuel cost

The effect of implementing cordon pricing and increasing fuel cost is shown in Figure 4 It can be

seen that these two policies have no interaction effect in decreasing car usage This result is verified

by this interactions insignificant marginal value in Table 6

Figure 4- Effect of implementing cordon pricing and increasing fuel cost

8 Conclusions

This study examined the role of TDM policies in individual mode choice for work trips in the city of

Tehran Five policies including increasing parking cost increasing fuel cost cordon pricing in the odd-

even zone of the city transit time reduction and transit access improvement were investigated Of

the five policies the former three were push policies and the latter two were pull policies The

design of experiments approach was used to design the questionnaire that would capture the stated

preferences of car commuters

A number of nested structures based on recognizing differences in the variances associated

with unobserved influences were examined and finally the MNL as a superior model for this study

was developed

25

The results show that for car users who faced TDM policies all of the five studied TDM

policies were factors in the mode choice process The model shows that in considering car usage

push policies play a main role and pull policies do not

Results also show that the interactions of policies are also significant The model shows that

the interaction of parking cost policy and fuel cost policy is significant in prompting car users to

consider other modes It also shows that the interaction of cordon pricing and increase in fuel cost is

effective to discourage motorcycle usage Furthermore the interaction of transit time reduction

policy and access improvements policy is significant in considering DampR as a transportation mode to

workplace

Pull policies in the study were expected to attract individuals to transit modes (pull policies

of this study are related to transit modes) This was verified by the model

Results show that although taxi usage is usually treated as an alternative for urban trips it

was not directly affected by the studied policies This may be due to its special function in Tehran as

a non-public and non-private mode

The results also show that except for the interaction of cordon pricing and increase in fuel

cost other interaction effects have significant marginal effects on mode choice

Results of the model show that to implement a single policy cordon pricing is the most

effective in decreasing car usage

This study assessed workplace commuters Thus future studies should explore the

individuals mode choice with other trip aims Additionally including more policies especially pull

policies in favor of non-transit modes is reasonable to rigorously assess the model

It can be seen that the studied modes are affected differently by the TDM policies and their

interactions The impact of the interaction of policies appeared in the consideration of three out of

six studied modes and also in two out of three developed graphs This issue needs more attention

within the transportation demand management context and further studies should be performed

26

Design of this study is based on the consideration of all two-way interactions with some

degrees of deficiency in D-efficient policy Focusing on the significant interactions of this study is

recommended to perform other studies with more D-efficient values

Further understanding of the impacts of TDM policies might be provided in future research

by focusing on the amount of their synergy in prohibiting car usage

9 References

Arentze TA Hofman F amp Timmermans HJP (2004) Predicting multi-faceted activityndashtravel adjustmentstrategies in response to possible congestion pricing scenarios using an Internet-based stated adaptationexperiment Transport Policy 11 p31ndash41

Burris M amp Patil S (2009) Estimating the Benefits of Managed Lanes Technical Report Texas TexasTransportation Institute University Transportation Center for Mobility

Cao X amp Mokhtarian PL (2005) How do individuals adapt their personal travel A conceptual exploration ofthe consideation of travel-related strategies Transport Policy 12(3) pp199-206

de Palma A amp Lindsey R (2001) Transportation Supply and congestion In International Encyclopedia of theSocial and Behavioral Sciences 1st ed Elsevier p15882ndash15888

Eriksson L Garvill J amp Nordlund AM (2008) Acceptability of single and combined transport policy measuresThe importance of environmental and policy specific beliefs Transportation Research Part A 42 p1117ndash1128

Eriksson L Nordlund AM amp Garvill J (2010) Expected car use reduction in response to structural traveldemand management measures Transportation Research Part F 13 p329ndash342

Golob TF amp Hensher DA (2007) The Trip chaining Activity of Sydney Residents A Cross-Section Assessmentby Age Group with a Focus on Seniors Journal of Transport Geography 15 p298ndash312

Hensher DA amp King J (2001) Parking demand and responsiveness to supply pricing and location in theSydney central business district Transportation Research A 35 pp177-96

Hensher DA amp Rose JM (2007) Development of commuter and non-commuter mode choice models for theassessment of new public transport infrastructure projects A case study Transportation Research Part A41 p428ndash443

Hensher DA Rose JM amp Greene WH (2005) Applied Choice Analysis A Primer New York CambridgeUniversity Press

Iranian Center of Statistics (ICS) (2009) Information of Iranian States [Online] Available at HYPERLINK httpwwwamarorgirUploadModulesContentsasset16tehrantehparthtmlhttpwwwamarorgirUploadModulesContentsasset16tehrantehparthtml [Accessed 13 November2009]

Kingham S Dickinson J amp Copsey S (2001) Travelling to work Will people move out of their cars TransportPolicy 8 pp151-60

Kuhfeld WF (2009) Marketing Research Methods in SAS Experimental Design Choice Conjoint andGraphical Techniques SAS 92 Edition NC USA SAS Institute Inc Cary Available on the Webwwwsupportsascomtechsuptnotetnote_stathtmlmarket

Litman T (2003) The online TDM Encyclopedia mobility management information gateway Transport Policy10 pp245-49

Litman T (2010) Online TDM Encyclopedia [Online] Available at HYPERLINK httpwwwvtpiorghttpwwwvtpiorg [Accessed 21 Augest 2010]

9

main effects with only 36 choice sets4 (See (Rose amp Bliemer 2009) or (Kuhfeld 2009) for more

details on efficient design) To avoid a time-consuming questionnaire 36 choice sets (scenarios)

were randomly ordered and divided into six separate questionnaire types coded as 1 to 6 Each of

the questionnaires had six scenarios and each scenario consisted of five policies

4 Survey

Two push policies are currently being implemented in the city of Tehran The first is car-free

planning in the CBD area of the city and the second one is an odd-even scheme based on the last

digit of car plates that attempt to enter a zone which is about three times larger than and includes

the CBD area Based on their occupation a few people can drive to the CBD area with a license

called permission A stated preferences survey was assigned for the morning car commuters to the

odd-even zone but they were asked to ignore these two policies to find the accurate sensitivity of

individuals to the study policies The odd-even zone is selected as study area for the two following

reasons 1) because of odd-even control respondents are familiar with the fringes and they can

better imagine the entrance pricing area and 2) respondents are familiar with the limits that they

face half of the week and are thus aware of the alternative existing modes Compared to the CBD

area this zone covers more car commuters and the entrance restriction is more imaginable for this

zone than the former one Respondents were interviewed face-to-face in their workplaces midway

through the year 2009 The interviews were enhanced with a special card to better define the

scenarios

For this study 2196 scenario observations from 366 individuals were adopted The sample included

308 men (ie 841) and 58 women (ie 159) The figures are close to the employment

percentages in the city according to the Iranian Center of Statistics (ICS) This source indicates that

825 of Tehran employees are men and 175 are women (Iranian Center of Statistics (ICS) 2009)

Because this study focuses on car-using commuters comparisons between the sample and city data

4 Efficient design is also adopted in other studies such as managed lanes (Burris amp Patil 2009)

10

especially regarding educational distribution were impossible Table 2 presents demographics of the

sample

Table 2- Demographics gender marital household (HH) size employee type age HH employee(s)

Amount Percent

Gender Male 308 841Female 58 159

Marital Single 100 273Married 266 727

HH Size 1 4 112 86 2353 129 3524 90 2465 42 1156+ 15 41

Age 18~29 122 33330~39 146 39940~49 58 15950~59 32 8760+ 8 22

HH employee(s) 1 156 4262 159 4343 41 1124+ 10 27

The first part of the questionnaire is dedicated to gathering the occupation state home and job

locations the distance between these locations round-trip time (from home to workplace and then

workplace to home) and all car trip characteristics in the previous day or the day before it based on

plate number It was necessary that the respondents drive hisher car in the day studied to complete

the trip diary portion of the questionnaire5 The general reasons for car usage and the scenarios

formed the next portion In each scenario every respondent was asked the question How would

you travel to the workplace if all of these changes were in place on the day studied For example

one may have to pay 4000 Rialsh for parking 50000 Rials per entrance to the cordon the same

amount in transit access and fuel cost and a 15 percent decrease in transit time simultaneously

Depending on individual responses six main options were distinguished6 These choices were still

5 In designing the questionnaire the general form of questionnaire which has mentioned in OFallons study was adopted6 In the pre-test survey 14 modes is distinguished

11

drive a car (C) walk to the station and catch public transit (WampR) drive to a public station and catch

public transit (DampR) ride a motorcycle (MC) catch a taxi7 (T) and catch a taxi by phone (T_T) DampR is

somewhat different than the more familiar ldquoPark amp Riderdquo In fact in the fringes of the odd-even

zone there were no specialized parking lots dedicated to this purpose and commuters considered

Drive amp Ride because they were not allowed to pass the fringes

After each scenario if the respondent changed hisher mode the reason(s) for the change were

asked It could be a sole policy or a bundle of them Furthermore travel-related information was

sought These data were not part of the stated choice but they might have important influences on

individual choices These data consisted of car dependency (need to drive someone or move freight

in the trip) parking place type and average weekly parking costs car and motorcycle ownership and

number of household driving licenses

Depending on the individuals activity in that day three types of activity patterns were detected

Pattern 1 described individuals who had no stop in their commute Pattern 2 was for individuals who

had at least one stop on their way to or from work and pattern 3 was for the individuals who went

to another workplace in their daily activities

Finally for the sake of data generalization and the examination of household characteristics gender

age and household type employment status and education level were also asked

5 Mode choice model

In order to detect the policies that affect individual mode choice the logit modeling approach was

adopted In this model one can determine if the interaction of two policies affects the mode choice

In the calibration step 152 variables were defined and their effects on consideration of each mode

were examined

7 Taxis in Iran are somewhat different than taxis in other countries of the world In fact taxis in Iran are not hiring by oneperson or a group of people at a time Taxis allow passengers to board or alight along their path with respect to theircapacity In other word this mode is functioning similar to transit vehicles but the stops are not predefined

12

51 Model structure

Initially a multinomial logit (MNL) model is developed (Figure 1a) By selecting a number of tree

structures based on recognizing differences in the variances associated with unobserved influences

we find that the greatest similarity in variance profiles is associated with public transport modes as

opposed to non-public modes (Figure 1b) This structure has two nests one including Car (C) and

Motorcycle (MC) as private modes and the other including Walk and Ride (WampR) Drive and Ride

(DampR) Taxi (T) and Tel-taxi (T_T) as non-private modes The result of this nested logit (NL) model is

shown in Table 3

Although it is not a statistically significant improvement overall on the MNL model the statistically

significant inclusive value8 (IV) of 0889 for non-public modes relative to the fixed parameter value of

10 for public modes suggests that there is a structural advantage in selecting the NL specification

The normal test of a statistically significant difference between NL and MNL is an IV parameter

relative to 10 calculated using a Wald-test via equation 1

)1(Wald-test = (IVparameter ndash 1)std error

a The MNL structure

b Final nested structure

Figure 1- Model structure

8 Also called scale parameter

Alternatives

MCCar T_TT WampR DampR

Alternatives

MCCar T_TT

Public

WampR DampR

Private

13

We have (0889-1)2508 =-075 which would be rejected at the usual acceptable significance levels

This suggests that the NL model could be collapsed into an MNL form

Table 3- Nested logit (NL) model resultValueParameter

0889IV (Private)1000IV (nPrivate)

-2668335L( )-4057684L(0)

0342sup2

After the calibration process the variables that were statistically significant were identified and are

presented in Table 4 Table 5 presents the final model of the study with a goodness of fit of 031 for

the six studied modes For a general review of the model calibration results the effective factors can

be grouped under the following three categories TDM policy characteristics commuting trip

characteristics and household socio-economic characteristics which are all treated as alternative-

specific variables

52 Model results

Car (C)

It is expected that push policies impel car-drivers to choose other modes Table 5 shows that cordon

pricing and increase in parking cost cause individuals to choose not to use their car This is in line

with other studies suggesting that these policies are effective to discourage car usage (Hensher amp

Rose 2007 OFallon et al 2004) In addition the interaction between the policies of fuel cost

increase and increase in parking cost shows similar car usage discourage effect Because fuel cost is

related to the distance between home and work locations and parking cost is related to work time

the time that an individual spends out of the home is negatively affected by hisher likelihood to use

a car

14

Table 4 - Definition of the significant variables

AbbreviationVariableTransportation demand management measures

Measures

ParkingParking cost increase Rials per hour

CordonCordon price Rials per entranceAccessTransit access time shortage percent

Interaction of push measures

ParkampFuelParking cost and fuel cost simultaneous effectsCordonampFuelCordon pricing and fuel cost simultaneous effects

Interaction of pull measures

PT_timeampaccessPT time reduction and access improvement simultaneous effectsCommuting trip characteristics

Trip distanceDistance between home and workplaceTrip timeTravel time between home and workplace

Exp FuelLikelihood of unsubsidized fuel use (self-reported on a Likert scale)NtripsNumber of daily tripsPattern2Commuting with 1+ stop(s) in go or return

Pattern3Commuting with 2 workplacesFirst trip timeStart time of first trip

PnocarwkLikelihood of going to work in absence of that car (self-reported)PTnwaccNon-walk access to transit (yes=1)First NaccoNumber of passengers in first trip

PassengerAny passenger on that day (yes=1)Park_paymentParking payment in last weekNhempfullNumber of full employees in HH

CardependencyBoardalight a passenger or move freight in the trip (yes=1)D car ownBe the owner of the used vehicle (yes=1)

Car accCar accessibility in household (number of cars to number of HH driving licenses ratio)NmotorcycleNumber of motorcycles owned by HHD home placeHome Location is in study area (yes=1)

PermissionPermission to enter to study area (yes=1)ComfortI use my car because it is comfortablePoor_PTI use my car because transit is not good

HH socio-economic characteristics

FemaleGender (Female=1)Age lt30Age younger than 30 (yes=1)Age 30_39Age between 30 to 39 (yes=1)Job_durationNumber of years that individual has been at hisher job

Emp_fullFull-time employee (yes=1)Edu BSDegree of education is BSc (yes=1)Edu BS+Degree of education is higher than BSc(yes=1)

D childlt=18Child younger than 18 in HH (yes=1)

15

Table 5 ndash The mode choice model

Tel-Taxi(T_T)

Motorcycle(MC)

Drive amp Ride(DampR)

Taxi (T)Walk amp Ride(WampR)

Car (C)Mode

Variable-471756-37067-147911Constant

Transportation demand management measure variables00019-00045Cordon

-000072Parking-004308Access

-28443D-05Parkampfuel-32475D-06Cordonampfuel

00029Pt_timeampaccess

Commuting trip characteristics-04709Trip distance

-02163-00831Trip time-96755163655Exp fuel-16253Ntrips

-114779Pattern2-71008Pattern3

00282-00270First trip time-02439-01549Pnocarwk

-11322992883-32765PTnwacc-133701First Nacco

-7778-73782Accompany-00049000010Park_payment

201646195554Nhempfull-160144ComfortCar1

-206142DependencyCar1-16101883385-121224DependencyCar2

42176Poor_PTCar1-24988Poor_PTCar2

- -27221D car own70960-39136Car acc

1 -71112-156123Nmotorcycle-1436322762D home place

2 78826Permission

HH socio-economic characteristics149490Female

297584-24548Agelt30-136490Age30_39

079430366303585Job_duration-108743Emp_full-203468-64900Edu BS

10932856687-4499984445Edu BS+102271D childlt=18

-2677366L( )-3849556L(0)0305sup2

112127178592580607N

Note = Positive significance at 1 5 10 level

As expected individuals with higher income are more likely to use their car This is indicated in the

model by the positive signs of individuals who use fuel with fixed (unsubsidized) cost and individuals

16

who pay more in parking charges in the previous week of study Negative sign of Pnocarwk variable

shows that the commuters who stated that their commute depends on car availability are more

likely to use their car Individuals in households with more full-time employees are more likely to use

their car which may be the result of higher household income Not surprisingly commuters who

have permission are more likely to maintain car usage Among the household socio-economic

parameters greater job experience (Job_duration) and higher graduate levels (EduBS+) increase the

probability of car usage

Public transit accessed by walking (WampR)

Access time to transit negatively impacts WampR choice which is expected This result is similar to

findings for the city of Sydney (Hensher amp Rose 2007) The negative coefficient of first trip time

indicates that individuals are more likely to use WampR in the early morning This result seems to

reflect the better weather for walking and faster speed of WampR mode early in the morning

Obviously individuals who are not able to access transit stations via walking (PTnwacc) are less likely

to consider this mode Furthermore serving passengers on daily trips is also a deterrent to using

WampR

Initially assessing the individuals who stated that their car usage is due to poor public transit service

(Poor_PT) led to an unexpected result in favor of considering WampR By introducing to this variable

the number of household cars as a proxy for household income (Poor_PTCar1) the model shows

that of the previously mentioned individuals those who have lower income are the ones who have

to consider WampR The result is understandable as these individuals may have no alternative when

they have to change their mode (they also are not likely to consider other modes) Individuals with

higher levels of income who have to use their car during before or after work (Dependencycar1+)

are not likely to use WampR

The greater the number of motorcycles in a household the less likely commuters is to consider

WampR There appears to be a competition between motorcycle and PT for access to the city center

17

Better PT services in the center of the city in terms of coverage and frequency increases the

likelihood that its residents will consider WampR This is verified by the positive sign of the

D_home_place variable Commuters with greater job experience (Job_duration) in their workplace

are more likely to use this mode Although individuals with higher levels of education are not likely

to use WampR as education level increases avoidance of WampR decreases

Taxi (T)

Table 5 shows that none of the studied policies are significant in considering taxi usage It seems that

taxi usage considering its function in Iran as a non-private and non-public mode of transport is not

affected by pull or push policies A negative sign for taxi travel time indicates that individuals are not

likely to use this mode for longer trips This seems reasonable given that longer trips are more

expensive Commuters who are more likely to use fuel with no subsidy are not likely to use taxis As

mentioned before they prefer to use their car A higher number of trips in a day are also a deterrent

to considering taxi usage which may be due to increased cost for more trips Results show that an

individual with more daily trips avoids using taxis Commuters who are employed in more than one

workplace (Pattern 3) are not likely to use taxis This may be due to the fact that they have a lower

level of income which forces them to dedicate more time on the job

Initial results showed that individuals who stated that their car usage is due to poor public transit

service (Poor_PT) are not likely to use taxis This result was far from our expectations By introducing

to this variable the number of household cars as a proxy for household income the model shows

that the previously mentioned individuals who have higher income (Poor_PTCar1+) are the ones

who are not likely to consider taxis Furthermore because such individuals are not considering any

other modes they may treat taxi usage as a kind of PT mode with poor service

As expected greater access to cars in a household (Car_acc) lessens the likelihood of considering

taxis as an alternative Furthermore individuals in households with more motorcycle ownership are

less likely to consider taxis It seems like there is a competition among motorcycles and taxis for

18

access to the city center Younger commuters are less likely to use taxis and individuals with at least

master degrees do consider this mode in addition to their car

Public transit accessed by Drive (DampR)

This mode is affected by the simultaneous interaction of transit time and transit access

(PT_TimeampAccess) which is reflected in the fact that individuals prefer to use this mode for longer

trips Comparing this mode and WampR the first trip start time affects the consideration of this mode

differently Later morning commuters prefer to use their car to access PT modes Such commuters

may have higher income levels or managerial jobs Obviously individuals who are not able to access

PT stations by walking (PTnwacc) are likely to use DampR Serving passengers in daily trips is also a

deterrent in considering this mode which is similar to WampR but with a lower coefficient

Commuters with higher income levels who depend on their car during before or after work

(Dependencycar1+) are likely to use DampR Individuals who use their own car are less likely to use

this mode which is unexpected As a city center develops better PT network coverage and residents

have smaller distances to their workplaces they are unlikely to use DampR This is proven in the model

by a negative sign for D_home_place

Motorcycle (MC)

Increasing fuel cost and cordon pricing simultaneously discourage motorcycle usages Although fuel

cost is expected to reduce motorcycle usage to some extent its combined effect with cordon pricing

also reduces motorcycle usage However this variable is not as strong as other policy variables

=10)

Of the studied modes motorcycle usage is affected by the most commuting variables This may be

due to the fact that this mode is not common Commuting distance has a negative effect on

motorcycle usage which is expected It is worth noting that trip distance appears only in this mode

which may be a reflection of the role of distance in regards to the safety risk in considering this

19

mode Commuters with more stops to serve passengers while commuting (Pattern 2) are not likely

to use this mode which may be due to the poor passenger service of this mode

Individuals who state that commuting is independent of the mode (Pnocarwk) are not likely to use

MC By looking at the (First_Nacco) negative sign this could stem from the fact that the more

passengers there are on the first trip the less likely individuals are to consider MC Regarding the

low capacity of MC and its safety concerns such commuters avoid using this mode Commuters who

pay more parking charges (Park_payment) are less likely to use MC which is expected Individuals

who are dependent on their car during before or after their work time are not likely to use MC

even if they have lower levels of income (DependencyCar1) Individuals who use their own car

(D_car_own) are less likely to use this mode As expected individuals who live in households with

more motorcycle ownership are more likely to use this mode The positive sign of (Permission)

indicates that commuters who have permission to enter the study area do consider MC Because

such commuters generally provide that permission just for car usage this result is unexpected

As with commute variables of all the studied modes MC is affected by the greatest number of

socio-economic variables As expected young commuters (Agelt30) are more likely to use this mode

Commuters with Bachelor of Science degree are less likely to use this mode among others Full time

employees (Emp_full) are less likely to consider MC whereas commuters with more experience in

their jobs prefer to use it Results show that individuals who live in a household with children

younger than 18 are more likely to consider using a car

Tel-Taxi (T_T)

Results show that cordon pricing causes higher probability of using T_T In fact individuals who use

T_T as a mode with similar level of service as cars9 are more willing to pay the cost and make use of

the mode It is worth noting that the effect of cordon pricing in pushing commuters from car usage

9 As this mode does not have driving stress and parking search time in some cases it may have more amount of utility thana car does

20

(000045) is greater than its effect on pulling them to Tel-taxi (000019) This is because of the

possibility of considering other non-car modes

Because consideration of this mode is a function of its operation travel time (Trip_time) appears as

a deterrent in this mode utility function Table 5 shows that individuals are more sensitive to the trip

time when using T_T mode versus taxi which is expected due to their relative costs

The greater the number of full time employees in a family (Nhempfull) the higher the probability of

considering T_T by its commuters which may be due to the higher income level of these

households This is verified by the greater likelihood of using T_T rather than taxis by such

commuters Individuals with higher levels of income who depend on their car during before or after

work time are less likely to use T_T Commuters with lower income levels who state that they use

their car for the sake of comfort (Comfortcar1) are less likely to use T_T which may be due to its

cost Although such individuals do not consider any other modes they specifically avoid T_T Greater

access to cars in a household leads to greater likelihood of T_T usage which could be due to the

higher income level of a household As mentioned before such individuals even avoid taxis

Females who drive to their workplace are more likely to use T_T It seems like this part of society

considers this mode when desiring to avoid the difficulties of driving Younger commuters are less

likely to use T_T and individuals between 30 and 39 years of age are specifically avoiding this mode

Results show that university graduated commuters are more likely to use this mode

6 Marginal effects

To explore the effects of each policy on mode choice and to answer the second issue raised at the

beginning of this paper the marginal effects approach can be adopted Although the coefficients of

the models utility functions show the drivers behavior when facing one or more policies the

marginal effects of policies or their interactions may appropriately show the results of their

implementation More specifically the marginal effect for this study is interpreted as the change in

21

probability given a unit change in a variable ceteris paribus In this section the variable is defined as

a specific policy or a specific interaction of two policies Table 6 presents the marginal effects of the

studied policies and their interactions with mode choice The results are shown in the form of trip

percentages transferred away from the car to the studied modes and the probability-weighted

sample enumeration approach is adopted to find the values It is worth noting that this table is fully

compatible with Table 5 but the marginal effects that were less significant than 90 percent have

been removed

Table 6 - Marginal effects of policies (percent)

Tel-Taxi(TT)

Motorcycle(MC)

Drive ampRide(DampR)

Taxi (T)Walk ampRide(WampR)

Car (C)Mode

Variable-000088Cordon-000140Parking

-09069Access-0000001ParkampFuel

00040PT_TimeampAccess

Table 5 shows that a 1 Rial increase in cordon pricing causes a 000088 percent decrease in car

usage Furthermore a 1 Rial per hour increase in parking cost decreases the car usage probability to

00014 percent By assuming 8 hours for the average parking duration the daily marginal value of

parking cost converts to 000018 percent These values show that cordon pricing is more effective in

forcing individuals not to use their car than increasing parking cost with the same value Results also

show that a 1 Rial increase in parking cost and fuel cost simultaneously decreases the probability of

choosing car usage by 0000001 percent Table 5 also shows that a 1 minute decrease in transit

access time would result in a 09 percent increase in probability of choosing this mode It also shows

that each 1 minute decrease in transit travel time and transit access leads to a 0004 decrease in the

probability of choosing the DampR mode

22

The above results show that just one of the policy interactions is not significantly different from

none (ie CordonampFuel) and push policies have a significant role in impelling car users to consider

other modes

7 Planning

To demonstrate the application of the model some of its planning aspects for the city of Tehran are

presented here In fact the model is able to assess the effect of all studied variables which were

grouped into three tiers on commuters mode choice In this section we assess the effect of TDM

policies by adopting their pair-wise combinations for the city Table 6 shows that pull policies in this

study did not impose a major effect on car usage Because the average access time of the sample

was 11 minutes improving it by 50 percent (about 5 minutes) increases the probability of WampR to

509069100 which is equal to a 048 percent increase in considering this mode (and decrease in all

other modes) On the other hand decreasing transit time about 50 percent (about 19 minutes based

on sample average) and similarly improving access time decreases the probability of considering

DampR to 0004195100 which is equal to a 038 percent decrease in considering this mode (and

increase in all other modes) Therefore in this section we focused on the push policies which

appear in the utility function of car mode It is worth noting that in assessing each combination of

policies all other variables were assumed to be fixed as the current state

71 Parking pricing and cordon pricing

The average hourly parking charge of individuals in the sample is 71 Rials This charge is 3000 Rials

per hour in the curb lane on some streets about 2000 Rials per hour in parking lots and free in

alleys and other streets The average amount shows that most of the commuters benefited from

free parking spaces in alleys and non-controlled streets Figure 2 depicts the result of implementing

this policy and the cordon pricing policy simultaneously In line with marginal effects it can be seen

that in implementing each policy separately cordon pricing is more effective than parking pricing in

the planning range It is worth noting that commuters in this study are less sensitive to parking

23

pricing than cordon pricing One reason may be that they thought in any case they would find free

parking space Figure 2 shows that the effect of each policy depends on the level of the other policy

In other words the effect of the simultaneous implementation of two policies is not equal to the

sum of their separate individual effects

Figure 2- Effect of implementing parking pricing and cordon pricing

72 Parking pricing and increasing fuel cost

Figure 3 shows the simultaneous effects of parking pricing and increasing fuel cost policies Although

both of these policies are not effective separately their simultaneous effect is significant It can be

seen that their synergy appears in the higher levels of both policies

Figure 3 - Effect of implementing parking pricing and increasing fuel cost

24

73 Cordon pricing and increasing fuel cost

The effect of implementing cordon pricing and increasing fuel cost is shown in Figure 4 It can be

seen that these two policies have no interaction effect in decreasing car usage This result is verified

by this interactions insignificant marginal value in Table 6

Figure 4- Effect of implementing cordon pricing and increasing fuel cost

8 Conclusions

This study examined the role of TDM policies in individual mode choice for work trips in the city of

Tehran Five policies including increasing parking cost increasing fuel cost cordon pricing in the odd-

even zone of the city transit time reduction and transit access improvement were investigated Of

the five policies the former three were push policies and the latter two were pull policies The

design of experiments approach was used to design the questionnaire that would capture the stated

preferences of car commuters

A number of nested structures based on recognizing differences in the variances associated

with unobserved influences were examined and finally the MNL as a superior model for this study

was developed

25

The results show that for car users who faced TDM policies all of the five studied TDM

policies were factors in the mode choice process The model shows that in considering car usage

push policies play a main role and pull policies do not

Results also show that the interactions of policies are also significant The model shows that

the interaction of parking cost policy and fuel cost policy is significant in prompting car users to

consider other modes It also shows that the interaction of cordon pricing and increase in fuel cost is

effective to discourage motorcycle usage Furthermore the interaction of transit time reduction

policy and access improvements policy is significant in considering DampR as a transportation mode to

workplace

Pull policies in the study were expected to attract individuals to transit modes (pull policies

of this study are related to transit modes) This was verified by the model

Results show that although taxi usage is usually treated as an alternative for urban trips it

was not directly affected by the studied policies This may be due to its special function in Tehran as

a non-public and non-private mode

The results also show that except for the interaction of cordon pricing and increase in fuel

cost other interaction effects have significant marginal effects on mode choice

Results of the model show that to implement a single policy cordon pricing is the most

effective in decreasing car usage

This study assessed workplace commuters Thus future studies should explore the

individuals mode choice with other trip aims Additionally including more policies especially pull

policies in favor of non-transit modes is reasonable to rigorously assess the model

It can be seen that the studied modes are affected differently by the TDM policies and their

interactions The impact of the interaction of policies appeared in the consideration of three out of

six studied modes and also in two out of three developed graphs This issue needs more attention

within the transportation demand management context and further studies should be performed

26

Design of this study is based on the consideration of all two-way interactions with some

degrees of deficiency in D-efficient policy Focusing on the significant interactions of this study is

recommended to perform other studies with more D-efficient values

Further understanding of the impacts of TDM policies might be provided in future research

by focusing on the amount of their synergy in prohibiting car usage

9 References

Arentze TA Hofman F amp Timmermans HJP (2004) Predicting multi-faceted activityndashtravel adjustmentstrategies in response to possible congestion pricing scenarios using an Internet-based stated adaptationexperiment Transport Policy 11 p31ndash41

Burris M amp Patil S (2009) Estimating the Benefits of Managed Lanes Technical Report Texas TexasTransportation Institute University Transportation Center for Mobility

Cao X amp Mokhtarian PL (2005) How do individuals adapt their personal travel A conceptual exploration ofthe consideation of travel-related strategies Transport Policy 12(3) pp199-206

de Palma A amp Lindsey R (2001) Transportation Supply and congestion In International Encyclopedia of theSocial and Behavioral Sciences 1st ed Elsevier p15882ndash15888

Eriksson L Garvill J amp Nordlund AM (2008) Acceptability of single and combined transport policy measuresThe importance of environmental and policy specific beliefs Transportation Research Part A 42 p1117ndash1128

Eriksson L Nordlund AM amp Garvill J (2010) Expected car use reduction in response to structural traveldemand management measures Transportation Research Part F 13 p329ndash342

Golob TF amp Hensher DA (2007) The Trip chaining Activity of Sydney Residents A Cross-Section Assessmentby Age Group with a Focus on Seniors Journal of Transport Geography 15 p298ndash312

Hensher DA amp King J (2001) Parking demand and responsiveness to supply pricing and location in theSydney central business district Transportation Research A 35 pp177-96

Hensher DA amp Rose JM (2007) Development of commuter and non-commuter mode choice models for theassessment of new public transport infrastructure projects A case study Transportation Research Part A41 p428ndash443

Hensher DA Rose JM amp Greene WH (2005) Applied Choice Analysis A Primer New York CambridgeUniversity Press

Iranian Center of Statistics (ICS) (2009) Information of Iranian States [Online] Available at HYPERLINK httpwwwamarorgirUploadModulesContentsasset16tehrantehparthtmlhttpwwwamarorgirUploadModulesContentsasset16tehrantehparthtml [Accessed 13 November2009]

Kingham S Dickinson J amp Copsey S (2001) Travelling to work Will people move out of their cars TransportPolicy 8 pp151-60

Kuhfeld WF (2009) Marketing Research Methods in SAS Experimental Design Choice Conjoint andGraphical Techniques SAS 92 Edition NC USA SAS Institute Inc Cary Available on the Webwwwsupportsascomtechsuptnotetnote_stathtmlmarket

Litman T (2003) The online TDM Encyclopedia mobility management information gateway Transport Policy10 pp245-49

Litman T (2010) Online TDM Encyclopedia [Online] Available at HYPERLINK httpwwwvtpiorghttpwwwvtpiorg [Accessed 21 Augest 2010]

10

especially regarding educational distribution were impossible Table 2 presents demographics of the

sample

Table 2- Demographics gender marital household (HH) size employee type age HH employee(s)

Amount Percent

Gender Male 308 841Female 58 159

Marital Single 100 273Married 266 727

HH Size 1 4 112 86 2353 129 3524 90 2465 42 1156+ 15 41

Age 18~29 122 33330~39 146 39940~49 58 15950~59 32 8760+ 8 22

HH employee(s) 1 156 4262 159 4343 41 1124+ 10 27

The first part of the questionnaire is dedicated to gathering the occupation state home and job

locations the distance between these locations round-trip time (from home to workplace and then

workplace to home) and all car trip characteristics in the previous day or the day before it based on

plate number It was necessary that the respondents drive hisher car in the day studied to complete

the trip diary portion of the questionnaire5 The general reasons for car usage and the scenarios

formed the next portion In each scenario every respondent was asked the question How would

you travel to the workplace if all of these changes were in place on the day studied For example

one may have to pay 4000 Rialsh for parking 50000 Rials per entrance to the cordon the same

amount in transit access and fuel cost and a 15 percent decrease in transit time simultaneously

Depending on individual responses six main options were distinguished6 These choices were still

5 In designing the questionnaire the general form of questionnaire which has mentioned in OFallons study was adopted6 In the pre-test survey 14 modes is distinguished

11

drive a car (C) walk to the station and catch public transit (WampR) drive to a public station and catch

public transit (DampR) ride a motorcycle (MC) catch a taxi7 (T) and catch a taxi by phone (T_T) DampR is

somewhat different than the more familiar ldquoPark amp Riderdquo In fact in the fringes of the odd-even

zone there were no specialized parking lots dedicated to this purpose and commuters considered

Drive amp Ride because they were not allowed to pass the fringes

After each scenario if the respondent changed hisher mode the reason(s) for the change were

asked It could be a sole policy or a bundle of them Furthermore travel-related information was

sought These data were not part of the stated choice but they might have important influences on

individual choices These data consisted of car dependency (need to drive someone or move freight

in the trip) parking place type and average weekly parking costs car and motorcycle ownership and

number of household driving licenses

Depending on the individuals activity in that day three types of activity patterns were detected

Pattern 1 described individuals who had no stop in their commute Pattern 2 was for individuals who

had at least one stop on their way to or from work and pattern 3 was for the individuals who went

to another workplace in their daily activities

Finally for the sake of data generalization and the examination of household characteristics gender

age and household type employment status and education level were also asked

5 Mode choice model

In order to detect the policies that affect individual mode choice the logit modeling approach was

adopted In this model one can determine if the interaction of two policies affects the mode choice

In the calibration step 152 variables were defined and their effects on consideration of each mode

were examined

7 Taxis in Iran are somewhat different than taxis in other countries of the world In fact taxis in Iran are not hiring by oneperson or a group of people at a time Taxis allow passengers to board or alight along their path with respect to theircapacity In other word this mode is functioning similar to transit vehicles but the stops are not predefined

12

51 Model structure

Initially a multinomial logit (MNL) model is developed (Figure 1a) By selecting a number of tree

structures based on recognizing differences in the variances associated with unobserved influences

we find that the greatest similarity in variance profiles is associated with public transport modes as

opposed to non-public modes (Figure 1b) This structure has two nests one including Car (C) and

Motorcycle (MC) as private modes and the other including Walk and Ride (WampR) Drive and Ride

(DampR) Taxi (T) and Tel-taxi (T_T) as non-private modes The result of this nested logit (NL) model is

shown in Table 3

Although it is not a statistically significant improvement overall on the MNL model the statistically

significant inclusive value8 (IV) of 0889 for non-public modes relative to the fixed parameter value of

10 for public modes suggests that there is a structural advantage in selecting the NL specification

The normal test of a statistically significant difference between NL and MNL is an IV parameter

relative to 10 calculated using a Wald-test via equation 1

)1(Wald-test = (IVparameter ndash 1)std error

a The MNL structure

b Final nested structure

Figure 1- Model structure

8 Also called scale parameter

Alternatives

MCCar T_TT WampR DampR

Alternatives

MCCar T_TT

Public

WampR DampR

Private

13

We have (0889-1)2508 =-075 which would be rejected at the usual acceptable significance levels

This suggests that the NL model could be collapsed into an MNL form

Table 3- Nested logit (NL) model resultValueParameter

0889IV (Private)1000IV (nPrivate)

-2668335L( )-4057684L(0)

0342sup2

After the calibration process the variables that were statistically significant were identified and are

presented in Table 4 Table 5 presents the final model of the study with a goodness of fit of 031 for

the six studied modes For a general review of the model calibration results the effective factors can

be grouped under the following three categories TDM policy characteristics commuting trip

characteristics and household socio-economic characteristics which are all treated as alternative-

specific variables

52 Model results

Car (C)

It is expected that push policies impel car-drivers to choose other modes Table 5 shows that cordon

pricing and increase in parking cost cause individuals to choose not to use their car This is in line

with other studies suggesting that these policies are effective to discourage car usage (Hensher amp

Rose 2007 OFallon et al 2004) In addition the interaction between the policies of fuel cost

increase and increase in parking cost shows similar car usage discourage effect Because fuel cost is

related to the distance between home and work locations and parking cost is related to work time

the time that an individual spends out of the home is negatively affected by hisher likelihood to use

a car

14

Table 4 - Definition of the significant variables

AbbreviationVariableTransportation demand management measures

Measures

ParkingParking cost increase Rials per hour

CordonCordon price Rials per entranceAccessTransit access time shortage percent

Interaction of push measures

ParkampFuelParking cost and fuel cost simultaneous effectsCordonampFuelCordon pricing and fuel cost simultaneous effects

Interaction of pull measures

PT_timeampaccessPT time reduction and access improvement simultaneous effectsCommuting trip characteristics

Trip distanceDistance between home and workplaceTrip timeTravel time between home and workplace

Exp FuelLikelihood of unsubsidized fuel use (self-reported on a Likert scale)NtripsNumber of daily tripsPattern2Commuting with 1+ stop(s) in go or return

Pattern3Commuting with 2 workplacesFirst trip timeStart time of first trip

PnocarwkLikelihood of going to work in absence of that car (self-reported)PTnwaccNon-walk access to transit (yes=1)First NaccoNumber of passengers in first trip

PassengerAny passenger on that day (yes=1)Park_paymentParking payment in last weekNhempfullNumber of full employees in HH

CardependencyBoardalight a passenger or move freight in the trip (yes=1)D car ownBe the owner of the used vehicle (yes=1)

Car accCar accessibility in household (number of cars to number of HH driving licenses ratio)NmotorcycleNumber of motorcycles owned by HHD home placeHome Location is in study area (yes=1)

PermissionPermission to enter to study area (yes=1)ComfortI use my car because it is comfortablePoor_PTI use my car because transit is not good

HH socio-economic characteristics

FemaleGender (Female=1)Age lt30Age younger than 30 (yes=1)Age 30_39Age between 30 to 39 (yes=1)Job_durationNumber of years that individual has been at hisher job

Emp_fullFull-time employee (yes=1)Edu BSDegree of education is BSc (yes=1)Edu BS+Degree of education is higher than BSc(yes=1)

D childlt=18Child younger than 18 in HH (yes=1)

15

Table 5 ndash The mode choice model

Tel-Taxi(T_T)

Motorcycle(MC)

Drive amp Ride(DampR)

Taxi (T)Walk amp Ride(WampR)

Car (C)Mode

Variable-471756-37067-147911Constant

Transportation demand management measure variables00019-00045Cordon

-000072Parking-004308Access

-28443D-05Parkampfuel-32475D-06Cordonampfuel

00029Pt_timeampaccess

Commuting trip characteristics-04709Trip distance

-02163-00831Trip time-96755163655Exp fuel-16253Ntrips

-114779Pattern2-71008Pattern3

00282-00270First trip time-02439-01549Pnocarwk

-11322992883-32765PTnwacc-133701First Nacco

-7778-73782Accompany-00049000010Park_payment

201646195554Nhempfull-160144ComfortCar1

-206142DependencyCar1-16101883385-121224DependencyCar2

42176Poor_PTCar1-24988Poor_PTCar2

- -27221D car own70960-39136Car acc

1 -71112-156123Nmotorcycle-1436322762D home place

2 78826Permission

HH socio-economic characteristics149490Female

297584-24548Agelt30-136490Age30_39

079430366303585Job_duration-108743Emp_full-203468-64900Edu BS

10932856687-4499984445Edu BS+102271D childlt=18

-2677366L( )-3849556L(0)0305sup2

112127178592580607N

Note = Positive significance at 1 5 10 level

As expected individuals with higher income are more likely to use their car This is indicated in the

model by the positive signs of individuals who use fuel with fixed (unsubsidized) cost and individuals

16

who pay more in parking charges in the previous week of study Negative sign of Pnocarwk variable

shows that the commuters who stated that their commute depends on car availability are more

likely to use their car Individuals in households with more full-time employees are more likely to use

their car which may be the result of higher household income Not surprisingly commuters who

have permission are more likely to maintain car usage Among the household socio-economic

parameters greater job experience (Job_duration) and higher graduate levels (EduBS+) increase the

probability of car usage

Public transit accessed by walking (WampR)

Access time to transit negatively impacts WampR choice which is expected This result is similar to

findings for the city of Sydney (Hensher amp Rose 2007) The negative coefficient of first trip time

indicates that individuals are more likely to use WampR in the early morning This result seems to

reflect the better weather for walking and faster speed of WampR mode early in the morning

Obviously individuals who are not able to access transit stations via walking (PTnwacc) are less likely

to consider this mode Furthermore serving passengers on daily trips is also a deterrent to using

WampR

Initially assessing the individuals who stated that their car usage is due to poor public transit service

(Poor_PT) led to an unexpected result in favor of considering WampR By introducing to this variable

the number of household cars as a proxy for household income (Poor_PTCar1) the model shows

that of the previously mentioned individuals those who have lower income are the ones who have

to consider WampR The result is understandable as these individuals may have no alternative when

they have to change their mode (they also are not likely to consider other modes) Individuals with

higher levels of income who have to use their car during before or after work (Dependencycar1+)

are not likely to use WampR

The greater the number of motorcycles in a household the less likely commuters is to consider

WampR There appears to be a competition between motorcycle and PT for access to the city center

17

Better PT services in the center of the city in terms of coverage and frequency increases the

likelihood that its residents will consider WampR This is verified by the positive sign of the

D_home_place variable Commuters with greater job experience (Job_duration) in their workplace

are more likely to use this mode Although individuals with higher levels of education are not likely

to use WampR as education level increases avoidance of WampR decreases

Taxi (T)

Table 5 shows that none of the studied policies are significant in considering taxi usage It seems that

taxi usage considering its function in Iran as a non-private and non-public mode of transport is not

affected by pull or push policies A negative sign for taxi travel time indicates that individuals are not

likely to use this mode for longer trips This seems reasonable given that longer trips are more

expensive Commuters who are more likely to use fuel with no subsidy are not likely to use taxis As

mentioned before they prefer to use their car A higher number of trips in a day are also a deterrent

to considering taxi usage which may be due to increased cost for more trips Results show that an

individual with more daily trips avoids using taxis Commuters who are employed in more than one

workplace (Pattern 3) are not likely to use taxis This may be due to the fact that they have a lower

level of income which forces them to dedicate more time on the job

Initial results showed that individuals who stated that their car usage is due to poor public transit

service (Poor_PT) are not likely to use taxis This result was far from our expectations By introducing

to this variable the number of household cars as a proxy for household income the model shows

that the previously mentioned individuals who have higher income (Poor_PTCar1+) are the ones

who are not likely to consider taxis Furthermore because such individuals are not considering any

other modes they may treat taxi usage as a kind of PT mode with poor service

As expected greater access to cars in a household (Car_acc) lessens the likelihood of considering

taxis as an alternative Furthermore individuals in households with more motorcycle ownership are

less likely to consider taxis It seems like there is a competition among motorcycles and taxis for

18

access to the city center Younger commuters are less likely to use taxis and individuals with at least

master degrees do consider this mode in addition to their car

Public transit accessed by Drive (DampR)

This mode is affected by the simultaneous interaction of transit time and transit access

(PT_TimeampAccess) which is reflected in the fact that individuals prefer to use this mode for longer

trips Comparing this mode and WampR the first trip start time affects the consideration of this mode

differently Later morning commuters prefer to use their car to access PT modes Such commuters

may have higher income levels or managerial jobs Obviously individuals who are not able to access

PT stations by walking (PTnwacc) are likely to use DampR Serving passengers in daily trips is also a

deterrent in considering this mode which is similar to WampR but with a lower coefficient

Commuters with higher income levels who depend on their car during before or after work

(Dependencycar1+) are likely to use DampR Individuals who use their own car are less likely to use

this mode which is unexpected As a city center develops better PT network coverage and residents

have smaller distances to their workplaces they are unlikely to use DampR This is proven in the model

by a negative sign for D_home_place

Motorcycle (MC)

Increasing fuel cost and cordon pricing simultaneously discourage motorcycle usages Although fuel

cost is expected to reduce motorcycle usage to some extent its combined effect with cordon pricing

also reduces motorcycle usage However this variable is not as strong as other policy variables

=10)

Of the studied modes motorcycle usage is affected by the most commuting variables This may be

due to the fact that this mode is not common Commuting distance has a negative effect on

motorcycle usage which is expected It is worth noting that trip distance appears only in this mode

which may be a reflection of the role of distance in regards to the safety risk in considering this

19

mode Commuters with more stops to serve passengers while commuting (Pattern 2) are not likely

to use this mode which may be due to the poor passenger service of this mode

Individuals who state that commuting is independent of the mode (Pnocarwk) are not likely to use

MC By looking at the (First_Nacco) negative sign this could stem from the fact that the more

passengers there are on the first trip the less likely individuals are to consider MC Regarding the

low capacity of MC and its safety concerns such commuters avoid using this mode Commuters who

pay more parking charges (Park_payment) are less likely to use MC which is expected Individuals

who are dependent on their car during before or after their work time are not likely to use MC

even if they have lower levels of income (DependencyCar1) Individuals who use their own car

(D_car_own) are less likely to use this mode As expected individuals who live in households with

more motorcycle ownership are more likely to use this mode The positive sign of (Permission)

indicates that commuters who have permission to enter the study area do consider MC Because

such commuters generally provide that permission just for car usage this result is unexpected

As with commute variables of all the studied modes MC is affected by the greatest number of

socio-economic variables As expected young commuters (Agelt30) are more likely to use this mode

Commuters with Bachelor of Science degree are less likely to use this mode among others Full time

employees (Emp_full) are less likely to consider MC whereas commuters with more experience in

their jobs prefer to use it Results show that individuals who live in a household with children

younger than 18 are more likely to consider using a car

Tel-Taxi (T_T)

Results show that cordon pricing causes higher probability of using T_T In fact individuals who use

T_T as a mode with similar level of service as cars9 are more willing to pay the cost and make use of

the mode It is worth noting that the effect of cordon pricing in pushing commuters from car usage

9 As this mode does not have driving stress and parking search time in some cases it may have more amount of utility thana car does

20

(000045) is greater than its effect on pulling them to Tel-taxi (000019) This is because of the

possibility of considering other non-car modes

Because consideration of this mode is a function of its operation travel time (Trip_time) appears as

a deterrent in this mode utility function Table 5 shows that individuals are more sensitive to the trip

time when using T_T mode versus taxi which is expected due to their relative costs

The greater the number of full time employees in a family (Nhempfull) the higher the probability of

considering T_T by its commuters which may be due to the higher income level of these

households This is verified by the greater likelihood of using T_T rather than taxis by such

commuters Individuals with higher levels of income who depend on their car during before or after

work time are less likely to use T_T Commuters with lower income levels who state that they use

their car for the sake of comfort (Comfortcar1) are less likely to use T_T which may be due to its

cost Although such individuals do not consider any other modes they specifically avoid T_T Greater

access to cars in a household leads to greater likelihood of T_T usage which could be due to the

higher income level of a household As mentioned before such individuals even avoid taxis

Females who drive to their workplace are more likely to use T_T It seems like this part of society

considers this mode when desiring to avoid the difficulties of driving Younger commuters are less

likely to use T_T and individuals between 30 and 39 years of age are specifically avoiding this mode

Results show that university graduated commuters are more likely to use this mode

6 Marginal effects

To explore the effects of each policy on mode choice and to answer the second issue raised at the

beginning of this paper the marginal effects approach can be adopted Although the coefficients of

the models utility functions show the drivers behavior when facing one or more policies the

marginal effects of policies or their interactions may appropriately show the results of their

implementation More specifically the marginal effect for this study is interpreted as the change in

21

probability given a unit change in a variable ceteris paribus In this section the variable is defined as

a specific policy or a specific interaction of two policies Table 6 presents the marginal effects of the

studied policies and their interactions with mode choice The results are shown in the form of trip

percentages transferred away from the car to the studied modes and the probability-weighted

sample enumeration approach is adopted to find the values It is worth noting that this table is fully

compatible with Table 5 but the marginal effects that were less significant than 90 percent have

been removed

Table 6 - Marginal effects of policies (percent)

Tel-Taxi(TT)

Motorcycle(MC)

Drive ampRide(DampR)

Taxi (T)Walk ampRide(WampR)

Car (C)Mode

Variable-000088Cordon-000140Parking

-09069Access-0000001ParkampFuel

00040PT_TimeampAccess

Table 5 shows that a 1 Rial increase in cordon pricing causes a 000088 percent decrease in car

usage Furthermore a 1 Rial per hour increase in parking cost decreases the car usage probability to

00014 percent By assuming 8 hours for the average parking duration the daily marginal value of

parking cost converts to 000018 percent These values show that cordon pricing is more effective in

forcing individuals not to use their car than increasing parking cost with the same value Results also

show that a 1 Rial increase in parking cost and fuel cost simultaneously decreases the probability of

choosing car usage by 0000001 percent Table 5 also shows that a 1 minute decrease in transit

access time would result in a 09 percent increase in probability of choosing this mode It also shows

that each 1 minute decrease in transit travel time and transit access leads to a 0004 decrease in the

probability of choosing the DampR mode

22

The above results show that just one of the policy interactions is not significantly different from

none (ie CordonampFuel) and push policies have a significant role in impelling car users to consider

other modes

7 Planning

To demonstrate the application of the model some of its planning aspects for the city of Tehran are

presented here In fact the model is able to assess the effect of all studied variables which were

grouped into three tiers on commuters mode choice In this section we assess the effect of TDM

policies by adopting their pair-wise combinations for the city Table 6 shows that pull policies in this

study did not impose a major effect on car usage Because the average access time of the sample

was 11 minutes improving it by 50 percent (about 5 minutes) increases the probability of WampR to

509069100 which is equal to a 048 percent increase in considering this mode (and decrease in all

other modes) On the other hand decreasing transit time about 50 percent (about 19 minutes based

on sample average) and similarly improving access time decreases the probability of considering

DampR to 0004195100 which is equal to a 038 percent decrease in considering this mode (and

increase in all other modes) Therefore in this section we focused on the push policies which

appear in the utility function of car mode It is worth noting that in assessing each combination of

policies all other variables were assumed to be fixed as the current state

71 Parking pricing and cordon pricing

The average hourly parking charge of individuals in the sample is 71 Rials This charge is 3000 Rials

per hour in the curb lane on some streets about 2000 Rials per hour in parking lots and free in

alleys and other streets The average amount shows that most of the commuters benefited from

free parking spaces in alleys and non-controlled streets Figure 2 depicts the result of implementing

this policy and the cordon pricing policy simultaneously In line with marginal effects it can be seen

that in implementing each policy separately cordon pricing is more effective than parking pricing in

the planning range It is worth noting that commuters in this study are less sensitive to parking

23

pricing than cordon pricing One reason may be that they thought in any case they would find free

parking space Figure 2 shows that the effect of each policy depends on the level of the other policy

In other words the effect of the simultaneous implementation of two policies is not equal to the

sum of their separate individual effects

Figure 2- Effect of implementing parking pricing and cordon pricing

72 Parking pricing and increasing fuel cost

Figure 3 shows the simultaneous effects of parking pricing and increasing fuel cost policies Although

both of these policies are not effective separately their simultaneous effect is significant It can be

seen that their synergy appears in the higher levels of both policies

Figure 3 - Effect of implementing parking pricing and increasing fuel cost

24

73 Cordon pricing and increasing fuel cost

The effect of implementing cordon pricing and increasing fuel cost is shown in Figure 4 It can be

seen that these two policies have no interaction effect in decreasing car usage This result is verified

by this interactions insignificant marginal value in Table 6

Figure 4- Effect of implementing cordon pricing and increasing fuel cost

8 Conclusions

This study examined the role of TDM policies in individual mode choice for work trips in the city of

Tehran Five policies including increasing parking cost increasing fuel cost cordon pricing in the odd-

even zone of the city transit time reduction and transit access improvement were investigated Of

the five policies the former three were push policies and the latter two were pull policies The

design of experiments approach was used to design the questionnaire that would capture the stated

preferences of car commuters

A number of nested structures based on recognizing differences in the variances associated

with unobserved influences were examined and finally the MNL as a superior model for this study

was developed

25

The results show that for car users who faced TDM policies all of the five studied TDM

policies were factors in the mode choice process The model shows that in considering car usage

push policies play a main role and pull policies do not

Results also show that the interactions of policies are also significant The model shows that

the interaction of parking cost policy and fuel cost policy is significant in prompting car users to

consider other modes It also shows that the interaction of cordon pricing and increase in fuel cost is

effective to discourage motorcycle usage Furthermore the interaction of transit time reduction

policy and access improvements policy is significant in considering DampR as a transportation mode to

workplace

Pull policies in the study were expected to attract individuals to transit modes (pull policies

of this study are related to transit modes) This was verified by the model

Results show that although taxi usage is usually treated as an alternative for urban trips it

was not directly affected by the studied policies This may be due to its special function in Tehran as

a non-public and non-private mode

The results also show that except for the interaction of cordon pricing and increase in fuel

cost other interaction effects have significant marginal effects on mode choice

Results of the model show that to implement a single policy cordon pricing is the most

effective in decreasing car usage

This study assessed workplace commuters Thus future studies should explore the

individuals mode choice with other trip aims Additionally including more policies especially pull

policies in favor of non-transit modes is reasonable to rigorously assess the model

It can be seen that the studied modes are affected differently by the TDM policies and their

interactions The impact of the interaction of policies appeared in the consideration of three out of

six studied modes and also in two out of three developed graphs This issue needs more attention

within the transportation demand management context and further studies should be performed

26

Design of this study is based on the consideration of all two-way interactions with some

degrees of deficiency in D-efficient policy Focusing on the significant interactions of this study is

recommended to perform other studies with more D-efficient values

Further understanding of the impacts of TDM policies might be provided in future research

by focusing on the amount of their synergy in prohibiting car usage

9 References

Arentze TA Hofman F amp Timmermans HJP (2004) Predicting multi-faceted activityndashtravel adjustmentstrategies in response to possible congestion pricing scenarios using an Internet-based stated adaptationexperiment Transport Policy 11 p31ndash41

Burris M amp Patil S (2009) Estimating the Benefits of Managed Lanes Technical Report Texas TexasTransportation Institute University Transportation Center for Mobility

Cao X amp Mokhtarian PL (2005) How do individuals adapt their personal travel A conceptual exploration ofthe consideation of travel-related strategies Transport Policy 12(3) pp199-206

de Palma A amp Lindsey R (2001) Transportation Supply and congestion In International Encyclopedia of theSocial and Behavioral Sciences 1st ed Elsevier p15882ndash15888

Eriksson L Garvill J amp Nordlund AM (2008) Acceptability of single and combined transport policy measuresThe importance of environmental and policy specific beliefs Transportation Research Part A 42 p1117ndash1128

Eriksson L Nordlund AM amp Garvill J (2010) Expected car use reduction in response to structural traveldemand management measures Transportation Research Part F 13 p329ndash342

Golob TF amp Hensher DA (2007) The Trip chaining Activity of Sydney Residents A Cross-Section Assessmentby Age Group with a Focus on Seniors Journal of Transport Geography 15 p298ndash312

Hensher DA amp King J (2001) Parking demand and responsiveness to supply pricing and location in theSydney central business district Transportation Research A 35 pp177-96

Hensher DA amp Rose JM (2007) Development of commuter and non-commuter mode choice models for theassessment of new public transport infrastructure projects A case study Transportation Research Part A41 p428ndash443

Hensher DA Rose JM amp Greene WH (2005) Applied Choice Analysis A Primer New York CambridgeUniversity Press

Iranian Center of Statistics (ICS) (2009) Information of Iranian States [Online] Available at HYPERLINK httpwwwamarorgirUploadModulesContentsasset16tehrantehparthtmlhttpwwwamarorgirUploadModulesContentsasset16tehrantehparthtml [Accessed 13 November2009]

Kingham S Dickinson J amp Copsey S (2001) Travelling to work Will people move out of their cars TransportPolicy 8 pp151-60

Kuhfeld WF (2009) Marketing Research Methods in SAS Experimental Design Choice Conjoint andGraphical Techniques SAS 92 Edition NC USA SAS Institute Inc Cary Available on the Webwwwsupportsascomtechsuptnotetnote_stathtmlmarket

Litman T (2003) The online TDM Encyclopedia mobility management information gateway Transport Policy10 pp245-49

Litman T (2010) Online TDM Encyclopedia [Online] Available at HYPERLINK httpwwwvtpiorghttpwwwvtpiorg [Accessed 21 Augest 2010]

11

drive a car (C) walk to the station and catch public transit (WampR) drive to a public station and catch

public transit (DampR) ride a motorcycle (MC) catch a taxi7 (T) and catch a taxi by phone (T_T) DampR is

somewhat different than the more familiar ldquoPark amp Riderdquo In fact in the fringes of the odd-even

zone there were no specialized parking lots dedicated to this purpose and commuters considered

Drive amp Ride because they were not allowed to pass the fringes

After each scenario if the respondent changed hisher mode the reason(s) for the change were

asked It could be a sole policy or a bundle of them Furthermore travel-related information was

sought These data were not part of the stated choice but they might have important influences on

individual choices These data consisted of car dependency (need to drive someone or move freight

in the trip) parking place type and average weekly parking costs car and motorcycle ownership and

number of household driving licenses

Depending on the individuals activity in that day three types of activity patterns were detected

Pattern 1 described individuals who had no stop in their commute Pattern 2 was for individuals who

had at least one stop on their way to or from work and pattern 3 was for the individuals who went

to another workplace in their daily activities

Finally for the sake of data generalization and the examination of household characteristics gender

age and household type employment status and education level were also asked

5 Mode choice model

In order to detect the policies that affect individual mode choice the logit modeling approach was

adopted In this model one can determine if the interaction of two policies affects the mode choice

In the calibration step 152 variables were defined and their effects on consideration of each mode

were examined

7 Taxis in Iran are somewhat different than taxis in other countries of the world In fact taxis in Iran are not hiring by oneperson or a group of people at a time Taxis allow passengers to board or alight along their path with respect to theircapacity In other word this mode is functioning similar to transit vehicles but the stops are not predefined

12

51 Model structure

Initially a multinomial logit (MNL) model is developed (Figure 1a) By selecting a number of tree

structures based on recognizing differences in the variances associated with unobserved influences

we find that the greatest similarity in variance profiles is associated with public transport modes as

opposed to non-public modes (Figure 1b) This structure has two nests one including Car (C) and

Motorcycle (MC) as private modes and the other including Walk and Ride (WampR) Drive and Ride

(DampR) Taxi (T) and Tel-taxi (T_T) as non-private modes The result of this nested logit (NL) model is

shown in Table 3

Although it is not a statistically significant improvement overall on the MNL model the statistically

significant inclusive value8 (IV) of 0889 for non-public modes relative to the fixed parameter value of

10 for public modes suggests that there is a structural advantage in selecting the NL specification

The normal test of a statistically significant difference between NL and MNL is an IV parameter

relative to 10 calculated using a Wald-test via equation 1

)1(Wald-test = (IVparameter ndash 1)std error

a The MNL structure

b Final nested structure

Figure 1- Model structure

8 Also called scale parameter

Alternatives

MCCar T_TT WampR DampR

Alternatives

MCCar T_TT

Public

WampR DampR

Private

13

We have (0889-1)2508 =-075 which would be rejected at the usual acceptable significance levels

This suggests that the NL model could be collapsed into an MNL form

Table 3- Nested logit (NL) model resultValueParameter

0889IV (Private)1000IV (nPrivate)

-2668335L( )-4057684L(0)

0342sup2

After the calibration process the variables that were statistically significant were identified and are

presented in Table 4 Table 5 presents the final model of the study with a goodness of fit of 031 for

the six studied modes For a general review of the model calibration results the effective factors can

be grouped under the following three categories TDM policy characteristics commuting trip

characteristics and household socio-economic characteristics which are all treated as alternative-

specific variables

52 Model results

Car (C)

It is expected that push policies impel car-drivers to choose other modes Table 5 shows that cordon

pricing and increase in parking cost cause individuals to choose not to use their car This is in line

with other studies suggesting that these policies are effective to discourage car usage (Hensher amp

Rose 2007 OFallon et al 2004) In addition the interaction between the policies of fuel cost

increase and increase in parking cost shows similar car usage discourage effect Because fuel cost is

related to the distance between home and work locations and parking cost is related to work time

the time that an individual spends out of the home is negatively affected by hisher likelihood to use

a car

14

Table 4 - Definition of the significant variables

AbbreviationVariableTransportation demand management measures

Measures

ParkingParking cost increase Rials per hour

CordonCordon price Rials per entranceAccessTransit access time shortage percent

Interaction of push measures

ParkampFuelParking cost and fuel cost simultaneous effectsCordonampFuelCordon pricing and fuel cost simultaneous effects

Interaction of pull measures

PT_timeampaccessPT time reduction and access improvement simultaneous effectsCommuting trip characteristics

Trip distanceDistance between home and workplaceTrip timeTravel time between home and workplace

Exp FuelLikelihood of unsubsidized fuel use (self-reported on a Likert scale)NtripsNumber of daily tripsPattern2Commuting with 1+ stop(s) in go or return

Pattern3Commuting with 2 workplacesFirst trip timeStart time of first trip

PnocarwkLikelihood of going to work in absence of that car (self-reported)PTnwaccNon-walk access to transit (yes=1)First NaccoNumber of passengers in first trip

PassengerAny passenger on that day (yes=1)Park_paymentParking payment in last weekNhempfullNumber of full employees in HH

CardependencyBoardalight a passenger or move freight in the trip (yes=1)D car ownBe the owner of the used vehicle (yes=1)

Car accCar accessibility in household (number of cars to number of HH driving licenses ratio)NmotorcycleNumber of motorcycles owned by HHD home placeHome Location is in study area (yes=1)

PermissionPermission to enter to study area (yes=1)ComfortI use my car because it is comfortablePoor_PTI use my car because transit is not good

HH socio-economic characteristics

FemaleGender (Female=1)Age lt30Age younger than 30 (yes=1)Age 30_39Age between 30 to 39 (yes=1)Job_durationNumber of years that individual has been at hisher job

Emp_fullFull-time employee (yes=1)Edu BSDegree of education is BSc (yes=1)Edu BS+Degree of education is higher than BSc(yes=1)

D childlt=18Child younger than 18 in HH (yes=1)

15

Table 5 ndash The mode choice model

Tel-Taxi(T_T)

Motorcycle(MC)

Drive amp Ride(DampR)

Taxi (T)Walk amp Ride(WampR)

Car (C)Mode

Variable-471756-37067-147911Constant

Transportation demand management measure variables00019-00045Cordon

-000072Parking-004308Access

-28443D-05Parkampfuel-32475D-06Cordonampfuel

00029Pt_timeampaccess

Commuting trip characteristics-04709Trip distance

-02163-00831Trip time-96755163655Exp fuel-16253Ntrips

-114779Pattern2-71008Pattern3

00282-00270First trip time-02439-01549Pnocarwk

-11322992883-32765PTnwacc-133701First Nacco

-7778-73782Accompany-00049000010Park_payment

201646195554Nhempfull-160144ComfortCar1

-206142DependencyCar1-16101883385-121224DependencyCar2

42176Poor_PTCar1-24988Poor_PTCar2

- -27221D car own70960-39136Car acc

1 -71112-156123Nmotorcycle-1436322762D home place

2 78826Permission

HH socio-economic characteristics149490Female

297584-24548Agelt30-136490Age30_39

079430366303585Job_duration-108743Emp_full-203468-64900Edu BS

10932856687-4499984445Edu BS+102271D childlt=18

-2677366L( )-3849556L(0)0305sup2

112127178592580607N

Note = Positive significance at 1 5 10 level

As expected individuals with higher income are more likely to use their car This is indicated in the

model by the positive signs of individuals who use fuel with fixed (unsubsidized) cost and individuals

16

who pay more in parking charges in the previous week of study Negative sign of Pnocarwk variable

shows that the commuters who stated that their commute depends on car availability are more

likely to use their car Individuals in households with more full-time employees are more likely to use

their car which may be the result of higher household income Not surprisingly commuters who

have permission are more likely to maintain car usage Among the household socio-economic

parameters greater job experience (Job_duration) and higher graduate levels (EduBS+) increase the

probability of car usage

Public transit accessed by walking (WampR)

Access time to transit negatively impacts WampR choice which is expected This result is similar to

findings for the city of Sydney (Hensher amp Rose 2007) The negative coefficient of first trip time

indicates that individuals are more likely to use WampR in the early morning This result seems to

reflect the better weather for walking and faster speed of WampR mode early in the morning

Obviously individuals who are not able to access transit stations via walking (PTnwacc) are less likely

to consider this mode Furthermore serving passengers on daily trips is also a deterrent to using

WampR

Initially assessing the individuals who stated that their car usage is due to poor public transit service

(Poor_PT) led to an unexpected result in favor of considering WampR By introducing to this variable

the number of household cars as a proxy for household income (Poor_PTCar1) the model shows

that of the previously mentioned individuals those who have lower income are the ones who have

to consider WampR The result is understandable as these individuals may have no alternative when

they have to change their mode (they also are not likely to consider other modes) Individuals with

higher levels of income who have to use their car during before or after work (Dependencycar1+)

are not likely to use WampR

The greater the number of motorcycles in a household the less likely commuters is to consider

WampR There appears to be a competition between motorcycle and PT for access to the city center

17

Better PT services in the center of the city in terms of coverage and frequency increases the

likelihood that its residents will consider WampR This is verified by the positive sign of the

D_home_place variable Commuters with greater job experience (Job_duration) in their workplace

are more likely to use this mode Although individuals with higher levels of education are not likely

to use WampR as education level increases avoidance of WampR decreases

Taxi (T)

Table 5 shows that none of the studied policies are significant in considering taxi usage It seems that

taxi usage considering its function in Iran as a non-private and non-public mode of transport is not

affected by pull or push policies A negative sign for taxi travel time indicates that individuals are not

likely to use this mode for longer trips This seems reasonable given that longer trips are more

expensive Commuters who are more likely to use fuel with no subsidy are not likely to use taxis As

mentioned before they prefer to use their car A higher number of trips in a day are also a deterrent

to considering taxi usage which may be due to increased cost for more trips Results show that an

individual with more daily trips avoids using taxis Commuters who are employed in more than one

workplace (Pattern 3) are not likely to use taxis This may be due to the fact that they have a lower

level of income which forces them to dedicate more time on the job

Initial results showed that individuals who stated that their car usage is due to poor public transit

service (Poor_PT) are not likely to use taxis This result was far from our expectations By introducing

to this variable the number of household cars as a proxy for household income the model shows

that the previously mentioned individuals who have higher income (Poor_PTCar1+) are the ones

who are not likely to consider taxis Furthermore because such individuals are not considering any

other modes they may treat taxi usage as a kind of PT mode with poor service

As expected greater access to cars in a household (Car_acc) lessens the likelihood of considering

taxis as an alternative Furthermore individuals in households with more motorcycle ownership are

less likely to consider taxis It seems like there is a competition among motorcycles and taxis for

18

access to the city center Younger commuters are less likely to use taxis and individuals with at least

master degrees do consider this mode in addition to their car

Public transit accessed by Drive (DampR)

This mode is affected by the simultaneous interaction of transit time and transit access

(PT_TimeampAccess) which is reflected in the fact that individuals prefer to use this mode for longer

trips Comparing this mode and WampR the first trip start time affects the consideration of this mode

differently Later morning commuters prefer to use their car to access PT modes Such commuters

may have higher income levels or managerial jobs Obviously individuals who are not able to access

PT stations by walking (PTnwacc) are likely to use DampR Serving passengers in daily trips is also a

deterrent in considering this mode which is similar to WampR but with a lower coefficient

Commuters with higher income levels who depend on their car during before or after work

(Dependencycar1+) are likely to use DampR Individuals who use their own car are less likely to use

this mode which is unexpected As a city center develops better PT network coverage and residents

have smaller distances to their workplaces they are unlikely to use DampR This is proven in the model

by a negative sign for D_home_place

Motorcycle (MC)

Increasing fuel cost and cordon pricing simultaneously discourage motorcycle usages Although fuel

cost is expected to reduce motorcycle usage to some extent its combined effect with cordon pricing

also reduces motorcycle usage However this variable is not as strong as other policy variables

=10)

Of the studied modes motorcycle usage is affected by the most commuting variables This may be

due to the fact that this mode is not common Commuting distance has a negative effect on

motorcycle usage which is expected It is worth noting that trip distance appears only in this mode

which may be a reflection of the role of distance in regards to the safety risk in considering this

19

mode Commuters with more stops to serve passengers while commuting (Pattern 2) are not likely

to use this mode which may be due to the poor passenger service of this mode

Individuals who state that commuting is independent of the mode (Pnocarwk) are not likely to use

MC By looking at the (First_Nacco) negative sign this could stem from the fact that the more

passengers there are on the first trip the less likely individuals are to consider MC Regarding the

low capacity of MC and its safety concerns such commuters avoid using this mode Commuters who

pay more parking charges (Park_payment) are less likely to use MC which is expected Individuals

who are dependent on their car during before or after their work time are not likely to use MC

even if they have lower levels of income (DependencyCar1) Individuals who use their own car

(D_car_own) are less likely to use this mode As expected individuals who live in households with

more motorcycle ownership are more likely to use this mode The positive sign of (Permission)

indicates that commuters who have permission to enter the study area do consider MC Because

such commuters generally provide that permission just for car usage this result is unexpected

As with commute variables of all the studied modes MC is affected by the greatest number of

socio-economic variables As expected young commuters (Agelt30) are more likely to use this mode

Commuters with Bachelor of Science degree are less likely to use this mode among others Full time

employees (Emp_full) are less likely to consider MC whereas commuters with more experience in

their jobs prefer to use it Results show that individuals who live in a household with children

younger than 18 are more likely to consider using a car

Tel-Taxi (T_T)

Results show that cordon pricing causes higher probability of using T_T In fact individuals who use

T_T as a mode with similar level of service as cars9 are more willing to pay the cost and make use of

the mode It is worth noting that the effect of cordon pricing in pushing commuters from car usage

9 As this mode does not have driving stress and parking search time in some cases it may have more amount of utility thana car does

20

(000045) is greater than its effect on pulling them to Tel-taxi (000019) This is because of the

possibility of considering other non-car modes

Because consideration of this mode is a function of its operation travel time (Trip_time) appears as

a deterrent in this mode utility function Table 5 shows that individuals are more sensitive to the trip

time when using T_T mode versus taxi which is expected due to their relative costs

The greater the number of full time employees in a family (Nhempfull) the higher the probability of

considering T_T by its commuters which may be due to the higher income level of these

households This is verified by the greater likelihood of using T_T rather than taxis by such

commuters Individuals with higher levels of income who depend on their car during before or after

work time are less likely to use T_T Commuters with lower income levels who state that they use

their car for the sake of comfort (Comfortcar1) are less likely to use T_T which may be due to its

cost Although such individuals do not consider any other modes they specifically avoid T_T Greater

access to cars in a household leads to greater likelihood of T_T usage which could be due to the

higher income level of a household As mentioned before such individuals even avoid taxis

Females who drive to their workplace are more likely to use T_T It seems like this part of society

considers this mode when desiring to avoid the difficulties of driving Younger commuters are less

likely to use T_T and individuals between 30 and 39 years of age are specifically avoiding this mode

Results show that university graduated commuters are more likely to use this mode

6 Marginal effects

To explore the effects of each policy on mode choice and to answer the second issue raised at the

beginning of this paper the marginal effects approach can be adopted Although the coefficients of

the models utility functions show the drivers behavior when facing one or more policies the

marginal effects of policies or their interactions may appropriately show the results of their

implementation More specifically the marginal effect for this study is interpreted as the change in

21

probability given a unit change in a variable ceteris paribus In this section the variable is defined as

a specific policy or a specific interaction of two policies Table 6 presents the marginal effects of the

studied policies and their interactions with mode choice The results are shown in the form of trip

percentages transferred away from the car to the studied modes and the probability-weighted

sample enumeration approach is adopted to find the values It is worth noting that this table is fully

compatible with Table 5 but the marginal effects that were less significant than 90 percent have

been removed

Table 6 - Marginal effects of policies (percent)

Tel-Taxi(TT)

Motorcycle(MC)

Drive ampRide(DampR)

Taxi (T)Walk ampRide(WampR)

Car (C)Mode

Variable-000088Cordon-000140Parking

-09069Access-0000001ParkampFuel

00040PT_TimeampAccess

Table 5 shows that a 1 Rial increase in cordon pricing causes a 000088 percent decrease in car

usage Furthermore a 1 Rial per hour increase in parking cost decreases the car usage probability to

00014 percent By assuming 8 hours for the average parking duration the daily marginal value of

parking cost converts to 000018 percent These values show that cordon pricing is more effective in

forcing individuals not to use their car than increasing parking cost with the same value Results also

show that a 1 Rial increase in parking cost and fuel cost simultaneously decreases the probability of

choosing car usage by 0000001 percent Table 5 also shows that a 1 minute decrease in transit

access time would result in a 09 percent increase in probability of choosing this mode It also shows

that each 1 minute decrease in transit travel time and transit access leads to a 0004 decrease in the

probability of choosing the DampR mode

22

The above results show that just one of the policy interactions is not significantly different from

none (ie CordonampFuel) and push policies have a significant role in impelling car users to consider

other modes

7 Planning

To demonstrate the application of the model some of its planning aspects for the city of Tehran are

presented here In fact the model is able to assess the effect of all studied variables which were

grouped into three tiers on commuters mode choice In this section we assess the effect of TDM

policies by adopting their pair-wise combinations for the city Table 6 shows that pull policies in this

study did not impose a major effect on car usage Because the average access time of the sample

was 11 minutes improving it by 50 percent (about 5 minutes) increases the probability of WampR to

509069100 which is equal to a 048 percent increase in considering this mode (and decrease in all

other modes) On the other hand decreasing transit time about 50 percent (about 19 minutes based

on sample average) and similarly improving access time decreases the probability of considering

DampR to 0004195100 which is equal to a 038 percent decrease in considering this mode (and

increase in all other modes) Therefore in this section we focused on the push policies which

appear in the utility function of car mode It is worth noting that in assessing each combination of

policies all other variables were assumed to be fixed as the current state

71 Parking pricing and cordon pricing

The average hourly parking charge of individuals in the sample is 71 Rials This charge is 3000 Rials

per hour in the curb lane on some streets about 2000 Rials per hour in parking lots and free in

alleys and other streets The average amount shows that most of the commuters benefited from

free parking spaces in alleys and non-controlled streets Figure 2 depicts the result of implementing

this policy and the cordon pricing policy simultaneously In line with marginal effects it can be seen

that in implementing each policy separately cordon pricing is more effective than parking pricing in

the planning range It is worth noting that commuters in this study are less sensitive to parking

23

pricing than cordon pricing One reason may be that they thought in any case they would find free

parking space Figure 2 shows that the effect of each policy depends on the level of the other policy

In other words the effect of the simultaneous implementation of two policies is not equal to the

sum of their separate individual effects

Figure 2- Effect of implementing parking pricing and cordon pricing

72 Parking pricing and increasing fuel cost

Figure 3 shows the simultaneous effects of parking pricing and increasing fuel cost policies Although

both of these policies are not effective separately their simultaneous effect is significant It can be

seen that their synergy appears in the higher levels of both policies

Figure 3 - Effect of implementing parking pricing and increasing fuel cost

24

73 Cordon pricing and increasing fuel cost

The effect of implementing cordon pricing and increasing fuel cost is shown in Figure 4 It can be

seen that these two policies have no interaction effect in decreasing car usage This result is verified

by this interactions insignificant marginal value in Table 6

Figure 4- Effect of implementing cordon pricing and increasing fuel cost

8 Conclusions

This study examined the role of TDM policies in individual mode choice for work trips in the city of

Tehran Five policies including increasing parking cost increasing fuel cost cordon pricing in the odd-

even zone of the city transit time reduction and transit access improvement were investigated Of

the five policies the former three were push policies and the latter two were pull policies The

design of experiments approach was used to design the questionnaire that would capture the stated

preferences of car commuters

A number of nested structures based on recognizing differences in the variances associated

with unobserved influences were examined and finally the MNL as a superior model for this study

was developed

25

The results show that for car users who faced TDM policies all of the five studied TDM

policies were factors in the mode choice process The model shows that in considering car usage

push policies play a main role and pull policies do not

Results also show that the interactions of policies are also significant The model shows that

the interaction of parking cost policy and fuel cost policy is significant in prompting car users to

consider other modes It also shows that the interaction of cordon pricing and increase in fuel cost is

effective to discourage motorcycle usage Furthermore the interaction of transit time reduction

policy and access improvements policy is significant in considering DampR as a transportation mode to

workplace

Pull policies in the study were expected to attract individuals to transit modes (pull policies

of this study are related to transit modes) This was verified by the model

Results show that although taxi usage is usually treated as an alternative for urban trips it

was not directly affected by the studied policies This may be due to its special function in Tehran as

a non-public and non-private mode

The results also show that except for the interaction of cordon pricing and increase in fuel

cost other interaction effects have significant marginal effects on mode choice

Results of the model show that to implement a single policy cordon pricing is the most

effective in decreasing car usage

This study assessed workplace commuters Thus future studies should explore the

individuals mode choice with other trip aims Additionally including more policies especially pull

policies in favor of non-transit modes is reasonable to rigorously assess the model

It can be seen that the studied modes are affected differently by the TDM policies and their

interactions The impact of the interaction of policies appeared in the consideration of three out of

six studied modes and also in two out of three developed graphs This issue needs more attention

within the transportation demand management context and further studies should be performed

26

Design of this study is based on the consideration of all two-way interactions with some

degrees of deficiency in D-efficient policy Focusing on the significant interactions of this study is

recommended to perform other studies with more D-efficient values

Further understanding of the impacts of TDM policies might be provided in future research

by focusing on the amount of their synergy in prohibiting car usage

9 References

Arentze TA Hofman F amp Timmermans HJP (2004) Predicting multi-faceted activityndashtravel adjustmentstrategies in response to possible congestion pricing scenarios using an Internet-based stated adaptationexperiment Transport Policy 11 p31ndash41

Burris M amp Patil S (2009) Estimating the Benefits of Managed Lanes Technical Report Texas TexasTransportation Institute University Transportation Center for Mobility

Cao X amp Mokhtarian PL (2005) How do individuals adapt their personal travel A conceptual exploration ofthe consideation of travel-related strategies Transport Policy 12(3) pp199-206

de Palma A amp Lindsey R (2001) Transportation Supply and congestion In International Encyclopedia of theSocial and Behavioral Sciences 1st ed Elsevier p15882ndash15888

Eriksson L Garvill J amp Nordlund AM (2008) Acceptability of single and combined transport policy measuresThe importance of environmental and policy specific beliefs Transportation Research Part A 42 p1117ndash1128

Eriksson L Nordlund AM amp Garvill J (2010) Expected car use reduction in response to structural traveldemand management measures Transportation Research Part F 13 p329ndash342

Golob TF amp Hensher DA (2007) The Trip chaining Activity of Sydney Residents A Cross-Section Assessmentby Age Group with a Focus on Seniors Journal of Transport Geography 15 p298ndash312

Hensher DA amp King J (2001) Parking demand and responsiveness to supply pricing and location in theSydney central business district Transportation Research A 35 pp177-96

Hensher DA amp Rose JM (2007) Development of commuter and non-commuter mode choice models for theassessment of new public transport infrastructure projects A case study Transportation Research Part A41 p428ndash443

Hensher DA Rose JM amp Greene WH (2005) Applied Choice Analysis A Primer New York CambridgeUniversity Press

Iranian Center of Statistics (ICS) (2009) Information of Iranian States [Online] Available at HYPERLINK httpwwwamarorgirUploadModulesContentsasset16tehrantehparthtmlhttpwwwamarorgirUploadModulesContentsasset16tehrantehparthtml [Accessed 13 November2009]

Kingham S Dickinson J amp Copsey S (2001) Travelling to work Will people move out of their cars TransportPolicy 8 pp151-60

Kuhfeld WF (2009) Marketing Research Methods in SAS Experimental Design Choice Conjoint andGraphical Techniques SAS 92 Edition NC USA SAS Institute Inc Cary Available on the Webwwwsupportsascomtechsuptnotetnote_stathtmlmarket

Litman T (2003) The online TDM Encyclopedia mobility management information gateway Transport Policy10 pp245-49

Litman T (2010) Online TDM Encyclopedia [Online] Available at HYPERLINK httpwwwvtpiorghttpwwwvtpiorg [Accessed 21 Augest 2010]

12

51 Model structure

Initially a multinomial logit (MNL) model is developed (Figure 1a) By selecting a number of tree

structures based on recognizing differences in the variances associated with unobserved influences

we find that the greatest similarity in variance profiles is associated with public transport modes as

opposed to non-public modes (Figure 1b) This structure has two nests one including Car (C) and

Motorcycle (MC) as private modes and the other including Walk and Ride (WampR) Drive and Ride

(DampR) Taxi (T) and Tel-taxi (T_T) as non-private modes The result of this nested logit (NL) model is

shown in Table 3

Although it is not a statistically significant improvement overall on the MNL model the statistically

significant inclusive value8 (IV) of 0889 for non-public modes relative to the fixed parameter value of

10 for public modes suggests that there is a structural advantage in selecting the NL specification

The normal test of a statistically significant difference between NL and MNL is an IV parameter

relative to 10 calculated using a Wald-test via equation 1

)1(Wald-test = (IVparameter ndash 1)std error

a The MNL structure

b Final nested structure

Figure 1- Model structure

8 Also called scale parameter

Alternatives

MCCar T_TT WampR DampR

Alternatives

MCCar T_TT

Public

WampR DampR

Private

13

We have (0889-1)2508 =-075 which would be rejected at the usual acceptable significance levels

This suggests that the NL model could be collapsed into an MNL form

Table 3- Nested logit (NL) model resultValueParameter

0889IV (Private)1000IV (nPrivate)

-2668335L( )-4057684L(0)

0342sup2

After the calibration process the variables that were statistically significant were identified and are

presented in Table 4 Table 5 presents the final model of the study with a goodness of fit of 031 for

the six studied modes For a general review of the model calibration results the effective factors can

be grouped under the following three categories TDM policy characteristics commuting trip

characteristics and household socio-economic characteristics which are all treated as alternative-

specific variables

52 Model results

Car (C)

It is expected that push policies impel car-drivers to choose other modes Table 5 shows that cordon

pricing and increase in parking cost cause individuals to choose not to use their car This is in line

with other studies suggesting that these policies are effective to discourage car usage (Hensher amp

Rose 2007 OFallon et al 2004) In addition the interaction between the policies of fuel cost

increase and increase in parking cost shows similar car usage discourage effect Because fuel cost is

related to the distance between home and work locations and parking cost is related to work time

the time that an individual spends out of the home is negatively affected by hisher likelihood to use

a car

14

Table 4 - Definition of the significant variables

AbbreviationVariableTransportation demand management measures

Measures

ParkingParking cost increase Rials per hour

CordonCordon price Rials per entranceAccessTransit access time shortage percent

Interaction of push measures

ParkampFuelParking cost and fuel cost simultaneous effectsCordonampFuelCordon pricing and fuel cost simultaneous effects

Interaction of pull measures

PT_timeampaccessPT time reduction and access improvement simultaneous effectsCommuting trip characteristics

Trip distanceDistance between home and workplaceTrip timeTravel time between home and workplace

Exp FuelLikelihood of unsubsidized fuel use (self-reported on a Likert scale)NtripsNumber of daily tripsPattern2Commuting with 1+ stop(s) in go or return

Pattern3Commuting with 2 workplacesFirst trip timeStart time of first trip

PnocarwkLikelihood of going to work in absence of that car (self-reported)PTnwaccNon-walk access to transit (yes=1)First NaccoNumber of passengers in first trip

PassengerAny passenger on that day (yes=1)Park_paymentParking payment in last weekNhempfullNumber of full employees in HH

CardependencyBoardalight a passenger or move freight in the trip (yes=1)D car ownBe the owner of the used vehicle (yes=1)

Car accCar accessibility in household (number of cars to number of HH driving licenses ratio)NmotorcycleNumber of motorcycles owned by HHD home placeHome Location is in study area (yes=1)

PermissionPermission to enter to study area (yes=1)ComfortI use my car because it is comfortablePoor_PTI use my car because transit is not good

HH socio-economic characteristics

FemaleGender (Female=1)Age lt30Age younger than 30 (yes=1)Age 30_39Age between 30 to 39 (yes=1)Job_durationNumber of years that individual has been at hisher job

Emp_fullFull-time employee (yes=1)Edu BSDegree of education is BSc (yes=1)Edu BS+Degree of education is higher than BSc(yes=1)

D childlt=18Child younger than 18 in HH (yes=1)

15

Table 5 ndash The mode choice model

Tel-Taxi(T_T)

Motorcycle(MC)

Drive amp Ride(DampR)

Taxi (T)Walk amp Ride(WampR)

Car (C)Mode

Variable-471756-37067-147911Constant

Transportation demand management measure variables00019-00045Cordon

-000072Parking-004308Access

-28443D-05Parkampfuel-32475D-06Cordonampfuel

00029Pt_timeampaccess

Commuting trip characteristics-04709Trip distance

-02163-00831Trip time-96755163655Exp fuel-16253Ntrips

-114779Pattern2-71008Pattern3

00282-00270First trip time-02439-01549Pnocarwk

-11322992883-32765PTnwacc-133701First Nacco

-7778-73782Accompany-00049000010Park_payment

201646195554Nhempfull-160144ComfortCar1

-206142DependencyCar1-16101883385-121224DependencyCar2

42176Poor_PTCar1-24988Poor_PTCar2

- -27221D car own70960-39136Car acc

1 -71112-156123Nmotorcycle-1436322762D home place

2 78826Permission

HH socio-economic characteristics149490Female

297584-24548Agelt30-136490Age30_39

079430366303585Job_duration-108743Emp_full-203468-64900Edu BS

10932856687-4499984445Edu BS+102271D childlt=18

-2677366L( )-3849556L(0)0305sup2

112127178592580607N

Note = Positive significance at 1 5 10 level

As expected individuals with higher income are more likely to use their car This is indicated in the

model by the positive signs of individuals who use fuel with fixed (unsubsidized) cost and individuals

16

who pay more in parking charges in the previous week of study Negative sign of Pnocarwk variable

shows that the commuters who stated that their commute depends on car availability are more

likely to use their car Individuals in households with more full-time employees are more likely to use

their car which may be the result of higher household income Not surprisingly commuters who

have permission are more likely to maintain car usage Among the household socio-economic

parameters greater job experience (Job_duration) and higher graduate levels (EduBS+) increase the

probability of car usage

Public transit accessed by walking (WampR)

Access time to transit negatively impacts WampR choice which is expected This result is similar to

findings for the city of Sydney (Hensher amp Rose 2007) The negative coefficient of first trip time

indicates that individuals are more likely to use WampR in the early morning This result seems to

reflect the better weather for walking and faster speed of WampR mode early in the morning

Obviously individuals who are not able to access transit stations via walking (PTnwacc) are less likely

to consider this mode Furthermore serving passengers on daily trips is also a deterrent to using

WampR

Initially assessing the individuals who stated that their car usage is due to poor public transit service

(Poor_PT) led to an unexpected result in favor of considering WampR By introducing to this variable

the number of household cars as a proxy for household income (Poor_PTCar1) the model shows

that of the previously mentioned individuals those who have lower income are the ones who have

to consider WampR The result is understandable as these individuals may have no alternative when

they have to change their mode (they also are not likely to consider other modes) Individuals with

higher levels of income who have to use their car during before or after work (Dependencycar1+)

are not likely to use WampR

The greater the number of motorcycles in a household the less likely commuters is to consider

WampR There appears to be a competition between motorcycle and PT for access to the city center

17

Better PT services in the center of the city in terms of coverage and frequency increases the

likelihood that its residents will consider WampR This is verified by the positive sign of the

D_home_place variable Commuters with greater job experience (Job_duration) in their workplace

are more likely to use this mode Although individuals with higher levels of education are not likely

to use WampR as education level increases avoidance of WampR decreases

Taxi (T)

Table 5 shows that none of the studied policies are significant in considering taxi usage It seems that

taxi usage considering its function in Iran as a non-private and non-public mode of transport is not

affected by pull or push policies A negative sign for taxi travel time indicates that individuals are not

likely to use this mode for longer trips This seems reasonable given that longer trips are more

expensive Commuters who are more likely to use fuel with no subsidy are not likely to use taxis As

mentioned before they prefer to use their car A higher number of trips in a day are also a deterrent

to considering taxi usage which may be due to increased cost for more trips Results show that an

individual with more daily trips avoids using taxis Commuters who are employed in more than one

workplace (Pattern 3) are not likely to use taxis This may be due to the fact that they have a lower

level of income which forces them to dedicate more time on the job

Initial results showed that individuals who stated that their car usage is due to poor public transit

service (Poor_PT) are not likely to use taxis This result was far from our expectations By introducing

to this variable the number of household cars as a proxy for household income the model shows

that the previously mentioned individuals who have higher income (Poor_PTCar1+) are the ones

who are not likely to consider taxis Furthermore because such individuals are not considering any

other modes they may treat taxi usage as a kind of PT mode with poor service

As expected greater access to cars in a household (Car_acc) lessens the likelihood of considering

taxis as an alternative Furthermore individuals in households with more motorcycle ownership are

less likely to consider taxis It seems like there is a competition among motorcycles and taxis for

18

access to the city center Younger commuters are less likely to use taxis and individuals with at least

master degrees do consider this mode in addition to their car

Public transit accessed by Drive (DampR)

This mode is affected by the simultaneous interaction of transit time and transit access

(PT_TimeampAccess) which is reflected in the fact that individuals prefer to use this mode for longer

trips Comparing this mode and WampR the first trip start time affects the consideration of this mode

differently Later morning commuters prefer to use their car to access PT modes Such commuters

may have higher income levels or managerial jobs Obviously individuals who are not able to access

PT stations by walking (PTnwacc) are likely to use DampR Serving passengers in daily trips is also a

deterrent in considering this mode which is similar to WampR but with a lower coefficient

Commuters with higher income levels who depend on their car during before or after work

(Dependencycar1+) are likely to use DampR Individuals who use their own car are less likely to use

this mode which is unexpected As a city center develops better PT network coverage and residents

have smaller distances to their workplaces they are unlikely to use DampR This is proven in the model

by a negative sign for D_home_place

Motorcycle (MC)

Increasing fuel cost and cordon pricing simultaneously discourage motorcycle usages Although fuel

cost is expected to reduce motorcycle usage to some extent its combined effect with cordon pricing

also reduces motorcycle usage However this variable is not as strong as other policy variables

=10)

Of the studied modes motorcycle usage is affected by the most commuting variables This may be

due to the fact that this mode is not common Commuting distance has a negative effect on

motorcycle usage which is expected It is worth noting that trip distance appears only in this mode

which may be a reflection of the role of distance in regards to the safety risk in considering this

19

mode Commuters with more stops to serve passengers while commuting (Pattern 2) are not likely

to use this mode which may be due to the poor passenger service of this mode

Individuals who state that commuting is independent of the mode (Pnocarwk) are not likely to use

MC By looking at the (First_Nacco) negative sign this could stem from the fact that the more

passengers there are on the first trip the less likely individuals are to consider MC Regarding the

low capacity of MC and its safety concerns such commuters avoid using this mode Commuters who

pay more parking charges (Park_payment) are less likely to use MC which is expected Individuals

who are dependent on their car during before or after their work time are not likely to use MC

even if they have lower levels of income (DependencyCar1) Individuals who use their own car

(D_car_own) are less likely to use this mode As expected individuals who live in households with

more motorcycle ownership are more likely to use this mode The positive sign of (Permission)

indicates that commuters who have permission to enter the study area do consider MC Because

such commuters generally provide that permission just for car usage this result is unexpected

As with commute variables of all the studied modes MC is affected by the greatest number of

socio-economic variables As expected young commuters (Agelt30) are more likely to use this mode

Commuters with Bachelor of Science degree are less likely to use this mode among others Full time

employees (Emp_full) are less likely to consider MC whereas commuters with more experience in

their jobs prefer to use it Results show that individuals who live in a household with children

younger than 18 are more likely to consider using a car

Tel-Taxi (T_T)

Results show that cordon pricing causes higher probability of using T_T In fact individuals who use

T_T as a mode with similar level of service as cars9 are more willing to pay the cost and make use of

the mode It is worth noting that the effect of cordon pricing in pushing commuters from car usage

9 As this mode does not have driving stress and parking search time in some cases it may have more amount of utility thana car does

20

(000045) is greater than its effect on pulling them to Tel-taxi (000019) This is because of the

possibility of considering other non-car modes

Because consideration of this mode is a function of its operation travel time (Trip_time) appears as

a deterrent in this mode utility function Table 5 shows that individuals are more sensitive to the trip

time when using T_T mode versus taxi which is expected due to their relative costs

The greater the number of full time employees in a family (Nhempfull) the higher the probability of

considering T_T by its commuters which may be due to the higher income level of these

households This is verified by the greater likelihood of using T_T rather than taxis by such

commuters Individuals with higher levels of income who depend on their car during before or after

work time are less likely to use T_T Commuters with lower income levels who state that they use

their car for the sake of comfort (Comfortcar1) are less likely to use T_T which may be due to its

cost Although such individuals do not consider any other modes they specifically avoid T_T Greater

access to cars in a household leads to greater likelihood of T_T usage which could be due to the

higher income level of a household As mentioned before such individuals even avoid taxis

Females who drive to their workplace are more likely to use T_T It seems like this part of society

considers this mode when desiring to avoid the difficulties of driving Younger commuters are less

likely to use T_T and individuals between 30 and 39 years of age are specifically avoiding this mode

Results show that university graduated commuters are more likely to use this mode

6 Marginal effects

To explore the effects of each policy on mode choice and to answer the second issue raised at the

beginning of this paper the marginal effects approach can be adopted Although the coefficients of

the models utility functions show the drivers behavior when facing one or more policies the

marginal effects of policies or their interactions may appropriately show the results of their

implementation More specifically the marginal effect for this study is interpreted as the change in

21

probability given a unit change in a variable ceteris paribus In this section the variable is defined as

a specific policy or a specific interaction of two policies Table 6 presents the marginal effects of the

studied policies and their interactions with mode choice The results are shown in the form of trip

percentages transferred away from the car to the studied modes and the probability-weighted

sample enumeration approach is adopted to find the values It is worth noting that this table is fully

compatible with Table 5 but the marginal effects that were less significant than 90 percent have

been removed

Table 6 - Marginal effects of policies (percent)

Tel-Taxi(TT)

Motorcycle(MC)

Drive ampRide(DampR)

Taxi (T)Walk ampRide(WampR)

Car (C)Mode

Variable-000088Cordon-000140Parking

-09069Access-0000001ParkampFuel

00040PT_TimeampAccess

Table 5 shows that a 1 Rial increase in cordon pricing causes a 000088 percent decrease in car

usage Furthermore a 1 Rial per hour increase in parking cost decreases the car usage probability to

00014 percent By assuming 8 hours for the average parking duration the daily marginal value of

parking cost converts to 000018 percent These values show that cordon pricing is more effective in

forcing individuals not to use their car than increasing parking cost with the same value Results also

show that a 1 Rial increase in parking cost and fuel cost simultaneously decreases the probability of

choosing car usage by 0000001 percent Table 5 also shows that a 1 minute decrease in transit

access time would result in a 09 percent increase in probability of choosing this mode It also shows

that each 1 minute decrease in transit travel time and transit access leads to a 0004 decrease in the

probability of choosing the DampR mode

22

The above results show that just one of the policy interactions is not significantly different from

none (ie CordonampFuel) and push policies have a significant role in impelling car users to consider

other modes

7 Planning

To demonstrate the application of the model some of its planning aspects for the city of Tehran are

presented here In fact the model is able to assess the effect of all studied variables which were

grouped into three tiers on commuters mode choice In this section we assess the effect of TDM

policies by adopting their pair-wise combinations for the city Table 6 shows that pull policies in this

study did not impose a major effect on car usage Because the average access time of the sample

was 11 minutes improving it by 50 percent (about 5 minutes) increases the probability of WampR to

509069100 which is equal to a 048 percent increase in considering this mode (and decrease in all

other modes) On the other hand decreasing transit time about 50 percent (about 19 minutes based

on sample average) and similarly improving access time decreases the probability of considering

DampR to 0004195100 which is equal to a 038 percent decrease in considering this mode (and

increase in all other modes) Therefore in this section we focused on the push policies which

appear in the utility function of car mode It is worth noting that in assessing each combination of

policies all other variables were assumed to be fixed as the current state

71 Parking pricing and cordon pricing

The average hourly parking charge of individuals in the sample is 71 Rials This charge is 3000 Rials

per hour in the curb lane on some streets about 2000 Rials per hour in parking lots and free in

alleys and other streets The average amount shows that most of the commuters benefited from

free parking spaces in alleys and non-controlled streets Figure 2 depicts the result of implementing

this policy and the cordon pricing policy simultaneously In line with marginal effects it can be seen

that in implementing each policy separately cordon pricing is more effective than parking pricing in

the planning range It is worth noting that commuters in this study are less sensitive to parking

23

pricing than cordon pricing One reason may be that they thought in any case they would find free

parking space Figure 2 shows that the effect of each policy depends on the level of the other policy

In other words the effect of the simultaneous implementation of two policies is not equal to the

sum of their separate individual effects

Figure 2- Effect of implementing parking pricing and cordon pricing

72 Parking pricing and increasing fuel cost

Figure 3 shows the simultaneous effects of parking pricing and increasing fuel cost policies Although

both of these policies are not effective separately their simultaneous effect is significant It can be

seen that their synergy appears in the higher levels of both policies

Figure 3 - Effect of implementing parking pricing and increasing fuel cost

24

73 Cordon pricing and increasing fuel cost

The effect of implementing cordon pricing and increasing fuel cost is shown in Figure 4 It can be

seen that these two policies have no interaction effect in decreasing car usage This result is verified

by this interactions insignificant marginal value in Table 6

Figure 4- Effect of implementing cordon pricing and increasing fuel cost

8 Conclusions

This study examined the role of TDM policies in individual mode choice for work trips in the city of

Tehran Five policies including increasing parking cost increasing fuel cost cordon pricing in the odd-

even zone of the city transit time reduction and transit access improvement were investigated Of

the five policies the former three were push policies and the latter two were pull policies The

design of experiments approach was used to design the questionnaire that would capture the stated

preferences of car commuters

A number of nested structures based on recognizing differences in the variances associated

with unobserved influences were examined and finally the MNL as a superior model for this study

was developed

25

The results show that for car users who faced TDM policies all of the five studied TDM

policies were factors in the mode choice process The model shows that in considering car usage

push policies play a main role and pull policies do not

Results also show that the interactions of policies are also significant The model shows that

the interaction of parking cost policy and fuel cost policy is significant in prompting car users to

consider other modes It also shows that the interaction of cordon pricing and increase in fuel cost is

effective to discourage motorcycle usage Furthermore the interaction of transit time reduction

policy and access improvements policy is significant in considering DampR as a transportation mode to

workplace

Pull policies in the study were expected to attract individuals to transit modes (pull policies

of this study are related to transit modes) This was verified by the model

Results show that although taxi usage is usually treated as an alternative for urban trips it

was not directly affected by the studied policies This may be due to its special function in Tehran as

a non-public and non-private mode

The results also show that except for the interaction of cordon pricing and increase in fuel

cost other interaction effects have significant marginal effects on mode choice

Results of the model show that to implement a single policy cordon pricing is the most

effective in decreasing car usage

This study assessed workplace commuters Thus future studies should explore the

individuals mode choice with other trip aims Additionally including more policies especially pull

policies in favor of non-transit modes is reasonable to rigorously assess the model

It can be seen that the studied modes are affected differently by the TDM policies and their

interactions The impact of the interaction of policies appeared in the consideration of three out of

six studied modes and also in two out of three developed graphs This issue needs more attention

within the transportation demand management context and further studies should be performed

26

Design of this study is based on the consideration of all two-way interactions with some

degrees of deficiency in D-efficient policy Focusing on the significant interactions of this study is

recommended to perform other studies with more D-efficient values

Further understanding of the impacts of TDM policies might be provided in future research

by focusing on the amount of their synergy in prohibiting car usage

9 References

Arentze TA Hofman F amp Timmermans HJP (2004) Predicting multi-faceted activityndashtravel adjustmentstrategies in response to possible congestion pricing scenarios using an Internet-based stated adaptationexperiment Transport Policy 11 p31ndash41

Burris M amp Patil S (2009) Estimating the Benefits of Managed Lanes Technical Report Texas TexasTransportation Institute University Transportation Center for Mobility

Cao X amp Mokhtarian PL (2005) How do individuals adapt their personal travel A conceptual exploration ofthe consideation of travel-related strategies Transport Policy 12(3) pp199-206

de Palma A amp Lindsey R (2001) Transportation Supply and congestion In International Encyclopedia of theSocial and Behavioral Sciences 1st ed Elsevier p15882ndash15888

Eriksson L Garvill J amp Nordlund AM (2008) Acceptability of single and combined transport policy measuresThe importance of environmental and policy specific beliefs Transportation Research Part A 42 p1117ndash1128

Eriksson L Nordlund AM amp Garvill J (2010) Expected car use reduction in response to structural traveldemand management measures Transportation Research Part F 13 p329ndash342

Golob TF amp Hensher DA (2007) The Trip chaining Activity of Sydney Residents A Cross-Section Assessmentby Age Group with a Focus on Seniors Journal of Transport Geography 15 p298ndash312

Hensher DA amp King J (2001) Parking demand and responsiveness to supply pricing and location in theSydney central business district Transportation Research A 35 pp177-96

Hensher DA amp Rose JM (2007) Development of commuter and non-commuter mode choice models for theassessment of new public transport infrastructure projects A case study Transportation Research Part A41 p428ndash443

Hensher DA Rose JM amp Greene WH (2005) Applied Choice Analysis A Primer New York CambridgeUniversity Press

Iranian Center of Statistics (ICS) (2009) Information of Iranian States [Online] Available at HYPERLINK httpwwwamarorgirUploadModulesContentsasset16tehrantehparthtmlhttpwwwamarorgirUploadModulesContentsasset16tehrantehparthtml [Accessed 13 November2009]

Kingham S Dickinson J amp Copsey S (2001) Travelling to work Will people move out of their cars TransportPolicy 8 pp151-60

Kuhfeld WF (2009) Marketing Research Methods in SAS Experimental Design Choice Conjoint andGraphical Techniques SAS 92 Edition NC USA SAS Institute Inc Cary Available on the Webwwwsupportsascomtechsuptnotetnote_stathtmlmarket

Litman T (2003) The online TDM Encyclopedia mobility management information gateway Transport Policy10 pp245-49

Litman T (2010) Online TDM Encyclopedia [Online] Available at HYPERLINK httpwwwvtpiorghttpwwwvtpiorg [Accessed 21 Augest 2010]

13

We have (0889-1)2508 =-075 which would be rejected at the usual acceptable significance levels

This suggests that the NL model could be collapsed into an MNL form

Table 3- Nested logit (NL) model resultValueParameter

0889IV (Private)1000IV (nPrivate)

-2668335L( )-4057684L(0)

0342sup2

After the calibration process the variables that were statistically significant were identified and are

presented in Table 4 Table 5 presents the final model of the study with a goodness of fit of 031 for

the six studied modes For a general review of the model calibration results the effective factors can

be grouped under the following three categories TDM policy characteristics commuting trip

characteristics and household socio-economic characteristics which are all treated as alternative-

specific variables

52 Model results

Car (C)

It is expected that push policies impel car-drivers to choose other modes Table 5 shows that cordon

pricing and increase in parking cost cause individuals to choose not to use their car This is in line

with other studies suggesting that these policies are effective to discourage car usage (Hensher amp

Rose 2007 OFallon et al 2004) In addition the interaction between the policies of fuel cost

increase and increase in parking cost shows similar car usage discourage effect Because fuel cost is

related to the distance between home and work locations and parking cost is related to work time

the time that an individual spends out of the home is negatively affected by hisher likelihood to use

a car

14

Table 4 - Definition of the significant variables

AbbreviationVariableTransportation demand management measures

Measures

ParkingParking cost increase Rials per hour

CordonCordon price Rials per entranceAccessTransit access time shortage percent

Interaction of push measures

ParkampFuelParking cost and fuel cost simultaneous effectsCordonampFuelCordon pricing and fuel cost simultaneous effects

Interaction of pull measures

PT_timeampaccessPT time reduction and access improvement simultaneous effectsCommuting trip characteristics

Trip distanceDistance between home and workplaceTrip timeTravel time between home and workplace

Exp FuelLikelihood of unsubsidized fuel use (self-reported on a Likert scale)NtripsNumber of daily tripsPattern2Commuting with 1+ stop(s) in go or return

Pattern3Commuting with 2 workplacesFirst trip timeStart time of first trip

PnocarwkLikelihood of going to work in absence of that car (self-reported)PTnwaccNon-walk access to transit (yes=1)First NaccoNumber of passengers in first trip

PassengerAny passenger on that day (yes=1)Park_paymentParking payment in last weekNhempfullNumber of full employees in HH

CardependencyBoardalight a passenger or move freight in the trip (yes=1)D car ownBe the owner of the used vehicle (yes=1)

Car accCar accessibility in household (number of cars to number of HH driving licenses ratio)NmotorcycleNumber of motorcycles owned by HHD home placeHome Location is in study area (yes=1)

PermissionPermission to enter to study area (yes=1)ComfortI use my car because it is comfortablePoor_PTI use my car because transit is not good

HH socio-economic characteristics

FemaleGender (Female=1)Age lt30Age younger than 30 (yes=1)Age 30_39Age between 30 to 39 (yes=1)Job_durationNumber of years that individual has been at hisher job

Emp_fullFull-time employee (yes=1)Edu BSDegree of education is BSc (yes=1)Edu BS+Degree of education is higher than BSc(yes=1)

D childlt=18Child younger than 18 in HH (yes=1)

15

Table 5 ndash The mode choice model

Tel-Taxi(T_T)

Motorcycle(MC)

Drive amp Ride(DampR)

Taxi (T)Walk amp Ride(WampR)

Car (C)Mode

Variable-471756-37067-147911Constant

Transportation demand management measure variables00019-00045Cordon

-000072Parking-004308Access

-28443D-05Parkampfuel-32475D-06Cordonampfuel

00029Pt_timeampaccess

Commuting trip characteristics-04709Trip distance

-02163-00831Trip time-96755163655Exp fuel-16253Ntrips

-114779Pattern2-71008Pattern3

00282-00270First trip time-02439-01549Pnocarwk

-11322992883-32765PTnwacc-133701First Nacco

-7778-73782Accompany-00049000010Park_payment

201646195554Nhempfull-160144ComfortCar1

-206142DependencyCar1-16101883385-121224DependencyCar2

42176Poor_PTCar1-24988Poor_PTCar2

- -27221D car own70960-39136Car acc

1 -71112-156123Nmotorcycle-1436322762D home place

2 78826Permission

HH socio-economic characteristics149490Female

297584-24548Agelt30-136490Age30_39

079430366303585Job_duration-108743Emp_full-203468-64900Edu BS

10932856687-4499984445Edu BS+102271D childlt=18

-2677366L( )-3849556L(0)0305sup2

112127178592580607N

Note = Positive significance at 1 5 10 level

As expected individuals with higher income are more likely to use their car This is indicated in the

model by the positive signs of individuals who use fuel with fixed (unsubsidized) cost and individuals

16

who pay more in parking charges in the previous week of study Negative sign of Pnocarwk variable

shows that the commuters who stated that their commute depends on car availability are more

likely to use their car Individuals in households with more full-time employees are more likely to use

their car which may be the result of higher household income Not surprisingly commuters who

have permission are more likely to maintain car usage Among the household socio-economic

parameters greater job experience (Job_duration) and higher graduate levels (EduBS+) increase the

probability of car usage

Public transit accessed by walking (WampR)

Access time to transit negatively impacts WampR choice which is expected This result is similar to

findings for the city of Sydney (Hensher amp Rose 2007) The negative coefficient of first trip time

indicates that individuals are more likely to use WampR in the early morning This result seems to

reflect the better weather for walking and faster speed of WampR mode early in the morning

Obviously individuals who are not able to access transit stations via walking (PTnwacc) are less likely

to consider this mode Furthermore serving passengers on daily trips is also a deterrent to using

WampR

Initially assessing the individuals who stated that their car usage is due to poor public transit service

(Poor_PT) led to an unexpected result in favor of considering WampR By introducing to this variable

the number of household cars as a proxy for household income (Poor_PTCar1) the model shows

that of the previously mentioned individuals those who have lower income are the ones who have

to consider WampR The result is understandable as these individuals may have no alternative when

they have to change their mode (they also are not likely to consider other modes) Individuals with

higher levels of income who have to use their car during before or after work (Dependencycar1+)

are not likely to use WampR

The greater the number of motorcycles in a household the less likely commuters is to consider

WampR There appears to be a competition between motorcycle and PT for access to the city center

17

Better PT services in the center of the city in terms of coverage and frequency increases the

likelihood that its residents will consider WampR This is verified by the positive sign of the

D_home_place variable Commuters with greater job experience (Job_duration) in their workplace

are more likely to use this mode Although individuals with higher levels of education are not likely

to use WampR as education level increases avoidance of WampR decreases

Taxi (T)

Table 5 shows that none of the studied policies are significant in considering taxi usage It seems that

taxi usage considering its function in Iran as a non-private and non-public mode of transport is not

affected by pull or push policies A negative sign for taxi travel time indicates that individuals are not

likely to use this mode for longer trips This seems reasonable given that longer trips are more

expensive Commuters who are more likely to use fuel with no subsidy are not likely to use taxis As

mentioned before they prefer to use their car A higher number of trips in a day are also a deterrent

to considering taxi usage which may be due to increased cost for more trips Results show that an

individual with more daily trips avoids using taxis Commuters who are employed in more than one

workplace (Pattern 3) are not likely to use taxis This may be due to the fact that they have a lower

level of income which forces them to dedicate more time on the job

Initial results showed that individuals who stated that their car usage is due to poor public transit

service (Poor_PT) are not likely to use taxis This result was far from our expectations By introducing

to this variable the number of household cars as a proxy for household income the model shows

that the previously mentioned individuals who have higher income (Poor_PTCar1+) are the ones

who are not likely to consider taxis Furthermore because such individuals are not considering any

other modes they may treat taxi usage as a kind of PT mode with poor service

As expected greater access to cars in a household (Car_acc) lessens the likelihood of considering

taxis as an alternative Furthermore individuals in households with more motorcycle ownership are

less likely to consider taxis It seems like there is a competition among motorcycles and taxis for

18

access to the city center Younger commuters are less likely to use taxis and individuals with at least

master degrees do consider this mode in addition to their car

Public transit accessed by Drive (DampR)

This mode is affected by the simultaneous interaction of transit time and transit access

(PT_TimeampAccess) which is reflected in the fact that individuals prefer to use this mode for longer

trips Comparing this mode and WampR the first trip start time affects the consideration of this mode

differently Later morning commuters prefer to use their car to access PT modes Such commuters

may have higher income levels or managerial jobs Obviously individuals who are not able to access

PT stations by walking (PTnwacc) are likely to use DampR Serving passengers in daily trips is also a

deterrent in considering this mode which is similar to WampR but with a lower coefficient

Commuters with higher income levels who depend on their car during before or after work

(Dependencycar1+) are likely to use DampR Individuals who use their own car are less likely to use

this mode which is unexpected As a city center develops better PT network coverage and residents

have smaller distances to their workplaces they are unlikely to use DampR This is proven in the model

by a negative sign for D_home_place

Motorcycle (MC)

Increasing fuel cost and cordon pricing simultaneously discourage motorcycle usages Although fuel

cost is expected to reduce motorcycle usage to some extent its combined effect with cordon pricing

also reduces motorcycle usage However this variable is not as strong as other policy variables

=10)

Of the studied modes motorcycle usage is affected by the most commuting variables This may be

due to the fact that this mode is not common Commuting distance has a negative effect on

motorcycle usage which is expected It is worth noting that trip distance appears only in this mode

which may be a reflection of the role of distance in regards to the safety risk in considering this

19

mode Commuters with more stops to serve passengers while commuting (Pattern 2) are not likely

to use this mode which may be due to the poor passenger service of this mode

Individuals who state that commuting is independent of the mode (Pnocarwk) are not likely to use

MC By looking at the (First_Nacco) negative sign this could stem from the fact that the more

passengers there are on the first trip the less likely individuals are to consider MC Regarding the

low capacity of MC and its safety concerns such commuters avoid using this mode Commuters who

pay more parking charges (Park_payment) are less likely to use MC which is expected Individuals

who are dependent on their car during before or after their work time are not likely to use MC

even if they have lower levels of income (DependencyCar1) Individuals who use their own car

(D_car_own) are less likely to use this mode As expected individuals who live in households with

more motorcycle ownership are more likely to use this mode The positive sign of (Permission)

indicates that commuters who have permission to enter the study area do consider MC Because

such commuters generally provide that permission just for car usage this result is unexpected

As with commute variables of all the studied modes MC is affected by the greatest number of

socio-economic variables As expected young commuters (Agelt30) are more likely to use this mode

Commuters with Bachelor of Science degree are less likely to use this mode among others Full time

employees (Emp_full) are less likely to consider MC whereas commuters with more experience in

their jobs prefer to use it Results show that individuals who live in a household with children

younger than 18 are more likely to consider using a car

Tel-Taxi (T_T)

Results show that cordon pricing causes higher probability of using T_T In fact individuals who use

T_T as a mode with similar level of service as cars9 are more willing to pay the cost and make use of

the mode It is worth noting that the effect of cordon pricing in pushing commuters from car usage

9 As this mode does not have driving stress and parking search time in some cases it may have more amount of utility thana car does

20

(000045) is greater than its effect on pulling them to Tel-taxi (000019) This is because of the

possibility of considering other non-car modes

Because consideration of this mode is a function of its operation travel time (Trip_time) appears as

a deterrent in this mode utility function Table 5 shows that individuals are more sensitive to the trip

time when using T_T mode versus taxi which is expected due to their relative costs

The greater the number of full time employees in a family (Nhempfull) the higher the probability of

considering T_T by its commuters which may be due to the higher income level of these

households This is verified by the greater likelihood of using T_T rather than taxis by such

commuters Individuals with higher levels of income who depend on their car during before or after

work time are less likely to use T_T Commuters with lower income levels who state that they use

their car for the sake of comfort (Comfortcar1) are less likely to use T_T which may be due to its

cost Although such individuals do not consider any other modes they specifically avoid T_T Greater

access to cars in a household leads to greater likelihood of T_T usage which could be due to the

higher income level of a household As mentioned before such individuals even avoid taxis

Females who drive to their workplace are more likely to use T_T It seems like this part of society

considers this mode when desiring to avoid the difficulties of driving Younger commuters are less

likely to use T_T and individuals between 30 and 39 years of age are specifically avoiding this mode

Results show that university graduated commuters are more likely to use this mode

6 Marginal effects

To explore the effects of each policy on mode choice and to answer the second issue raised at the

beginning of this paper the marginal effects approach can be adopted Although the coefficients of

the models utility functions show the drivers behavior when facing one or more policies the

marginal effects of policies or their interactions may appropriately show the results of their

implementation More specifically the marginal effect for this study is interpreted as the change in

21

probability given a unit change in a variable ceteris paribus In this section the variable is defined as

a specific policy or a specific interaction of two policies Table 6 presents the marginal effects of the

studied policies and their interactions with mode choice The results are shown in the form of trip

percentages transferred away from the car to the studied modes and the probability-weighted

sample enumeration approach is adopted to find the values It is worth noting that this table is fully

compatible with Table 5 but the marginal effects that were less significant than 90 percent have

been removed

Table 6 - Marginal effects of policies (percent)

Tel-Taxi(TT)

Motorcycle(MC)

Drive ampRide(DampR)

Taxi (T)Walk ampRide(WampR)

Car (C)Mode

Variable-000088Cordon-000140Parking

-09069Access-0000001ParkampFuel

00040PT_TimeampAccess

Table 5 shows that a 1 Rial increase in cordon pricing causes a 000088 percent decrease in car

usage Furthermore a 1 Rial per hour increase in parking cost decreases the car usage probability to

00014 percent By assuming 8 hours for the average parking duration the daily marginal value of

parking cost converts to 000018 percent These values show that cordon pricing is more effective in

forcing individuals not to use their car than increasing parking cost with the same value Results also

show that a 1 Rial increase in parking cost and fuel cost simultaneously decreases the probability of

choosing car usage by 0000001 percent Table 5 also shows that a 1 minute decrease in transit

access time would result in a 09 percent increase in probability of choosing this mode It also shows

that each 1 minute decrease in transit travel time and transit access leads to a 0004 decrease in the

probability of choosing the DampR mode

22

The above results show that just one of the policy interactions is not significantly different from

none (ie CordonampFuel) and push policies have a significant role in impelling car users to consider

other modes

7 Planning

To demonstrate the application of the model some of its planning aspects for the city of Tehran are

presented here In fact the model is able to assess the effect of all studied variables which were

grouped into three tiers on commuters mode choice In this section we assess the effect of TDM

policies by adopting their pair-wise combinations for the city Table 6 shows that pull policies in this

study did not impose a major effect on car usage Because the average access time of the sample

was 11 minutes improving it by 50 percent (about 5 minutes) increases the probability of WampR to

509069100 which is equal to a 048 percent increase in considering this mode (and decrease in all

other modes) On the other hand decreasing transit time about 50 percent (about 19 minutes based

on sample average) and similarly improving access time decreases the probability of considering

DampR to 0004195100 which is equal to a 038 percent decrease in considering this mode (and

increase in all other modes) Therefore in this section we focused on the push policies which

appear in the utility function of car mode It is worth noting that in assessing each combination of

policies all other variables were assumed to be fixed as the current state

71 Parking pricing and cordon pricing

The average hourly parking charge of individuals in the sample is 71 Rials This charge is 3000 Rials

per hour in the curb lane on some streets about 2000 Rials per hour in parking lots and free in

alleys and other streets The average amount shows that most of the commuters benefited from

free parking spaces in alleys and non-controlled streets Figure 2 depicts the result of implementing

this policy and the cordon pricing policy simultaneously In line with marginal effects it can be seen

that in implementing each policy separately cordon pricing is more effective than parking pricing in

the planning range It is worth noting that commuters in this study are less sensitive to parking

23

pricing than cordon pricing One reason may be that they thought in any case they would find free

parking space Figure 2 shows that the effect of each policy depends on the level of the other policy

In other words the effect of the simultaneous implementation of two policies is not equal to the

sum of their separate individual effects

Figure 2- Effect of implementing parking pricing and cordon pricing

72 Parking pricing and increasing fuel cost

Figure 3 shows the simultaneous effects of parking pricing and increasing fuel cost policies Although

both of these policies are not effective separately their simultaneous effect is significant It can be

seen that their synergy appears in the higher levels of both policies

Figure 3 - Effect of implementing parking pricing and increasing fuel cost

24

73 Cordon pricing and increasing fuel cost

The effect of implementing cordon pricing and increasing fuel cost is shown in Figure 4 It can be

seen that these two policies have no interaction effect in decreasing car usage This result is verified

by this interactions insignificant marginal value in Table 6

Figure 4- Effect of implementing cordon pricing and increasing fuel cost

8 Conclusions

This study examined the role of TDM policies in individual mode choice for work trips in the city of

Tehran Five policies including increasing parking cost increasing fuel cost cordon pricing in the odd-

even zone of the city transit time reduction and transit access improvement were investigated Of

the five policies the former three were push policies and the latter two were pull policies The

design of experiments approach was used to design the questionnaire that would capture the stated

preferences of car commuters

A number of nested structures based on recognizing differences in the variances associated

with unobserved influences were examined and finally the MNL as a superior model for this study

was developed

25

The results show that for car users who faced TDM policies all of the five studied TDM

policies were factors in the mode choice process The model shows that in considering car usage

push policies play a main role and pull policies do not

Results also show that the interactions of policies are also significant The model shows that

the interaction of parking cost policy and fuel cost policy is significant in prompting car users to

consider other modes It also shows that the interaction of cordon pricing and increase in fuel cost is

effective to discourage motorcycle usage Furthermore the interaction of transit time reduction

policy and access improvements policy is significant in considering DampR as a transportation mode to

workplace

Pull policies in the study were expected to attract individuals to transit modes (pull policies

of this study are related to transit modes) This was verified by the model

Results show that although taxi usage is usually treated as an alternative for urban trips it

was not directly affected by the studied policies This may be due to its special function in Tehran as

a non-public and non-private mode

The results also show that except for the interaction of cordon pricing and increase in fuel

cost other interaction effects have significant marginal effects on mode choice

Results of the model show that to implement a single policy cordon pricing is the most

effective in decreasing car usage

This study assessed workplace commuters Thus future studies should explore the

individuals mode choice with other trip aims Additionally including more policies especially pull

policies in favor of non-transit modes is reasonable to rigorously assess the model

It can be seen that the studied modes are affected differently by the TDM policies and their

interactions The impact of the interaction of policies appeared in the consideration of three out of

six studied modes and also in two out of three developed graphs This issue needs more attention

within the transportation demand management context and further studies should be performed

26

Design of this study is based on the consideration of all two-way interactions with some

degrees of deficiency in D-efficient policy Focusing on the significant interactions of this study is

recommended to perform other studies with more D-efficient values

Further understanding of the impacts of TDM policies might be provided in future research

by focusing on the amount of their synergy in prohibiting car usage

9 References

Arentze TA Hofman F amp Timmermans HJP (2004) Predicting multi-faceted activityndashtravel adjustmentstrategies in response to possible congestion pricing scenarios using an Internet-based stated adaptationexperiment Transport Policy 11 p31ndash41

Burris M amp Patil S (2009) Estimating the Benefits of Managed Lanes Technical Report Texas TexasTransportation Institute University Transportation Center for Mobility

Cao X amp Mokhtarian PL (2005) How do individuals adapt their personal travel A conceptual exploration ofthe consideation of travel-related strategies Transport Policy 12(3) pp199-206

de Palma A amp Lindsey R (2001) Transportation Supply and congestion In International Encyclopedia of theSocial and Behavioral Sciences 1st ed Elsevier p15882ndash15888

Eriksson L Garvill J amp Nordlund AM (2008) Acceptability of single and combined transport policy measuresThe importance of environmental and policy specific beliefs Transportation Research Part A 42 p1117ndash1128

Eriksson L Nordlund AM amp Garvill J (2010) Expected car use reduction in response to structural traveldemand management measures Transportation Research Part F 13 p329ndash342

Golob TF amp Hensher DA (2007) The Trip chaining Activity of Sydney Residents A Cross-Section Assessmentby Age Group with a Focus on Seniors Journal of Transport Geography 15 p298ndash312

Hensher DA amp King J (2001) Parking demand and responsiveness to supply pricing and location in theSydney central business district Transportation Research A 35 pp177-96

Hensher DA amp Rose JM (2007) Development of commuter and non-commuter mode choice models for theassessment of new public transport infrastructure projects A case study Transportation Research Part A41 p428ndash443

Hensher DA Rose JM amp Greene WH (2005) Applied Choice Analysis A Primer New York CambridgeUniversity Press

Iranian Center of Statistics (ICS) (2009) Information of Iranian States [Online] Available at HYPERLINK httpwwwamarorgirUploadModulesContentsasset16tehrantehparthtmlhttpwwwamarorgirUploadModulesContentsasset16tehrantehparthtml [Accessed 13 November2009]

Kingham S Dickinson J amp Copsey S (2001) Travelling to work Will people move out of their cars TransportPolicy 8 pp151-60

Kuhfeld WF (2009) Marketing Research Methods in SAS Experimental Design Choice Conjoint andGraphical Techniques SAS 92 Edition NC USA SAS Institute Inc Cary Available on the Webwwwsupportsascomtechsuptnotetnote_stathtmlmarket

Litman T (2003) The online TDM Encyclopedia mobility management information gateway Transport Policy10 pp245-49

Litman T (2010) Online TDM Encyclopedia [Online] Available at HYPERLINK httpwwwvtpiorghttpwwwvtpiorg [Accessed 21 Augest 2010]

14

Table 4 - Definition of the significant variables

AbbreviationVariableTransportation demand management measures

Measures

ParkingParking cost increase Rials per hour

CordonCordon price Rials per entranceAccessTransit access time shortage percent

Interaction of push measures

ParkampFuelParking cost and fuel cost simultaneous effectsCordonampFuelCordon pricing and fuel cost simultaneous effects

Interaction of pull measures

PT_timeampaccessPT time reduction and access improvement simultaneous effectsCommuting trip characteristics

Trip distanceDistance between home and workplaceTrip timeTravel time between home and workplace

Exp FuelLikelihood of unsubsidized fuel use (self-reported on a Likert scale)NtripsNumber of daily tripsPattern2Commuting with 1+ stop(s) in go or return

Pattern3Commuting with 2 workplacesFirst trip timeStart time of first trip

PnocarwkLikelihood of going to work in absence of that car (self-reported)PTnwaccNon-walk access to transit (yes=1)First NaccoNumber of passengers in first trip

PassengerAny passenger on that day (yes=1)Park_paymentParking payment in last weekNhempfullNumber of full employees in HH

CardependencyBoardalight a passenger or move freight in the trip (yes=1)D car ownBe the owner of the used vehicle (yes=1)

Car accCar accessibility in household (number of cars to number of HH driving licenses ratio)NmotorcycleNumber of motorcycles owned by HHD home placeHome Location is in study area (yes=1)

PermissionPermission to enter to study area (yes=1)ComfortI use my car because it is comfortablePoor_PTI use my car because transit is not good

HH socio-economic characteristics

FemaleGender (Female=1)Age lt30Age younger than 30 (yes=1)Age 30_39Age between 30 to 39 (yes=1)Job_durationNumber of years that individual has been at hisher job

Emp_fullFull-time employee (yes=1)Edu BSDegree of education is BSc (yes=1)Edu BS+Degree of education is higher than BSc(yes=1)

D childlt=18Child younger than 18 in HH (yes=1)

15

Table 5 ndash The mode choice model

Tel-Taxi(T_T)

Motorcycle(MC)

Drive amp Ride(DampR)

Taxi (T)Walk amp Ride(WampR)

Car (C)Mode

Variable-471756-37067-147911Constant

Transportation demand management measure variables00019-00045Cordon

-000072Parking-004308Access

-28443D-05Parkampfuel-32475D-06Cordonampfuel

00029Pt_timeampaccess

Commuting trip characteristics-04709Trip distance

-02163-00831Trip time-96755163655Exp fuel-16253Ntrips

-114779Pattern2-71008Pattern3

00282-00270First trip time-02439-01549Pnocarwk

-11322992883-32765PTnwacc-133701First Nacco

-7778-73782Accompany-00049000010Park_payment

201646195554Nhempfull-160144ComfortCar1

-206142DependencyCar1-16101883385-121224DependencyCar2

42176Poor_PTCar1-24988Poor_PTCar2

- -27221D car own70960-39136Car acc

1 -71112-156123Nmotorcycle-1436322762D home place

2 78826Permission

HH socio-economic characteristics149490Female

297584-24548Agelt30-136490Age30_39

079430366303585Job_duration-108743Emp_full-203468-64900Edu BS

10932856687-4499984445Edu BS+102271D childlt=18

-2677366L( )-3849556L(0)0305sup2

112127178592580607N

Note = Positive significance at 1 5 10 level

As expected individuals with higher income are more likely to use their car This is indicated in the

model by the positive signs of individuals who use fuel with fixed (unsubsidized) cost and individuals

16

who pay more in parking charges in the previous week of study Negative sign of Pnocarwk variable

shows that the commuters who stated that their commute depends on car availability are more

likely to use their car Individuals in households with more full-time employees are more likely to use

their car which may be the result of higher household income Not surprisingly commuters who

have permission are more likely to maintain car usage Among the household socio-economic

parameters greater job experience (Job_duration) and higher graduate levels (EduBS+) increase the

probability of car usage

Public transit accessed by walking (WampR)

Access time to transit negatively impacts WampR choice which is expected This result is similar to

findings for the city of Sydney (Hensher amp Rose 2007) The negative coefficient of first trip time

indicates that individuals are more likely to use WampR in the early morning This result seems to

reflect the better weather for walking and faster speed of WampR mode early in the morning

Obviously individuals who are not able to access transit stations via walking (PTnwacc) are less likely

to consider this mode Furthermore serving passengers on daily trips is also a deterrent to using

WampR

Initially assessing the individuals who stated that their car usage is due to poor public transit service

(Poor_PT) led to an unexpected result in favor of considering WampR By introducing to this variable

the number of household cars as a proxy for household income (Poor_PTCar1) the model shows

that of the previously mentioned individuals those who have lower income are the ones who have

to consider WampR The result is understandable as these individuals may have no alternative when

they have to change their mode (they also are not likely to consider other modes) Individuals with

higher levels of income who have to use their car during before or after work (Dependencycar1+)

are not likely to use WampR

The greater the number of motorcycles in a household the less likely commuters is to consider

WampR There appears to be a competition between motorcycle and PT for access to the city center

17

Better PT services in the center of the city in terms of coverage and frequency increases the

likelihood that its residents will consider WampR This is verified by the positive sign of the

D_home_place variable Commuters with greater job experience (Job_duration) in their workplace

are more likely to use this mode Although individuals with higher levels of education are not likely

to use WampR as education level increases avoidance of WampR decreases

Taxi (T)

Table 5 shows that none of the studied policies are significant in considering taxi usage It seems that

taxi usage considering its function in Iran as a non-private and non-public mode of transport is not

affected by pull or push policies A negative sign for taxi travel time indicates that individuals are not

likely to use this mode for longer trips This seems reasonable given that longer trips are more

expensive Commuters who are more likely to use fuel with no subsidy are not likely to use taxis As

mentioned before they prefer to use their car A higher number of trips in a day are also a deterrent

to considering taxi usage which may be due to increased cost for more trips Results show that an

individual with more daily trips avoids using taxis Commuters who are employed in more than one

workplace (Pattern 3) are not likely to use taxis This may be due to the fact that they have a lower

level of income which forces them to dedicate more time on the job

Initial results showed that individuals who stated that their car usage is due to poor public transit

service (Poor_PT) are not likely to use taxis This result was far from our expectations By introducing

to this variable the number of household cars as a proxy for household income the model shows

that the previously mentioned individuals who have higher income (Poor_PTCar1+) are the ones

who are not likely to consider taxis Furthermore because such individuals are not considering any

other modes they may treat taxi usage as a kind of PT mode with poor service

As expected greater access to cars in a household (Car_acc) lessens the likelihood of considering

taxis as an alternative Furthermore individuals in households with more motorcycle ownership are

less likely to consider taxis It seems like there is a competition among motorcycles and taxis for

18

access to the city center Younger commuters are less likely to use taxis and individuals with at least

master degrees do consider this mode in addition to their car

Public transit accessed by Drive (DampR)

This mode is affected by the simultaneous interaction of transit time and transit access

(PT_TimeampAccess) which is reflected in the fact that individuals prefer to use this mode for longer

trips Comparing this mode and WampR the first trip start time affects the consideration of this mode

differently Later morning commuters prefer to use their car to access PT modes Such commuters

may have higher income levels or managerial jobs Obviously individuals who are not able to access

PT stations by walking (PTnwacc) are likely to use DampR Serving passengers in daily trips is also a

deterrent in considering this mode which is similar to WampR but with a lower coefficient

Commuters with higher income levels who depend on their car during before or after work

(Dependencycar1+) are likely to use DampR Individuals who use their own car are less likely to use

this mode which is unexpected As a city center develops better PT network coverage and residents

have smaller distances to their workplaces they are unlikely to use DampR This is proven in the model

by a negative sign for D_home_place

Motorcycle (MC)

Increasing fuel cost and cordon pricing simultaneously discourage motorcycle usages Although fuel

cost is expected to reduce motorcycle usage to some extent its combined effect with cordon pricing

also reduces motorcycle usage However this variable is not as strong as other policy variables

=10)

Of the studied modes motorcycle usage is affected by the most commuting variables This may be

due to the fact that this mode is not common Commuting distance has a negative effect on

motorcycle usage which is expected It is worth noting that trip distance appears only in this mode

which may be a reflection of the role of distance in regards to the safety risk in considering this

19

mode Commuters with more stops to serve passengers while commuting (Pattern 2) are not likely

to use this mode which may be due to the poor passenger service of this mode

Individuals who state that commuting is independent of the mode (Pnocarwk) are not likely to use

MC By looking at the (First_Nacco) negative sign this could stem from the fact that the more

passengers there are on the first trip the less likely individuals are to consider MC Regarding the

low capacity of MC and its safety concerns such commuters avoid using this mode Commuters who

pay more parking charges (Park_payment) are less likely to use MC which is expected Individuals

who are dependent on their car during before or after their work time are not likely to use MC

even if they have lower levels of income (DependencyCar1) Individuals who use their own car

(D_car_own) are less likely to use this mode As expected individuals who live in households with

more motorcycle ownership are more likely to use this mode The positive sign of (Permission)

indicates that commuters who have permission to enter the study area do consider MC Because

such commuters generally provide that permission just for car usage this result is unexpected

As with commute variables of all the studied modes MC is affected by the greatest number of

socio-economic variables As expected young commuters (Agelt30) are more likely to use this mode

Commuters with Bachelor of Science degree are less likely to use this mode among others Full time

employees (Emp_full) are less likely to consider MC whereas commuters with more experience in

their jobs prefer to use it Results show that individuals who live in a household with children

younger than 18 are more likely to consider using a car

Tel-Taxi (T_T)

Results show that cordon pricing causes higher probability of using T_T In fact individuals who use

T_T as a mode with similar level of service as cars9 are more willing to pay the cost and make use of

the mode It is worth noting that the effect of cordon pricing in pushing commuters from car usage

9 As this mode does not have driving stress and parking search time in some cases it may have more amount of utility thana car does

20

(000045) is greater than its effect on pulling them to Tel-taxi (000019) This is because of the

possibility of considering other non-car modes

Because consideration of this mode is a function of its operation travel time (Trip_time) appears as

a deterrent in this mode utility function Table 5 shows that individuals are more sensitive to the trip

time when using T_T mode versus taxi which is expected due to their relative costs

The greater the number of full time employees in a family (Nhempfull) the higher the probability of

considering T_T by its commuters which may be due to the higher income level of these

households This is verified by the greater likelihood of using T_T rather than taxis by such

commuters Individuals with higher levels of income who depend on their car during before or after

work time are less likely to use T_T Commuters with lower income levels who state that they use

their car for the sake of comfort (Comfortcar1) are less likely to use T_T which may be due to its

cost Although such individuals do not consider any other modes they specifically avoid T_T Greater

access to cars in a household leads to greater likelihood of T_T usage which could be due to the

higher income level of a household As mentioned before such individuals even avoid taxis

Females who drive to their workplace are more likely to use T_T It seems like this part of society

considers this mode when desiring to avoid the difficulties of driving Younger commuters are less

likely to use T_T and individuals between 30 and 39 years of age are specifically avoiding this mode

Results show that university graduated commuters are more likely to use this mode

6 Marginal effects

To explore the effects of each policy on mode choice and to answer the second issue raised at the

beginning of this paper the marginal effects approach can be adopted Although the coefficients of

the models utility functions show the drivers behavior when facing one or more policies the

marginal effects of policies or their interactions may appropriately show the results of their

implementation More specifically the marginal effect for this study is interpreted as the change in

21

probability given a unit change in a variable ceteris paribus In this section the variable is defined as

a specific policy or a specific interaction of two policies Table 6 presents the marginal effects of the

studied policies and their interactions with mode choice The results are shown in the form of trip

percentages transferred away from the car to the studied modes and the probability-weighted

sample enumeration approach is adopted to find the values It is worth noting that this table is fully

compatible with Table 5 but the marginal effects that were less significant than 90 percent have

been removed

Table 6 - Marginal effects of policies (percent)

Tel-Taxi(TT)

Motorcycle(MC)

Drive ampRide(DampR)

Taxi (T)Walk ampRide(WampR)

Car (C)Mode

Variable-000088Cordon-000140Parking

-09069Access-0000001ParkampFuel

00040PT_TimeampAccess

Table 5 shows that a 1 Rial increase in cordon pricing causes a 000088 percent decrease in car

usage Furthermore a 1 Rial per hour increase in parking cost decreases the car usage probability to

00014 percent By assuming 8 hours for the average parking duration the daily marginal value of

parking cost converts to 000018 percent These values show that cordon pricing is more effective in

forcing individuals not to use their car than increasing parking cost with the same value Results also

show that a 1 Rial increase in parking cost and fuel cost simultaneously decreases the probability of

choosing car usage by 0000001 percent Table 5 also shows that a 1 minute decrease in transit

access time would result in a 09 percent increase in probability of choosing this mode It also shows

that each 1 minute decrease in transit travel time and transit access leads to a 0004 decrease in the

probability of choosing the DampR mode

22

The above results show that just one of the policy interactions is not significantly different from

none (ie CordonampFuel) and push policies have a significant role in impelling car users to consider

other modes

7 Planning

To demonstrate the application of the model some of its planning aspects for the city of Tehran are

presented here In fact the model is able to assess the effect of all studied variables which were

grouped into three tiers on commuters mode choice In this section we assess the effect of TDM

policies by adopting their pair-wise combinations for the city Table 6 shows that pull policies in this

study did not impose a major effect on car usage Because the average access time of the sample

was 11 minutes improving it by 50 percent (about 5 minutes) increases the probability of WampR to

509069100 which is equal to a 048 percent increase in considering this mode (and decrease in all

other modes) On the other hand decreasing transit time about 50 percent (about 19 minutes based

on sample average) and similarly improving access time decreases the probability of considering

DampR to 0004195100 which is equal to a 038 percent decrease in considering this mode (and

increase in all other modes) Therefore in this section we focused on the push policies which

appear in the utility function of car mode It is worth noting that in assessing each combination of

policies all other variables were assumed to be fixed as the current state

71 Parking pricing and cordon pricing

The average hourly parking charge of individuals in the sample is 71 Rials This charge is 3000 Rials

per hour in the curb lane on some streets about 2000 Rials per hour in parking lots and free in

alleys and other streets The average amount shows that most of the commuters benefited from

free parking spaces in alleys and non-controlled streets Figure 2 depicts the result of implementing

this policy and the cordon pricing policy simultaneously In line with marginal effects it can be seen

that in implementing each policy separately cordon pricing is more effective than parking pricing in

the planning range It is worth noting that commuters in this study are less sensitive to parking

23

pricing than cordon pricing One reason may be that they thought in any case they would find free

parking space Figure 2 shows that the effect of each policy depends on the level of the other policy

In other words the effect of the simultaneous implementation of two policies is not equal to the

sum of their separate individual effects

Figure 2- Effect of implementing parking pricing and cordon pricing

72 Parking pricing and increasing fuel cost

Figure 3 shows the simultaneous effects of parking pricing and increasing fuel cost policies Although

both of these policies are not effective separately their simultaneous effect is significant It can be

seen that their synergy appears in the higher levels of both policies

Figure 3 - Effect of implementing parking pricing and increasing fuel cost

24

73 Cordon pricing and increasing fuel cost

The effect of implementing cordon pricing and increasing fuel cost is shown in Figure 4 It can be

seen that these two policies have no interaction effect in decreasing car usage This result is verified

by this interactions insignificant marginal value in Table 6

Figure 4- Effect of implementing cordon pricing and increasing fuel cost

8 Conclusions

This study examined the role of TDM policies in individual mode choice for work trips in the city of

Tehran Five policies including increasing parking cost increasing fuel cost cordon pricing in the odd-

even zone of the city transit time reduction and transit access improvement were investigated Of

the five policies the former three were push policies and the latter two were pull policies The

design of experiments approach was used to design the questionnaire that would capture the stated

preferences of car commuters

A number of nested structures based on recognizing differences in the variances associated

with unobserved influences were examined and finally the MNL as a superior model for this study

was developed

25

The results show that for car users who faced TDM policies all of the five studied TDM

policies were factors in the mode choice process The model shows that in considering car usage

push policies play a main role and pull policies do not

Results also show that the interactions of policies are also significant The model shows that

the interaction of parking cost policy and fuel cost policy is significant in prompting car users to

consider other modes It also shows that the interaction of cordon pricing and increase in fuel cost is

effective to discourage motorcycle usage Furthermore the interaction of transit time reduction

policy and access improvements policy is significant in considering DampR as a transportation mode to

workplace

Pull policies in the study were expected to attract individuals to transit modes (pull policies

of this study are related to transit modes) This was verified by the model

Results show that although taxi usage is usually treated as an alternative for urban trips it

was not directly affected by the studied policies This may be due to its special function in Tehran as

a non-public and non-private mode

The results also show that except for the interaction of cordon pricing and increase in fuel

cost other interaction effects have significant marginal effects on mode choice

Results of the model show that to implement a single policy cordon pricing is the most

effective in decreasing car usage

This study assessed workplace commuters Thus future studies should explore the

individuals mode choice with other trip aims Additionally including more policies especially pull

policies in favor of non-transit modes is reasonable to rigorously assess the model

It can be seen that the studied modes are affected differently by the TDM policies and their

interactions The impact of the interaction of policies appeared in the consideration of three out of

six studied modes and also in two out of three developed graphs This issue needs more attention

within the transportation demand management context and further studies should be performed

26

Design of this study is based on the consideration of all two-way interactions with some

degrees of deficiency in D-efficient policy Focusing on the significant interactions of this study is

recommended to perform other studies with more D-efficient values

Further understanding of the impacts of TDM policies might be provided in future research

by focusing on the amount of their synergy in prohibiting car usage

9 References

Arentze TA Hofman F amp Timmermans HJP (2004) Predicting multi-faceted activityndashtravel adjustmentstrategies in response to possible congestion pricing scenarios using an Internet-based stated adaptationexperiment Transport Policy 11 p31ndash41

Burris M amp Patil S (2009) Estimating the Benefits of Managed Lanes Technical Report Texas TexasTransportation Institute University Transportation Center for Mobility

Cao X amp Mokhtarian PL (2005) How do individuals adapt their personal travel A conceptual exploration ofthe consideation of travel-related strategies Transport Policy 12(3) pp199-206

de Palma A amp Lindsey R (2001) Transportation Supply and congestion In International Encyclopedia of theSocial and Behavioral Sciences 1st ed Elsevier p15882ndash15888

Eriksson L Garvill J amp Nordlund AM (2008) Acceptability of single and combined transport policy measuresThe importance of environmental and policy specific beliefs Transportation Research Part A 42 p1117ndash1128

Eriksson L Nordlund AM amp Garvill J (2010) Expected car use reduction in response to structural traveldemand management measures Transportation Research Part F 13 p329ndash342

Golob TF amp Hensher DA (2007) The Trip chaining Activity of Sydney Residents A Cross-Section Assessmentby Age Group with a Focus on Seniors Journal of Transport Geography 15 p298ndash312

Hensher DA amp King J (2001) Parking demand and responsiveness to supply pricing and location in theSydney central business district Transportation Research A 35 pp177-96

Hensher DA amp Rose JM (2007) Development of commuter and non-commuter mode choice models for theassessment of new public transport infrastructure projects A case study Transportation Research Part A41 p428ndash443

Hensher DA Rose JM amp Greene WH (2005) Applied Choice Analysis A Primer New York CambridgeUniversity Press

Iranian Center of Statistics (ICS) (2009) Information of Iranian States [Online] Available at HYPERLINK httpwwwamarorgirUploadModulesContentsasset16tehrantehparthtmlhttpwwwamarorgirUploadModulesContentsasset16tehrantehparthtml [Accessed 13 November2009]

Kingham S Dickinson J amp Copsey S (2001) Travelling to work Will people move out of their cars TransportPolicy 8 pp151-60

Kuhfeld WF (2009) Marketing Research Methods in SAS Experimental Design Choice Conjoint andGraphical Techniques SAS 92 Edition NC USA SAS Institute Inc Cary Available on the Webwwwsupportsascomtechsuptnotetnote_stathtmlmarket

Litman T (2003) The online TDM Encyclopedia mobility management information gateway Transport Policy10 pp245-49

Litman T (2010) Online TDM Encyclopedia [Online] Available at HYPERLINK httpwwwvtpiorghttpwwwvtpiorg [Accessed 21 Augest 2010]

15

Table 5 ndash The mode choice model

Tel-Taxi(T_T)

Motorcycle(MC)

Drive amp Ride(DampR)

Taxi (T)Walk amp Ride(WampR)

Car (C)Mode

Variable-471756-37067-147911Constant

Transportation demand management measure variables00019-00045Cordon

-000072Parking-004308Access

-28443D-05Parkampfuel-32475D-06Cordonampfuel

00029Pt_timeampaccess

Commuting trip characteristics-04709Trip distance

-02163-00831Trip time-96755163655Exp fuel-16253Ntrips

-114779Pattern2-71008Pattern3

00282-00270First trip time-02439-01549Pnocarwk

-11322992883-32765PTnwacc-133701First Nacco

-7778-73782Accompany-00049000010Park_payment

201646195554Nhempfull-160144ComfortCar1

-206142DependencyCar1-16101883385-121224DependencyCar2

42176Poor_PTCar1-24988Poor_PTCar2

- -27221D car own70960-39136Car acc

1 -71112-156123Nmotorcycle-1436322762D home place

2 78826Permission

HH socio-economic characteristics149490Female

297584-24548Agelt30-136490Age30_39

079430366303585Job_duration-108743Emp_full-203468-64900Edu BS

10932856687-4499984445Edu BS+102271D childlt=18

-2677366L( )-3849556L(0)0305sup2

112127178592580607N

Note = Positive significance at 1 5 10 level

As expected individuals with higher income are more likely to use their car This is indicated in the

model by the positive signs of individuals who use fuel with fixed (unsubsidized) cost and individuals

16

who pay more in parking charges in the previous week of study Negative sign of Pnocarwk variable

shows that the commuters who stated that their commute depends on car availability are more

likely to use their car Individuals in households with more full-time employees are more likely to use

their car which may be the result of higher household income Not surprisingly commuters who

have permission are more likely to maintain car usage Among the household socio-economic

parameters greater job experience (Job_duration) and higher graduate levels (EduBS+) increase the

probability of car usage

Public transit accessed by walking (WampR)

Access time to transit negatively impacts WampR choice which is expected This result is similar to

findings for the city of Sydney (Hensher amp Rose 2007) The negative coefficient of first trip time

indicates that individuals are more likely to use WampR in the early morning This result seems to

reflect the better weather for walking and faster speed of WampR mode early in the morning

Obviously individuals who are not able to access transit stations via walking (PTnwacc) are less likely

to consider this mode Furthermore serving passengers on daily trips is also a deterrent to using

WampR

Initially assessing the individuals who stated that their car usage is due to poor public transit service

(Poor_PT) led to an unexpected result in favor of considering WampR By introducing to this variable

the number of household cars as a proxy for household income (Poor_PTCar1) the model shows

that of the previously mentioned individuals those who have lower income are the ones who have

to consider WampR The result is understandable as these individuals may have no alternative when

they have to change their mode (they also are not likely to consider other modes) Individuals with

higher levels of income who have to use their car during before or after work (Dependencycar1+)

are not likely to use WampR

The greater the number of motorcycles in a household the less likely commuters is to consider

WampR There appears to be a competition between motorcycle and PT for access to the city center

17

Better PT services in the center of the city in terms of coverage and frequency increases the

likelihood that its residents will consider WampR This is verified by the positive sign of the

D_home_place variable Commuters with greater job experience (Job_duration) in their workplace

are more likely to use this mode Although individuals with higher levels of education are not likely

to use WampR as education level increases avoidance of WampR decreases

Taxi (T)

Table 5 shows that none of the studied policies are significant in considering taxi usage It seems that

taxi usage considering its function in Iran as a non-private and non-public mode of transport is not

affected by pull or push policies A negative sign for taxi travel time indicates that individuals are not

likely to use this mode for longer trips This seems reasonable given that longer trips are more

expensive Commuters who are more likely to use fuel with no subsidy are not likely to use taxis As

mentioned before they prefer to use their car A higher number of trips in a day are also a deterrent

to considering taxi usage which may be due to increased cost for more trips Results show that an

individual with more daily trips avoids using taxis Commuters who are employed in more than one

workplace (Pattern 3) are not likely to use taxis This may be due to the fact that they have a lower

level of income which forces them to dedicate more time on the job

Initial results showed that individuals who stated that their car usage is due to poor public transit

service (Poor_PT) are not likely to use taxis This result was far from our expectations By introducing

to this variable the number of household cars as a proxy for household income the model shows

that the previously mentioned individuals who have higher income (Poor_PTCar1+) are the ones

who are not likely to consider taxis Furthermore because such individuals are not considering any

other modes they may treat taxi usage as a kind of PT mode with poor service

As expected greater access to cars in a household (Car_acc) lessens the likelihood of considering

taxis as an alternative Furthermore individuals in households with more motorcycle ownership are

less likely to consider taxis It seems like there is a competition among motorcycles and taxis for

18

access to the city center Younger commuters are less likely to use taxis and individuals with at least

master degrees do consider this mode in addition to their car

Public transit accessed by Drive (DampR)

This mode is affected by the simultaneous interaction of transit time and transit access

(PT_TimeampAccess) which is reflected in the fact that individuals prefer to use this mode for longer

trips Comparing this mode and WampR the first trip start time affects the consideration of this mode

differently Later morning commuters prefer to use their car to access PT modes Such commuters

may have higher income levels or managerial jobs Obviously individuals who are not able to access

PT stations by walking (PTnwacc) are likely to use DampR Serving passengers in daily trips is also a

deterrent in considering this mode which is similar to WampR but with a lower coefficient

Commuters with higher income levels who depend on their car during before or after work

(Dependencycar1+) are likely to use DampR Individuals who use their own car are less likely to use

this mode which is unexpected As a city center develops better PT network coverage and residents

have smaller distances to their workplaces they are unlikely to use DampR This is proven in the model

by a negative sign for D_home_place

Motorcycle (MC)

Increasing fuel cost and cordon pricing simultaneously discourage motorcycle usages Although fuel

cost is expected to reduce motorcycle usage to some extent its combined effect with cordon pricing

also reduces motorcycle usage However this variable is not as strong as other policy variables

=10)

Of the studied modes motorcycle usage is affected by the most commuting variables This may be

due to the fact that this mode is not common Commuting distance has a negative effect on

motorcycle usage which is expected It is worth noting that trip distance appears only in this mode

which may be a reflection of the role of distance in regards to the safety risk in considering this

19

mode Commuters with more stops to serve passengers while commuting (Pattern 2) are not likely

to use this mode which may be due to the poor passenger service of this mode

Individuals who state that commuting is independent of the mode (Pnocarwk) are not likely to use

MC By looking at the (First_Nacco) negative sign this could stem from the fact that the more

passengers there are on the first trip the less likely individuals are to consider MC Regarding the

low capacity of MC and its safety concerns such commuters avoid using this mode Commuters who

pay more parking charges (Park_payment) are less likely to use MC which is expected Individuals

who are dependent on their car during before or after their work time are not likely to use MC

even if they have lower levels of income (DependencyCar1) Individuals who use their own car

(D_car_own) are less likely to use this mode As expected individuals who live in households with

more motorcycle ownership are more likely to use this mode The positive sign of (Permission)

indicates that commuters who have permission to enter the study area do consider MC Because

such commuters generally provide that permission just for car usage this result is unexpected

As with commute variables of all the studied modes MC is affected by the greatest number of

socio-economic variables As expected young commuters (Agelt30) are more likely to use this mode

Commuters with Bachelor of Science degree are less likely to use this mode among others Full time

employees (Emp_full) are less likely to consider MC whereas commuters with more experience in

their jobs prefer to use it Results show that individuals who live in a household with children

younger than 18 are more likely to consider using a car

Tel-Taxi (T_T)

Results show that cordon pricing causes higher probability of using T_T In fact individuals who use

T_T as a mode with similar level of service as cars9 are more willing to pay the cost and make use of

the mode It is worth noting that the effect of cordon pricing in pushing commuters from car usage

9 As this mode does not have driving stress and parking search time in some cases it may have more amount of utility thana car does

20

(000045) is greater than its effect on pulling them to Tel-taxi (000019) This is because of the

possibility of considering other non-car modes

Because consideration of this mode is a function of its operation travel time (Trip_time) appears as

a deterrent in this mode utility function Table 5 shows that individuals are more sensitive to the trip

time when using T_T mode versus taxi which is expected due to their relative costs

The greater the number of full time employees in a family (Nhempfull) the higher the probability of

considering T_T by its commuters which may be due to the higher income level of these

households This is verified by the greater likelihood of using T_T rather than taxis by such

commuters Individuals with higher levels of income who depend on their car during before or after

work time are less likely to use T_T Commuters with lower income levels who state that they use

their car for the sake of comfort (Comfortcar1) are less likely to use T_T which may be due to its

cost Although such individuals do not consider any other modes they specifically avoid T_T Greater

access to cars in a household leads to greater likelihood of T_T usage which could be due to the

higher income level of a household As mentioned before such individuals even avoid taxis

Females who drive to their workplace are more likely to use T_T It seems like this part of society

considers this mode when desiring to avoid the difficulties of driving Younger commuters are less

likely to use T_T and individuals between 30 and 39 years of age are specifically avoiding this mode

Results show that university graduated commuters are more likely to use this mode

6 Marginal effects

To explore the effects of each policy on mode choice and to answer the second issue raised at the

beginning of this paper the marginal effects approach can be adopted Although the coefficients of

the models utility functions show the drivers behavior when facing one or more policies the

marginal effects of policies or their interactions may appropriately show the results of their

implementation More specifically the marginal effect for this study is interpreted as the change in

21

probability given a unit change in a variable ceteris paribus In this section the variable is defined as

a specific policy or a specific interaction of two policies Table 6 presents the marginal effects of the

studied policies and their interactions with mode choice The results are shown in the form of trip

percentages transferred away from the car to the studied modes and the probability-weighted

sample enumeration approach is adopted to find the values It is worth noting that this table is fully

compatible with Table 5 but the marginal effects that were less significant than 90 percent have

been removed

Table 6 - Marginal effects of policies (percent)

Tel-Taxi(TT)

Motorcycle(MC)

Drive ampRide(DampR)

Taxi (T)Walk ampRide(WampR)

Car (C)Mode

Variable-000088Cordon-000140Parking

-09069Access-0000001ParkampFuel

00040PT_TimeampAccess

Table 5 shows that a 1 Rial increase in cordon pricing causes a 000088 percent decrease in car

usage Furthermore a 1 Rial per hour increase in parking cost decreases the car usage probability to

00014 percent By assuming 8 hours for the average parking duration the daily marginal value of

parking cost converts to 000018 percent These values show that cordon pricing is more effective in

forcing individuals not to use their car than increasing parking cost with the same value Results also

show that a 1 Rial increase in parking cost and fuel cost simultaneously decreases the probability of

choosing car usage by 0000001 percent Table 5 also shows that a 1 minute decrease in transit

access time would result in a 09 percent increase in probability of choosing this mode It also shows

that each 1 minute decrease in transit travel time and transit access leads to a 0004 decrease in the

probability of choosing the DampR mode

22

The above results show that just one of the policy interactions is not significantly different from

none (ie CordonampFuel) and push policies have a significant role in impelling car users to consider

other modes

7 Planning

To demonstrate the application of the model some of its planning aspects for the city of Tehran are

presented here In fact the model is able to assess the effect of all studied variables which were

grouped into three tiers on commuters mode choice In this section we assess the effect of TDM

policies by adopting their pair-wise combinations for the city Table 6 shows that pull policies in this

study did not impose a major effect on car usage Because the average access time of the sample

was 11 minutes improving it by 50 percent (about 5 minutes) increases the probability of WampR to

509069100 which is equal to a 048 percent increase in considering this mode (and decrease in all

other modes) On the other hand decreasing transit time about 50 percent (about 19 minutes based

on sample average) and similarly improving access time decreases the probability of considering

DampR to 0004195100 which is equal to a 038 percent decrease in considering this mode (and

increase in all other modes) Therefore in this section we focused on the push policies which

appear in the utility function of car mode It is worth noting that in assessing each combination of

policies all other variables were assumed to be fixed as the current state

71 Parking pricing and cordon pricing

The average hourly parking charge of individuals in the sample is 71 Rials This charge is 3000 Rials

per hour in the curb lane on some streets about 2000 Rials per hour in parking lots and free in

alleys and other streets The average amount shows that most of the commuters benefited from

free parking spaces in alleys and non-controlled streets Figure 2 depicts the result of implementing

this policy and the cordon pricing policy simultaneously In line with marginal effects it can be seen

that in implementing each policy separately cordon pricing is more effective than parking pricing in

the planning range It is worth noting that commuters in this study are less sensitive to parking

23

pricing than cordon pricing One reason may be that they thought in any case they would find free

parking space Figure 2 shows that the effect of each policy depends on the level of the other policy

In other words the effect of the simultaneous implementation of two policies is not equal to the

sum of their separate individual effects

Figure 2- Effect of implementing parking pricing and cordon pricing

72 Parking pricing and increasing fuel cost

Figure 3 shows the simultaneous effects of parking pricing and increasing fuel cost policies Although

both of these policies are not effective separately their simultaneous effect is significant It can be

seen that their synergy appears in the higher levels of both policies

Figure 3 - Effect of implementing parking pricing and increasing fuel cost

24

73 Cordon pricing and increasing fuel cost

The effect of implementing cordon pricing and increasing fuel cost is shown in Figure 4 It can be

seen that these two policies have no interaction effect in decreasing car usage This result is verified

by this interactions insignificant marginal value in Table 6

Figure 4- Effect of implementing cordon pricing and increasing fuel cost

8 Conclusions

This study examined the role of TDM policies in individual mode choice for work trips in the city of

Tehran Five policies including increasing parking cost increasing fuel cost cordon pricing in the odd-

even zone of the city transit time reduction and transit access improvement were investigated Of

the five policies the former three were push policies and the latter two were pull policies The

design of experiments approach was used to design the questionnaire that would capture the stated

preferences of car commuters

A number of nested structures based on recognizing differences in the variances associated

with unobserved influences were examined and finally the MNL as a superior model for this study

was developed

25

The results show that for car users who faced TDM policies all of the five studied TDM

policies were factors in the mode choice process The model shows that in considering car usage

push policies play a main role and pull policies do not

Results also show that the interactions of policies are also significant The model shows that

the interaction of parking cost policy and fuel cost policy is significant in prompting car users to

consider other modes It also shows that the interaction of cordon pricing and increase in fuel cost is

effective to discourage motorcycle usage Furthermore the interaction of transit time reduction

policy and access improvements policy is significant in considering DampR as a transportation mode to

workplace

Pull policies in the study were expected to attract individuals to transit modes (pull policies

of this study are related to transit modes) This was verified by the model

Results show that although taxi usage is usually treated as an alternative for urban trips it

was not directly affected by the studied policies This may be due to its special function in Tehran as

a non-public and non-private mode

The results also show that except for the interaction of cordon pricing and increase in fuel

cost other interaction effects have significant marginal effects on mode choice

Results of the model show that to implement a single policy cordon pricing is the most

effective in decreasing car usage

This study assessed workplace commuters Thus future studies should explore the

individuals mode choice with other trip aims Additionally including more policies especially pull

policies in favor of non-transit modes is reasonable to rigorously assess the model

It can be seen that the studied modes are affected differently by the TDM policies and their

interactions The impact of the interaction of policies appeared in the consideration of three out of

six studied modes and also in two out of three developed graphs This issue needs more attention

within the transportation demand management context and further studies should be performed

26

Design of this study is based on the consideration of all two-way interactions with some

degrees of deficiency in D-efficient policy Focusing on the significant interactions of this study is

recommended to perform other studies with more D-efficient values

Further understanding of the impacts of TDM policies might be provided in future research

by focusing on the amount of their synergy in prohibiting car usage

9 References

Arentze TA Hofman F amp Timmermans HJP (2004) Predicting multi-faceted activityndashtravel adjustmentstrategies in response to possible congestion pricing scenarios using an Internet-based stated adaptationexperiment Transport Policy 11 p31ndash41

Burris M amp Patil S (2009) Estimating the Benefits of Managed Lanes Technical Report Texas TexasTransportation Institute University Transportation Center for Mobility

Cao X amp Mokhtarian PL (2005) How do individuals adapt their personal travel A conceptual exploration ofthe consideation of travel-related strategies Transport Policy 12(3) pp199-206

de Palma A amp Lindsey R (2001) Transportation Supply and congestion In International Encyclopedia of theSocial and Behavioral Sciences 1st ed Elsevier p15882ndash15888

Eriksson L Garvill J amp Nordlund AM (2008) Acceptability of single and combined transport policy measuresThe importance of environmental and policy specific beliefs Transportation Research Part A 42 p1117ndash1128

Eriksson L Nordlund AM amp Garvill J (2010) Expected car use reduction in response to structural traveldemand management measures Transportation Research Part F 13 p329ndash342

Golob TF amp Hensher DA (2007) The Trip chaining Activity of Sydney Residents A Cross-Section Assessmentby Age Group with a Focus on Seniors Journal of Transport Geography 15 p298ndash312

Hensher DA amp King J (2001) Parking demand and responsiveness to supply pricing and location in theSydney central business district Transportation Research A 35 pp177-96

Hensher DA amp Rose JM (2007) Development of commuter and non-commuter mode choice models for theassessment of new public transport infrastructure projects A case study Transportation Research Part A41 p428ndash443

Hensher DA Rose JM amp Greene WH (2005) Applied Choice Analysis A Primer New York CambridgeUniversity Press

Iranian Center of Statistics (ICS) (2009) Information of Iranian States [Online] Available at HYPERLINK httpwwwamarorgirUploadModulesContentsasset16tehrantehparthtmlhttpwwwamarorgirUploadModulesContentsasset16tehrantehparthtml [Accessed 13 November2009]

Kingham S Dickinson J amp Copsey S (2001) Travelling to work Will people move out of their cars TransportPolicy 8 pp151-60

Kuhfeld WF (2009) Marketing Research Methods in SAS Experimental Design Choice Conjoint andGraphical Techniques SAS 92 Edition NC USA SAS Institute Inc Cary Available on the Webwwwsupportsascomtechsuptnotetnote_stathtmlmarket

Litman T (2003) The online TDM Encyclopedia mobility management information gateway Transport Policy10 pp245-49

Litman T (2010) Online TDM Encyclopedia [Online] Available at HYPERLINK httpwwwvtpiorghttpwwwvtpiorg [Accessed 21 Augest 2010]

16

who pay more in parking charges in the previous week of study Negative sign of Pnocarwk variable

shows that the commuters who stated that their commute depends on car availability are more

likely to use their car Individuals in households with more full-time employees are more likely to use

their car which may be the result of higher household income Not surprisingly commuters who

have permission are more likely to maintain car usage Among the household socio-economic

parameters greater job experience (Job_duration) and higher graduate levels (EduBS+) increase the

probability of car usage

Public transit accessed by walking (WampR)

Access time to transit negatively impacts WampR choice which is expected This result is similar to

findings for the city of Sydney (Hensher amp Rose 2007) The negative coefficient of first trip time

indicates that individuals are more likely to use WampR in the early morning This result seems to

reflect the better weather for walking and faster speed of WampR mode early in the morning

Obviously individuals who are not able to access transit stations via walking (PTnwacc) are less likely

to consider this mode Furthermore serving passengers on daily trips is also a deterrent to using

WampR

Initially assessing the individuals who stated that their car usage is due to poor public transit service

(Poor_PT) led to an unexpected result in favor of considering WampR By introducing to this variable

the number of household cars as a proxy for household income (Poor_PTCar1) the model shows

that of the previously mentioned individuals those who have lower income are the ones who have

to consider WampR The result is understandable as these individuals may have no alternative when

they have to change their mode (they also are not likely to consider other modes) Individuals with

higher levels of income who have to use their car during before or after work (Dependencycar1+)

are not likely to use WampR

The greater the number of motorcycles in a household the less likely commuters is to consider

WampR There appears to be a competition between motorcycle and PT for access to the city center

17

Better PT services in the center of the city in terms of coverage and frequency increases the

likelihood that its residents will consider WampR This is verified by the positive sign of the

D_home_place variable Commuters with greater job experience (Job_duration) in their workplace

are more likely to use this mode Although individuals with higher levels of education are not likely

to use WampR as education level increases avoidance of WampR decreases

Taxi (T)

Table 5 shows that none of the studied policies are significant in considering taxi usage It seems that

taxi usage considering its function in Iran as a non-private and non-public mode of transport is not

affected by pull or push policies A negative sign for taxi travel time indicates that individuals are not

likely to use this mode for longer trips This seems reasonable given that longer trips are more

expensive Commuters who are more likely to use fuel with no subsidy are not likely to use taxis As

mentioned before they prefer to use their car A higher number of trips in a day are also a deterrent

to considering taxi usage which may be due to increased cost for more trips Results show that an

individual with more daily trips avoids using taxis Commuters who are employed in more than one

workplace (Pattern 3) are not likely to use taxis This may be due to the fact that they have a lower

level of income which forces them to dedicate more time on the job

Initial results showed that individuals who stated that their car usage is due to poor public transit

service (Poor_PT) are not likely to use taxis This result was far from our expectations By introducing

to this variable the number of household cars as a proxy for household income the model shows

that the previously mentioned individuals who have higher income (Poor_PTCar1+) are the ones

who are not likely to consider taxis Furthermore because such individuals are not considering any

other modes they may treat taxi usage as a kind of PT mode with poor service

As expected greater access to cars in a household (Car_acc) lessens the likelihood of considering

taxis as an alternative Furthermore individuals in households with more motorcycle ownership are

less likely to consider taxis It seems like there is a competition among motorcycles and taxis for

18

access to the city center Younger commuters are less likely to use taxis and individuals with at least

master degrees do consider this mode in addition to their car

Public transit accessed by Drive (DampR)

This mode is affected by the simultaneous interaction of transit time and transit access

(PT_TimeampAccess) which is reflected in the fact that individuals prefer to use this mode for longer

trips Comparing this mode and WampR the first trip start time affects the consideration of this mode

differently Later morning commuters prefer to use their car to access PT modes Such commuters

may have higher income levels or managerial jobs Obviously individuals who are not able to access

PT stations by walking (PTnwacc) are likely to use DampR Serving passengers in daily trips is also a

deterrent in considering this mode which is similar to WampR but with a lower coefficient

Commuters with higher income levels who depend on their car during before or after work

(Dependencycar1+) are likely to use DampR Individuals who use their own car are less likely to use

this mode which is unexpected As a city center develops better PT network coverage and residents

have smaller distances to their workplaces they are unlikely to use DampR This is proven in the model

by a negative sign for D_home_place

Motorcycle (MC)

Increasing fuel cost and cordon pricing simultaneously discourage motorcycle usages Although fuel

cost is expected to reduce motorcycle usage to some extent its combined effect with cordon pricing

also reduces motorcycle usage However this variable is not as strong as other policy variables

=10)

Of the studied modes motorcycle usage is affected by the most commuting variables This may be

due to the fact that this mode is not common Commuting distance has a negative effect on

motorcycle usage which is expected It is worth noting that trip distance appears only in this mode

which may be a reflection of the role of distance in regards to the safety risk in considering this

19

mode Commuters with more stops to serve passengers while commuting (Pattern 2) are not likely

to use this mode which may be due to the poor passenger service of this mode

Individuals who state that commuting is independent of the mode (Pnocarwk) are not likely to use

MC By looking at the (First_Nacco) negative sign this could stem from the fact that the more

passengers there are on the first trip the less likely individuals are to consider MC Regarding the

low capacity of MC and its safety concerns such commuters avoid using this mode Commuters who

pay more parking charges (Park_payment) are less likely to use MC which is expected Individuals

who are dependent on their car during before or after their work time are not likely to use MC

even if they have lower levels of income (DependencyCar1) Individuals who use their own car

(D_car_own) are less likely to use this mode As expected individuals who live in households with

more motorcycle ownership are more likely to use this mode The positive sign of (Permission)

indicates that commuters who have permission to enter the study area do consider MC Because

such commuters generally provide that permission just for car usage this result is unexpected

As with commute variables of all the studied modes MC is affected by the greatest number of

socio-economic variables As expected young commuters (Agelt30) are more likely to use this mode

Commuters with Bachelor of Science degree are less likely to use this mode among others Full time

employees (Emp_full) are less likely to consider MC whereas commuters with more experience in

their jobs prefer to use it Results show that individuals who live in a household with children

younger than 18 are more likely to consider using a car

Tel-Taxi (T_T)

Results show that cordon pricing causes higher probability of using T_T In fact individuals who use

T_T as a mode with similar level of service as cars9 are more willing to pay the cost and make use of

the mode It is worth noting that the effect of cordon pricing in pushing commuters from car usage

9 As this mode does not have driving stress and parking search time in some cases it may have more amount of utility thana car does

20

(000045) is greater than its effect on pulling them to Tel-taxi (000019) This is because of the

possibility of considering other non-car modes

Because consideration of this mode is a function of its operation travel time (Trip_time) appears as

a deterrent in this mode utility function Table 5 shows that individuals are more sensitive to the trip

time when using T_T mode versus taxi which is expected due to their relative costs

The greater the number of full time employees in a family (Nhempfull) the higher the probability of

considering T_T by its commuters which may be due to the higher income level of these

households This is verified by the greater likelihood of using T_T rather than taxis by such

commuters Individuals with higher levels of income who depend on their car during before or after

work time are less likely to use T_T Commuters with lower income levels who state that they use

their car for the sake of comfort (Comfortcar1) are less likely to use T_T which may be due to its

cost Although such individuals do not consider any other modes they specifically avoid T_T Greater

access to cars in a household leads to greater likelihood of T_T usage which could be due to the

higher income level of a household As mentioned before such individuals even avoid taxis

Females who drive to their workplace are more likely to use T_T It seems like this part of society

considers this mode when desiring to avoid the difficulties of driving Younger commuters are less

likely to use T_T and individuals between 30 and 39 years of age are specifically avoiding this mode

Results show that university graduated commuters are more likely to use this mode

6 Marginal effects

To explore the effects of each policy on mode choice and to answer the second issue raised at the

beginning of this paper the marginal effects approach can be adopted Although the coefficients of

the models utility functions show the drivers behavior when facing one or more policies the

marginal effects of policies or their interactions may appropriately show the results of their

implementation More specifically the marginal effect for this study is interpreted as the change in

21

probability given a unit change in a variable ceteris paribus In this section the variable is defined as

a specific policy or a specific interaction of two policies Table 6 presents the marginal effects of the

studied policies and their interactions with mode choice The results are shown in the form of trip

percentages transferred away from the car to the studied modes and the probability-weighted

sample enumeration approach is adopted to find the values It is worth noting that this table is fully

compatible with Table 5 but the marginal effects that were less significant than 90 percent have

been removed

Table 6 - Marginal effects of policies (percent)

Tel-Taxi(TT)

Motorcycle(MC)

Drive ampRide(DampR)

Taxi (T)Walk ampRide(WampR)

Car (C)Mode

Variable-000088Cordon-000140Parking

-09069Access-0000001ParkampFuel

00040PT_TimeampAccess

Table 5 shows that a 1 Rial increase in cordon pricing causes a 000088 percent decrease in car

usage Furthermore a 1 Rial per hour increase in parking cost decreases the car usage probability to

00014 percent By assuming 8 hours for the average parking duration the daily marginal value of

parking cost converts to 000018 percent These values show that cordon pricing is more effective in

forcing individuals not to use their car than increasing parking cost with the same value Results also

show that a 1 Rial increase in parking cost and fuel cost simultaneously decreases the probability of

choosing car usage by 0000001 percent Table 5 also shows that a 1 minute decrease in transit

access time would result in a 09 percent increase in probability of choosing this mode It also shows

that each 1 minute decrease in transit travel time and transit access leads to a 0004 decrease in the

probability of choosing the DampR mode

22

The above results show that just one of the policy interactions is not significantly different from

none (ie CordonampFuel) and push policies have a significant role in impelling car users to consider

other modes

7 Planning

To demonstrate the application of the model some of its planning aspects for the city of Tehran are

presented here In fact the model is able to assess the effect of all studied variables which were

grouped into three tiers on commuters mode choice In this section we assess the effect of TDM

policies by adopting their pair-wise combinations for the city Table 6 shows that pull policies in this

study did not impose a major effect on car usage Because the average access time of the sample

was 11 minutes improving it by 50 percent (about 5 minutes) increases the probability of WampR to

509069100 which is equal to a 048 percent increase in considering this mode (and decrease in all

other modes) On the other hand decreasing transit time about 50 percent (about 19 minutes based

on sample average) and similarly improving access time decreases the probability of considering

DampR to 0004195100 which is equal to a 038 percent decrease in considering this mode (and

increase in all other modes) Therefore in this section we focused on the push policies which

appear in the utility function of car mode It is worth noting that in assessing each combination of

policies all other variables were assumed to be fixed as the current state

71 Parking pricing and cordon pricing

The average hourly parking charge of individuals in the sample is 71 Rials This charge is 3000 Rials

per hour in the curb lane on some streets about 2000 Rials per hour in parking lots and free in

alleys and other streets The average amount shows that most of the commuters benefited from

free parking spaces in alleys and non-controlled streets Figure 2 depicts the result of implementing

this policy and the cordon pricing policy simultaneously In line with marginal effects it can be seen

that in implementing each policy separately cordon pricing is more effective than parking pricing in

the planning range It is worth noting that commuters in this study are less sensitive to parking

23

pricing than cordon pricing One reason may be that they thought in any case they would find free

parking space Figure 2 shows that the effect of each policy depends on the level of the other policy

In other words the effect of the simultaneous implementation of two policies is not equal to the

sum of their separate individual effects

Figure 2- Effect of implementing parking pricing and cordon pricing

72 Parking pricing and increasing fuel cost

Figure 3 shows the simultaneous effects of parking pricing and increasing fuel cost policies Although

both of these policies are not effective separately their simultaneous effect is significant It can be

seen that their synergy appears in the higher levels of both policies

Figure 3 - Effect of implementing parking pricing and increasing fuel cost

24

73 Cordon pricing and increasing fuel cost

The effect of implementing cordon pricing and increasing fuel cost is shown in Figure 4 It can be

seen that these two policies have no interaction effect in decreasing car usage This result is verified

by this interactions insignificant marginal value in Table 6

Figure 4- Effect of implementing cordon pricing and increasing fuel cost

8 Conclusions

This study examined the role of TDM policies in individual mode choice for work trips in the city of

Tehran Five policies including increasing parking cost increasing fuel cost cordon pricing in the odd-

even zone of the city transit time reduction and transit access improvement were investigated Of

the five policies the former three were push policies and the latter two were pull policies The

design of experiments approach was used to design the questionnaire that would capture the stated

preferences of car commuters

A number of nested structures based on recognizing differences in the variances associated

with unobserved influences were examined and finally the MNL as a superior model for this study

was developed

25

The results show that for car users who faced TDM policies all of the five studied TDM

policies were factors in the mode choice process The model shows that in considering car usage

push policies play a main role and pull policies do not

Results also show that the interactions of policies are also significant The model shows that

the interaction of parking cost policy and fuel cost policy is significant in prompting car users to

consider other modes It also shows that the interaction of cordon pricing and increase in fuel cost is

effective to discourage motorcycle usage Furthermore the interaction of transit time reduction

policy and access improvements policy is significant in considering DampR as a transportation mode to

workplace

Pull policies in the study were expected to attract individuals to transit modes (pull policies

of this study are related to transit modes) This was verified by the model

Results show that although taxi usage is usually treated as an alternative for urban trips it

was not directly affected by the studied policies This may be due to its special function in Tehran as

a non-public and non-private mode

The results also show that except for the interaction of cordon pricing and increase in fuel

cost other interaction effects have significant marginal effects on mode choice

Results of the model show that to implement a single policy cordon pricing is the most

effective in decreasing car usage

This study assessed workplace commuters Thus future studies should explore the

individuals mode choice with other trip aims Additionally including more policies especially pull

policies in favor of non-transit modes is reasonable to rigorously assess the model

It can be seen that the studied modes are affected differently by the TDM policies and their

interactions The impact of the interaction of policies appeared in the consideration of three out of

six studied modes and also in two out of three developed graphs This issue needs more attention

within the transportation demand management context and further studies should be performed

26

Design of this study is based on the consideration of all two-way interactions with some

degrees of deficiency in D-efficient policy Focusing on the significant interactions of this study is

recommended to perform other studies with more D-efficient values

Further understanding of the impacts of TDM policies might be provided in future research

by focusing on the amount of their synergy in prohibiting car usage

9 References

Arentze TA Hofman F amp Timmermans HJP (2004) Predicting multi-faceted activityndashtravel adjustmentstrategies in response to possible congestion pricing scenarios using an Internet-based stated adaptationexperiment Transport Policy 11 p31ndash41

Burris M amp Patil S (2009) Estimating the Benefits of Managed Lanes Technical Report Texas TexasTransportation Institute University Transportation Center for Mobility

Cao X amp Mokhtarian PL (2005) How do individuals adapt their personal travel A conceptual exploration ofthe consideation of travel-related strategies Transport Policy 12(3) pp199-206

de Palma A amp Lindsey R (2001) Transportation Supply and congestion In International Encyclopedia of theSocial and Behavioral Sciences 1st ed Elsevier p15882ndash15888

Eriksson L Garvill J amp Nordlund AM (2008) Acceptability of single and combined transport policy measuresThe importance of environmental and policy specific beliefs Transportation Research Part A 42 p1117ndash1128

Eriksson L Nordlund AM amp Garvill J (2010) Expected car use reduction in response to structural traveldemand management measures Transportation Research Part F 13 p329ndash342

Golob TF amp Hensher DA (2007) The Trip chaining Activity of Sydney Residents A Cross-Section Assessmentby Age Group with a Focus on Seniors Journal of Transport Geography 15 p298ndash312

Hensher DA amp King J (2001) Parking demand and responsiveness to supply pricing and location in theSydney central business district Transportation Research A 35 pp177-96

Hensher DA amp Rose JM (2007) Development of commuter and non-commuter mode choice models for theassessment of new public transport infrastructure projects A case study Transportation Research Part A41 p428ndash443

Hensher DA Rose JM amp Greene WH (2005) Applied Choice Analysis A Primer New York CambridgeUniversity Press

Iranian Center of Statistics (ICS) (2009) Information of Iranian States [Online] Available at HYPERLINK httpwwwamarorgirUploadModulesContentsasset16tehrantehparthtmlhttpwwwamarorgirUploadModulesContentsasset16tehrantehparthtml [Accessed 13 November2009]

Kingham S Dickinson J amp Copsey S (2001) Travelling to work Will people move out of their cars TransportPolicy 8 pp151-60

Kuhfeld WF (2009) Marketing Research Methods in SAS Experimental Design Choice Conjoint andGraphical Techniques SAS 92 Edition NC USA SAS Institute Inc Cary Available on the Webwwwsupportsascomtechsuptnotetnote_stathtmlmarket

Litman T (2003) The online TDM Encyclopedia mobility management information gateway Transport Policy10 pp245-49

Litman T (2010) Online TDM Encyclopedia [Online] Available at HYPERLINK httpwwwvtpiorghttpwwwvtpiorg [Accessed 21 Augest 2010]

17

Better PT services in the center of the city in terms of coverage and frequency increases the

likelihood that its residents will consider WampR This is verified by the positive sign of the

D_home_place variable Commuters with greater job experience (Job_duration) in their workplace

are more likely to use this mode Although individuals with higher levels of education are not likely

to use WampR as education level increases avoidance of WampR decreases

Taxi (T)

Table 5 shows that none of the studied policies are significant in considering taxi usage It seems that

taxi usage considering its function in Iran as a non-private and non-public mode of transport is not

affected by pull or push policies A negative sign for taxi travel time indicates that individuals are not

likely to use this mode for longer trips This seems reasonable given that longer trips are more

expensive Commuters who are more likely to use fuel with no subsidy are not likely to use taxis As

mentioned before they prefer to use their car A higher number of trips in a day are also a deterrent

to considering taxi usage which may be due to increased cost for more trips Results show that an

individual with more daily trips avoids using taxis Commuters who are employed in more than one

workplace (Pattern 3) are not likely to use taxis This may be due to the fact that they have a lower

level of income which forces them to dedicate more time on the job

Initial results showed that individuals who stated that their car usage is due to poor public transit

service (Poor_PT) are not likely to use taxis This result was far from our expectations By introducing

to this variable the number of household cars as a proxy for household income the model shows

that the previously mentioned individuals who have higher income (Poor_PTCar1+) are the ones

who are not likely to consider taxis Furthermore because such individuals are not considering any

other modes they may treat taxi usage as a kind of PT mode with poor service

As expected greater access to cars in a household (Car_acc) lessens the likelihood of considering

taxis as an alternative Furthermore individuals in households with more motorcycle ownership are

less likely to consider taxis It seems like there is a competition among motorcycles and taxis for

18

access to the city center Younger commuters are less likely to use taxis and individuals with at least

master degrees do consider this mode in addition to their car

Public transit accessed by Drive (DampR)

This mode is affected by the simultaneous interaction of transit time and transit access

(PT_TimeampAccess) which is reflected in the fact that individuals prefer to use this mode for longer

trips Comparing this mode and WampR the first trip start time affects the consideration of this mode

differently Later morning commuters prefer to use their car to access PT modes Such commuters

may have higher income levels or managerial jobs Obviously individuals who are not able to access

PT stations by walking (PTnwacc) are likely to use DampR Serving passengers in daily trips is also a

deterrent in considering this mode which is similar to WampR but with a lower coefficient

Commuters with higher income levels who depend on their car during before or after work

(Dependencycar1+) are likely to use DampR Individuals who use their own car are less likely to use

this mode which is unexpected As a city center develops better PT network coverage and residents

have smaller distances to their workplaces they are unlikely to use DampR This is proven in the model

by a negative sign for D_home_place

Motorcycle (MC)

Increasing fuel cost and cordon pricing simultaneously discourage motorcycle usages Although fuel

cost is expected to reduce motorcycle usage to some extent its combined effect with cordon pricing

also reduces motorcycle usage However this variable is not as strong as other policy variables

=10)

Of the studied modes motorcycle usage is affected by the most commuting variables This may be

due to the fact that this mode is not common Commuting distance has a negative effect on

motorcycle usage which is expected It is worth noting that trip distance appears only in this mode

which may be a reflection of the role of distance in regards to the safety risk in considering this

19

mode Commuters with more stops to serve passengers while commuting (Pattern 2) are not likely

to use this mode which may be due to the poor passenger service of this mode

Individuals who state that commuting is independent of the mode (Pnocarwk) are not likely to use

MC By looking at the (First_Nacco) negative sign this could stem from the fact that the more

passengers there are on the first trip the less likely individuals are to consider MC Regarding the

low capacity of MC and its safety concerns such commuters avoid using this mode Commuters who

pay more parking charges (Park_payment) are less likely to use MC which is expected Individuals

who are dependent on their car during before or after their work time are not likely to use MC

even if they have lower levels of income (DependencyCar1) Individuals who use their own car

(D_car_own) are less likely to use this mode As expected individuals who live in households with

more motorcycle ownership are more likely to use this mode The positive sign of (Permission)

indicates that commuters who have permission to enter the study area do consider MC Because

such commuters generally provide that permission just for car usage this result is unexpected

As with commute variables of all the studied modes MC is affected by the greatest number of

socio-economic variables As expected young commuters (Agelt30) are more likely to use this mode

Commuters with Bachelor of Science degree are less likely to use this mode among others Full time

employees (Emp_full) are less likely to consider MC whereas commuters with more experience in

their jobs prefer to use it Results show that individuals who live in a household with children

younger than 18 are more likely to consider using a car

Tel-Taxi (T_T)

Results show that cordon pricing causes higher probability of using T_T In fact individuals who use

T_T as a mode with similar level of service as cars9 are more willing to pay the cost and make use of

the mode It is worth noting that the effect of cordon pricing in pushing commuters from car usage

9 As this mode does not have driving stress and parking search time in some cases it may have more amount of utility thana car does

20

(000045) is greater than its effect on pulling them to Tel-taxi (000019) This is because of the

possibility of considering other non-car modes

Because consideration of this mode is a function of its operation travel time (Trip_time) appears as

a deterrent in this mode utility function Table 5 shows that individuals are more sensitive to the trip

time when using T_T mode versus taxi which is expected due to their relative costs

The greater the number of full time employees in a family (Nhempfull) the higher the probability of

considering T_T by its commuters which may be due to the higher income level of these

households This is verified by the greater likelihood of using T_T rather than taxis by such

commuters Individuals with higher levels of income who depend on their car during before or after

work time are less likely to use T_T Commuters with lower income levels who state that they use

their car for the sake of comfort (Comfortcar1) are less likely to use T_T which may be due to its

cost Although such individuals do not consider any other modes they specifically avoid T_T Greater

access to cars in a household leads to greater likelihood of T_T usage which could be due to the

higher income level of a household As mentioned before such individuals even avoid taxis

Females who drive to their workplace are more likely to use T_T It seems like this part of society

considers this mode when desiring to avoid the difficulties of driving Younger commuters are less

likely to use T_T and individuals between 30 and 39 years of age are specifically avoiding this mode

Results show that university graduated commuters are more likely to use this mode

6 Marginal effects

To explore the effects of each policy on mode choice and to answer the second issue raised at the

beginning of this paper the marginal effects approach can be adopted Although the coefficients of

the models utility functions show the drivers behavior when facing one or more policies the

marginal effects of policies or their interactions may appropriately show the results of their

implementation More specifically the marginal effect for this study is interpreted as the change in

21

probability given a unit change in a variable ceteris paribus In this section the variable is defined as

a specific policy or a specific interaction of two policies Table 6 presents the marginal effects of the

studied policies and their interactions with mode choice The results are shown in the form of trip

percentages transferred away from the car to the studied modes and the probability-weighted

sample enumeration approach is adopted to find the values It is worth noting that this table is fully

compatible with Table 5 but the marginal effects that were less significant than 90 percent have

been removed

Table 6 - Marginal effects of policies (percent)

Tel-Taxi(TT)

Motorcycle(MC)

Drive ampRide(DampR)

Taxi (T)Walk ampRide(WampR)

Car (C)Mode

Variable-000088Cordon-000140Parking

-09069Access-0000001ParkampFuel

00040PT_TimeampAccess

Table 5 shows that a 1 Rial increase in cordon pricing causes a 000088 percent decrease in car

usage Furthermore a 1 Rial per hour increase in parking cost decreases the car usage probability to

00014 percent By assuming 8 hours for the average parking duration the daily marginal value of

parking cost converts to 000018 percent These values show that cordon pricing is more effective in

forcing individuals not to use their car than increasing parking cost with the same value Results also

show that a 1 Rial increase in parking cost and fuel cost simultaneously decreases the probability of

choosing car usage by 0000001 percent Table 5 also shows that a 1 minute decrease in transit

access time would result in a 09 percent increase in probability of choosing this mode It also shows

that each 1 minute decrease in transit travel time and transit access leads to a 0004 decrease in the

probability of choosing the DampR mode

22

The above results show that just one of the policy interactions is not significantly different from

none (ie CordonampFuel) and push policies have a significant role in impelling car users to consider

other modes

7 Planning

To demonstrate the application of the model some of its planning aspects for the city of Tehran are

presented here In fact the model is able to assess the effect of all studied variables which were

grouped into three tiers on commuters mode choice In this section we assess the effect of TDM

policies by adopting their pair-wise combinations for the city Table 6 shows that pull policies in this

study did not impose a major effect on car usage Because the average access time of the sample

was 11 minutes improving it by 50 percent (about 5 minutes) increases the probability of WampR to

509069100 which is equal to a 048 percent increase in considering this mode (and decrease in all

other modes) On the other hand decreasing transit time about 50 percent (about 19 minutes based

on sample average) and similarly improving access time decreases the probability of considering

DampR to 0004195100 which is equal to a 038 percent decrease in considering this mode (and

increase in all other modes) Therefore in this section we focused on the push policies which

appear in the utility function of car mode It is worth noting that in assessing each combination of

policies all other variables were assumed to be fixed as the current state

71 Parking pricing and cordon pricing

The average hourly parking charge of individuals in the sample is 71 Rials This charge is 3000 Rials

per hour in the curb lane on some streets about 2000 Rials per hour in parking lots and free in

alleys and other streets The average amount shows that most of the commuters benefited from

free parking spaces in alleys and non-controlled streets Figure 2 depicts the result of implementing

this policy and the cordon pricing policy simultaneously In line with marginal effects it can be seen

that in implementing each policy separately cordon pricing is more effective than parking pricing in

the planning range It is worth noting that commuters in this study are less sensitive to parking

23

pricing than cordon pricing One reason may be that they thought in any case they would find free

parking space Figure 2 shows that the effect of each policy depends on the level of the other policy

In other words the effect of the simultaneous implementation of two policies is not equal to the

sum of their separate individual effects

Figure 2- Effect of implementing parking pricing and cordon pricing

72 Parking pricing and increasing fuel cost

Figure 3 shows the simultaneous effects of parking pricing and increasing fuel cost policies Although

both of these policies are not effective separately their simultaneous effect is significant It can be

seen that their synergy appears in the higher levels of both policies

Figure 3 - Effect of implementing parking pricing and increasing fuel cost

24

73 Cordon pricing and increasing fuel cost

The effect of implementing cordon pricing and increasing fuel cost is shown in Figure 4 It can be

seen that these two policies have no interaction effect in decreasing car usage This result is verified

by this interactions insignificant marginal value in Table 6

Figure 4- Effect of implementing cordon pricing and increasing fuel cost

8 Conclusions

This study examined the role of TDM policies in individual mode choice for work trips in the city of

Tehran Five policies including increasing parking cost increasing fuel cost cordon pricing in the odd-

even zone of the city transit time reduction and transit access improvement were investigated Of

the five policies the former three were push policies and the latter two were pull policies The

design of experiments approach was used to design the questionnaire that would capture the stated

preferences of car commuters

A number of nested structures based on recognizing differences in the variances associated

with unobserved influences were examined and finally the MNL as a superior model for this study

was developed

25

The results show that for car users who faced TDM policies all of the five studied TDM

policies were factors in the mode choice process The model shows that in considering car usage

push policies play a main role and pull policies do not

Results also show that the interactions of policies are also significant The model shows that

the interaction of parking cost policy and fuel cost policy is significant in prompting car users to

consider other modes It also shows that the interaction of cordon pricing and increase in fuel cost is

effective to discourage motorcycle usage Furthermore the interaction of transit time reduction

policy and access improvements policy is significant in considering DampR as a transportation mode to

workplace

Pull policies in the study were expected to attract individuals to transit modes (pull policies

of this study are related to transit modes) This was verified by the model

Results show that although taxi usage is usually treated as an alternative for urban trips it

was not directly affected by the studied policies This may be due to its special function in Tehran as

a non-public and non-private mode

The results also show that except for the interaction of cordon pricing and increase in fuel

cost other interaction effects have significant marginal effects on mode choice

Results of the model show that to implement a single policy cordon pricing is the most

effective in decreasing car usage

This study assessed workplace commuters Thus future studies should explore the

individuals mode choice with other trip aims Additionally including more policies especially pull

policies in favor of non-transit modes is reasonable to rigorously assess the model

It can be seen that the studied modes are affected differently by the TDM policies and their

interactions The impact of the interaction of policies appeared in the consideration of three out of

six studied modes and also in two out of three developed graphs This issue needs more attention

within the transportation demand management context and further studies should be performed

26

Design of this study is based on the consideration of all two-way interactions with some

degrees of deficiency in D-efficient policy Focusing on the significant interactions of this study is

recommended to perform other studies with more D-efficient values

Further understanding of the impacts of TDM policies might be provided in future research

by focusing on the amount of their synergy in prohibiting car usage

9 References

Arentze TA Hofman F amp Timmermans HJP (2004) Predicting multi-faceted activityndashtravel adjustmentstrategies in response to possible congestion pricing scenarios using an Internet-based stated adaptationexperiment Transport Policy 11 p31ndash41

Burris M amp Patil S (2009) Estimating the Benefits of Managed Lanes Technical Report Texas TexasTransportation Institute University Transportation Center for Mobility

Cao X amp Mokhtarian PL (2005) How do individuals adapt their personal travel A conceptual exploration ofthe consideation of travel-related strategies Transport Policy 12(3) pp199-206

de Palma A amp Lindsey R (2001) Transportation Supply and congestion In International Encyclopedia of theSocial and Behavioral Sciences 1st ed Elsevier p15882ndash15888

Eriksson L Garvill J amp Nordlund AM (2008) Acceptability of single and combined transport policy measuresThe importance of environmental and policy specific beliefs Transportation Research Part A 42 p1117ndash1128

Eriksson L Nordlund AM amp Garvill J (2010) Expected car use reduction in response to structural traveldemand management measures Transportation Research Part F 13 p329ndash342

Golob TF amp Hensher DA (2007) The Trip chaining Activity of Sydney Residents A Cross-Section Assessmentby Age Group with a Focus on Seniors Journal of Transport Geography 15 p298ndash312

Hensher DA amp King J (2001) Parking demand and responsiveness to supply pricing and location in theSydney central business district Transportation Research A 35 pp177-96

Hensher DA amp Rose JM (2007) Development of commuter and non-commuter mode choice models for theassessment of new public transport infrastructure projects A case study Transportation Research Part A41 p428ndash443

Hensher DA Rose JM amp Greene WH (2005) Applied Choice Analysis A Primer New York CambridgeUniversity Press

Iranian Center of Statistics (ICS) (2009) Information of Iranian States [Online] Available at HYPERLINK httpwwwamarorgirUploadModulesContentsasset16tehrantehparthtmlhttpwwwamarorgirUploadModulesContentsasset16tehrantehparthtml [Accessed 13 November2009]

Kingham S Dickinson J amp Copsey S (2001) Travelling to work Will people move out of their cars TransportPolicy 8 pp151-60

Kuhfeld WF (2009) Marketing Research Methods in SAS Experimental Design Choice Conjoint andGraphical Techniques SAS 92 Edition NC USA SAS Institute Inc Cary Available on the Webwwwsupportsascomtechsuptnotetnote_stathtmlmarket

Litman T (2003) The online TDM Encyclopedia mobility management information gateway Transport Policy10 pp245-49

Litman T (2010) Online TDM Encyclopedia [Online] Available at HYPERLINK httpwwwvtpiorghttpwwwvtpiorg [Accessed 21 Augest 2010]

18

access to the city center Younger commuters are less likely to use taxis and individuals with at least

master degrees do consider this mode in addition to their car

Public transit accessed by Drive (DampR)

This mode is affected by the simultaneous interaction of transit time and transit access

(PT_TimeampAccess) which is reflected in the fact that individuals prefer to use this mode for longer

trips Comparing this mode and WampR the first trip start time affects the consideration of this mode

differently Later morning commuters prefer to use their car to access PT modes Such commuters

may have higher income levels or managerial jobs Obviously individuals who are not able to access

PT stations by walking (PTnwacc) are likely to use DampR Serving passengers in daily trips is also a

deterrent in considering this mode which is similar to WampR but with a lower coefficient

Commuters with higher income levels who depend on their car during before or after work

(Dependencycar1+) are likely to use DampR Individuals who use their own car are less likely to use

this mode which is unexpected As a city center develops better PT network coverage and residents

have smaller distances to their workplaces they are unlikely to use DampR This is proven in the model

by a negative sign for D_home_place

Motorcycle (MC)

Increasing fuel cost and cordon pricing simultaneously discourage motorcycle usages Although fuel

cost is expected to reduce motorcycle usage to some extent its combined effect with cordon pricing

also reduces motorcycle usage However this variable is not as strong as other policy variables

=10)

Of the studied modes motorcycle usage is affected by the most commuting variables This may be

due to the fact that this mode is not common Commuting distance has a negative effect on

motorcycle usage which is expected It is worth noting that trip distance appears only in this mode

which may be a reflection of the role of distance in regards to the safety risk in considering this

19

mode Commuters with more stops to serve passengers while commuting (Pattern 2) are not likely

to use this mode which may be due to the poor passenger service of this mode

Individuals who state that commuting is independent of the mode (Pnocarwk) are not likely to use

MC By looking at the (First_Nacco) negative sign this could stem from the fact that the more

passengers there are on the first trip the less likely individuals are to consider MC Regarding the

low capacity of MC and its safety concerns such commuters avoid using this mode Commuters who

pay more parking charges (Park_payment) are less likely to use MC which is expected Individuals

who are dependent on their car during before or after their work time are not likely to use MC

even if they have lower levels of income (DependencyCar1) Individuals who use their own car

(D_car_own) are less likely to use this mode As expected individuals who live in households with

more motorcycle ownership are more likely to use this mode The positive sign of (Permission)

indicates that commuters who have permission to enter the study area do consider MC Because

such commuters generally provide that permission just for car usage this result is unexpected

As with commute variables of all the studied modes MC is affected by the greatest number of

socio-economic variables As expected young commuters (Agelt30) are more likely to use this mode

Commuters with Bachelor of Science degree are less likely to use this mode among others Full time

employees (Emp_full) are less likely to consider MC whereas commuters with more experience in

their jobs prefer to use it Results show that individuals who live in a household with children

younger than 18 are more likely to consider using a car

Tel-Taxi (T_T)

Results show that cordon pricing causes higher probability of using T_T In fact individuals who use

T_T as a mode with similar level of service as cars9 are more willing to pay the cost and make use of

the mode It is worth noting that the effect of cordon pricing in pushing commuters from car usage

9 As this mode does not have driving stress and parking search time in some cases it may have more amount of utility thana car does

20

(000045) is greater than its effect on pulling them to Tel-taxi (000019) This is because of the

possibility of considering other non-car modes

Because consideration of this mode is a function of its operation travel time (Trip_time) appears as

a deterrent in this mode utility function Table 5 shows that individuals are more sensitive to the trip

time when using T_T mode versus taxi which is expected due to their relative costs

The greater the number of full time employees in a family (Nhempfull) the higher the probability of

considering T_T by its commuters which may be due to the higher income level of these

households This is verified by the greater likelihood of using T_T rather than taxis by such

commuters Individuals with higher levels of income who depend on their car during before or after

work time are less likely to use T_T Commuters with lower income levels who state that they use

their car for the sake of comfort (Comfortcar1) are less likely to use T_T which may be due to its

cost Although such individuals do not consider any other modes they specifically avoid T_T Greater

access to cars in a household leads to greater likelihood of T_T usage which could be due to the

higher income level of a household As mentioned before such individuals even avoid taxis

Females who drive to their workplace are more likely to use T_T It seems like this part of society

considers this mode when desiring to avoid the difficulties of driving Younger commuters are less

likely to use T_T and individuals between 30 and 39 years of age are specifically avoiding this mode

Results show that university graduated commuters are more likely to use this mode

6 Marginal effects

To explore the effects of each policy on mode choice and to answer the second issue raised at the

beginning of this paper the marginal effects approach can be adopted Although the coefficients of

the models utility functions show the drivers behavior when facing one or more policies the

marginal effects of policies or their interactions may appropriately show the results of their

implementation More specifically the marginal effect for this study is interpreted as the change in

21

probability given a unit change in a variable ceteris paribus In this section the variable is defined as

a specific policy or a specific interaction of two policies Table 6 presents the marginal effects of the

studied policies and their interactions with mode choice The results are shown in the form of trip

percentages transferred away from the car to the studied modes and the probability-weighted

sample enumeration approach is adopted to find the values It is worth noting that this table is fully

compatible with Table 5 but the marginal effects that were less significant than 90 percent have

been removed

Table 6 - Marginal effects of policies (percent)

Tel-Taxi(TT)

Motorcycle(MC)

Drive ampRide(DampR)

Taxi (T)Walk ampRide(WampR)

Car (C)Mode

Variable-000088Cordon-000140Parking

-09069Access-0000001ParkampFuel

00040PT_TimeampAccess

Table 5 shows that a 1 Rial increase in cordon pricing causes a 000088 percent decrease in car

usage Furthermore a 1 Rial per hour increase in parking cost decreases the car usage probability to

00014 percent By assuming 8 hours for the average parking duration the daily marginal value of

parking cost converts to 000018 percent These values show that cordon pricing is more effective in

forcing individuals not to use their car than increasing parking cost with the same value Results also

show that a 1 Rial increase in parking cost and fuel cost simultaneously decreases the probability of

choosing car usage by 0000001 percent Table 5 also shows that a 1 minute decrease in transit

access time would result in a 09 percent increase in probability of choosing this mode It also shows

that each 1 minute decrease in transit travel time and transit access leads to a 0004 decrease in the

probability of choosing the DampR mode

22

The above results show that just one of the policy interactions is not significantly different from

none (ie CordonampFuel) and push policies have a significant role in impelling car users to consider

other modes

7 Planning

To demonstrate the application of the model some of its planning aspects for the city of Tehran are

presented here In fact the model is able to assess the effect of all studied variables which were

grouped into three tiers on commuters mode choice In this section we assess the effect of TDM

policies by adopting their pair-wise combinations for the city Table 6 shows that pull policies in this

study did not impose a major effect on car usage Because the average access time of the sample

was 11 minutes improving it by 50 percent (about 5 minutes) increases the probability of WampR to

509069100 which is equal to a 048 percent increase in considering this mode (and decrease in all

other modes) On the other hand decreasing transit time about 50 percent (about 19 minutes based

on sample average) and similarly improving access time decreases the probability of considering

DampR to 0004195100 which is equal to a 038 percent decrease in considering this mode (and

increase in all other modes) Therefore in this section we focused on the push policies which

appear in the utility function of car mode It is worth noting that in assessing each combination of

policies all other variables were assumed to be fixed as the current state

71 Parking pricing and cordon pricing

The average hourly parking charge of individuals in the sample is 71 Rials This charge is 3000 Rials

per hour in the curb lane on some streets about 2000 Rials per hour in parking lots and free in

alleys and other streets The average amount shows that most of the commuters benefited from

free parking spaces in alleys and non-controlled streets Figure 2 depicts the result of implementing

this policy and the cordon pricing policy simultaneously In line with marginal effects it can be seen

that in implementing each policy separately cordon pricing is more effective than parking pricing in

the planning range It is worth noting that commuters in this study are less sensitive to parking

23

pricing than cordon pricing One reason may be that they thought in any case they would find free

parking space Figure 2 shows that the effect of each policy depends on the level of the other policy

In other words the effect of the simultaneous implementation of two policies is not equal to the

sum of their separate individual effects

Figure 2- Effect of implementing parking pricing and cordon pricing

72 Parking pricing and increasing fuel cost

Figure 3 shows the simultaneous effects of parking pricing and increasing fuel cost policies Although

both of these policies are not effective separately their simultaneous effect is significant It can be

seen that their synergy appears in the higher levels of both policies

Figure 3 - Effect of implementing parking pricing and increasing fuel cost

24

73 Cordon pricing and increasing fuel cost

The effect of implementing cordon pricing and increasing fuel cost is shown in Figure 4 It can be

seen that these two policies have no interaction effect in decreasing car usage This result is verified

by this interactions insignificant marginal value in Table 6

Figure 4- Effect of implementing cordon pricing and increasing fuel cost

8 Conclusions

This study examined the role of TDM policies in individual mode choice for work trips in the city of

Tehran Five policies including increasing parking cost increasing fuel cost cordon pricing in the odd-

even zone of the city transit time reduction and transit access improvement were investigated Of

the five policies the former three were push policies and the latter two were pull policies The

design of experiments approach was used to design the questionnaire that would capture the stated

preferences of car commuters

A number of nested structures based on recognizing differences in the variances associated

with unobserved influences were examined and finally the MNL as a superior model for this study

was developed

25

The results show that for car users who faced TDM policies all of the five studied TDM

policies were factors in the mode choice process The model shows that in considering car usage

push policies play a main role and pull policies do not

Results also show that the interactions of policies are also significant The model shows that

the interaction of parking cost policy and fuel cost policy is significant in prompting car users to

consider other modes It also shows that the interaction of cordon pricing and increase in fuel cost is

effective to discourage motorcycle usage Furthermore the interaction of transit time reduction

policy and access improvements policy is significant in considering DampR as a transportation mode to

workplace

Pull policies in the study were expected to attract individuals to transit modes (pull policies

of this study are related to transit modes) This was verified by the model

Results show that although taxi usage is usually treated as an alternative for urban trips it

was not directly affected by the studied policies This may be due to its special function in Tehran as

a non-public and non-private mode

The results also show that except for the interaction of cordon pricing and increase in fuel

cost other interaction effects have significant marginal effects on mode choice

Results of the model show that to implement a single policy cordon pricing is the most

effective in decreasing car usage

This study assessed workplace commuters Thus future studies should explore the

individuals mode choice with other trip aims Additionally including more policies especially pull

policies in favor of non-transit modes is reasonable to rigorously assess the model

It can be seen that the studied modes are affected differently by the TDM policies and their

interactions The impact of the interaction of policies appeared in the consideration of three out of

six studied modes and also in two out of three developed graphs This issue needs more attention

within the transportation demand management context and further studies should be performed

26

Design of this study is based on the consideration of all two-way interactions with some

degrees of deficiency in D-efficient policy Focusing on the significant interactions of this study is

recommended to perform other studies with more D-efficient values

Further understanding of the impacts of TDM policies might be provided in future research

by focusing on the amount of their synergy in prohibiting car usage

9 References

Arentze TA Hofman F amp Timmermans HJP (2004) Predicting multi-faceted activityndashtravel adjustmentstrategies in response to possible congestion pricing scenarios using an Internet-based stated adaptationexperiment Transport Policy 11 p31ndash41

Burris M amp Patil S (2009) Estimating the Benefits of Managed Lanes Technical Report Texas TexasTransportation Institute University Transportation Center for Mobility

Cao X amp Mokhtarian PL (2005) How do individuals adapt their personal travel A conceptual exploration ofthe consideation of travel-related strategies Transport Policy 12(3) pp199-206

de Palma A amp Lindsey R (2001) Transportation Supply and congestion In International Encyclopedia of theSocial and Behavioral Sciences 1st ed Elsevier p15882ndash15888

Eriksson L Garvill J amp Nordlund AM (2008) Acceptability of single and combined transport policy measuresThe importance of environmental and policy specific beliefs Transportation Research Part A 42 p1117ndash1128

Eriksson L Nordlund AM amp Garvill J (2010) Expected car use reduction in response to structural traveldemand management measures Transportation Research Part F 13 p329ndash342

Golob TF amp Hensher DA (2007) The Trip chaining Activity of Sydney Residents A Cross-Section Assessmentby Age Group with a Focus on Seniors Journal of Transport Geography 15 p298ndash312

Hensher DA amp King J (2001) Parking demand and responsiveness to supply pricing and location in theSydney central business district Transportation Research A 35 pp177-96

Hensher DA amp Rose JM (2007) Development of commuter and non-commuter mode choice models for theassessment of new public transport infrastructure projects A case study Transportation Research Part A41 p428ndash443

Hensher DA Rose JM amp Greene WH (2005) Applied Choice Analysis A Primer New York CambridgeUniversity Press

Iranian Center of Statistics (ICS) (2009) Information of Iranian States [Online] Available at HYPERLINK httpwwwamarorgirUploadModulesContentsasset16tehrantehparthtmlhttpwwwamarorgirUploadModulesContentsasset16tehrantehparthtml [Accessed 13 November2009]

Kingham S Dickinson J amp Copsey S (2001) Travelling to work Will people move out of their cars TransportPolicy 8 pp151-60

Kuhfeld WF (2009) Marketing Research Methods in SAS Experimental Design Choice Conjoint andGraphical Techniques SAS 92 Edition NC USA SAS Institute Inc Cary Available on the Webwwwsupportsascomtechsuptnotetnote_stathtmlmarket

Litman T (2003) The online TDM Encyclopedia mobility management information gateway Transport Policy10 pp245-49

Litman T (2010) Online TDM Encyclopedia [Online] Available at HYPERLINK httpwwwvtpiorghttpwwwvtpiorg [Accessed 21 Augest 2010]

19

mode Commuters with more stops to serve passengers while commuting (Pattern 2) are not likely

to use this mode which may be due to the poor passenger service of this mode

Individuals who state that commuting is independent of the mode (Pnocarwk) are not likely to use

MC By looking at the (First_Nacco) negative sign this could stem from the fact that the more

passengers there are on the first trip the less likely individuals are to consider MC Regarding the

low capacity of MC and its safety concerns such commuters avoid using this mode Commuters who

pay more parking charges (Park_payment) are less likely to use MC which is expected Individuals

who are dependent on their car during before or after their work time are not likely to use MC

even if they have lower levels of income (DependencyCar1) Individuals who use their own car

(D_car_own) are less likely to use this mode As expected individuals who live in households with

more motorcycle ownership are more likely to use this mode The positive sign of (Permission)

indicates that commuters who have permission to enter the study area do consider MC Because

such commuters generally provide that permission just for car usage this result is unexpected

As with commute variables of all the studied modes MC is affected by the greatest number of

socio-economic variables As expected young commuters (Agelt30) are more likely to use this mode

Commuters with Bachelor of Science degree are less likely to use this mode among others Full time

employees (Emp_full) are less likely to consider MC whereas commuters with more experience in

their jobs prefer to use it Results show that individuals who live in a household with children

younger than 18 are more likely to consider using a car

Tel-Taxi (T_T)

Results show that cordon pricing causes higher probability of using T_T In fact individuals who use

T_T as a mode with similar level of service as cars9 are more willing to pay the cost and make use of

the mode It is worth noting that the effect of cordon pricing in pushing commuters from car usage

9 As this mode does not have driving stress and parking search time in some cases it may have more amount of utility thana car does

20

(000045) is greater than its effect on pulling them to Tel-taxi (000019) This is because of the

possibility of considering other non-car modes

Because consideration of this mode is a function of its operation travel time (Trip_time) appears as

a deterrent in this mode utility function Table 5 shows that individuals are more sensitive to the trip

time when using T_T mode versus taxi which is expected due to their relative costs

The greater the number of full time employees in a family (Nhempfull) the higher the probability of

considering T_T by its commuters which may be due to the higher income level of these

households This is verified by the greater likelihood of using T_T rather than taxis by such

commuters Individuals with higher levels of income who depend on their car during before or after

work time are less likely to use T_T Commuters with lower income levels who state that they use

their car for the sake of comfort (Comfortcar1) are less likely to use T_T which may be due to its

cost Although such individuals do not consider any other modes they specifically avoid T_T Greater

access to cars in a household leads to greater likelihood of T_T usage which could be due to the

higher income level of a household As mentioned before such individuals even avoid taxis

Females who drive to their workplace are more likely to use T_T It seems like this part of society

considers this mode when desiring to avoid the difficulties of driving Younger commuters are less

likely to use T_T and individuals between 30 and 39 years of age are specifically avoiding this mode

Results show that university graduated commuters are more likely to use this mode

6 Marginal effects

To explore the effects of each policy on mode choice and to answer the second issue raised at the

beginning of this paper the marginal effects approach can be adopted Although the coefficients of

the models utility functions show the drivers behavior when facing one or more policies the

marginal effects of policies or their interactions may appropriately show the results of their

implementation More specifically the marginal effect for this study is interpreted as the change in

21

probability given a unit change in a variable ceteris paribus In this section the variable is defined as

a specific policy or a specific interaction of two policies Table 6 presents the marginal effects of the

studied policies and their interactions with mode choice The results are shown in the form of trip

percentages transferred away from the car to the studied modes and the probability-weighted

sample enumeration approach is adopted to find the values It is worth noting that this table is fully

compatible with Table 5 but the marginal effects that were less significant than 90 percent have

been removed

Table 6 - Marginal effects of policies (percent)

Tel-Taxi(TT)

Motorcycle(MC)

Drive ampRide(DampR)

Taxi (T)Walk ampRide(WampR)

Car (C)Mode

Variable-000088Cordon-000140Parking

-09069Access-0000001ParkampFuel

00040PT_TimeampAccess

Table 5 shows that a 1 Rial increase in cordon pricing causes a 000088 percent decrease in car

usage Furthermore a 1 Rial per hour increase in parking cost decreases the car usage probability to

00014 percent By assuming 8 hours for the average parking duration the daily marginal value of

parking cost converts to 000018 percent These values show that cordon pricing is more effective in

forcing individuals not to use their car than increasing parking cost with the same value Results also

show that a 1 Rial increase in parking cost and fuel cost simultaneously decreases the probability of

choosing car usage by 0000001 percent Table 5 also shows that a 1 minute decrease in transit

access time would result in a 09 percent increase in probability of choosing this mode It also shows

that each 1 minute decrease in transit travel time and transit access leads to a 0004 decrease in the

probability of choosing the DampR mode

22

The above results show that just one of the policy interactions is not significantly different from

none (ie CordonampFuel) and push policies have a significant role in impelling car users to consider

other modes

7 Planning

To demonstrate the application of the model some of its planning aspects for the city of Tehran are

presented here In fact the model is able to assess the effect of all studied variables which were

grouped into three tiers on commuters mode choice In this section we assess the effect of TDM

policies by adopting their pair-wise combinations for the city Table 6 shows that pull policies in this

study did not impose a major effect on car usage Because the average access time of the sample

was 11 minutes improving it by 50 percent (about 5 minutes) increases the probability of WampR to

509069100 which is equal to a 048 percent increase in considering this mode (and decrease in all

other modes) On the other hand decreasing transit time about 50 percent (about 19 minutes based

on sample average) and similarly improving access time decreases the probability of considering

DampR to 0004195100 which is equal to a 038 percent decrease in considering this mode (and

increase in all other modes) Therefore in this section we focused on the push policies which

appear in the utility function of car mode It is worth noting that in assessing each combination of

policies all other variables were assumed to be fixed as the current state

71 Parking pricing and cordon pricing

The average hourly parking charge of individuals in the sample is 71 Rials This charge is 3000 Rials

per hour in the curb lane on some streets about 2000 Rials per hour in parking lots and free in

alleys and other streets The average amount shows that most of the commuters benefited from

free parking spaces in alleys and non-controlled streets Figure 2 depicts the result of implementing

this policy and the cordon pricing policy simultaneously In line with marginal effects it can be seen

that in implementing each policy separately cordon pricing is more effective than parking pricing in

the planning range It is worth noting that commuters in this study are less sensitive to parking

23

pricing than cordon pricing One reason may be that they thought in any case they would find free

parking space Figure 2 shows that the effect of each policy depends on the level of the other policy

In other words the effect of the simultaneous implementation of two policies is not equal to the

sum of their separate individual effects

Figure 2- Effect of implementing parking pricing and cordon pricing

72 Parking pricing and increasing fuel cost

Figure 3 shows the simultaneous effects of parking pricing and increasing fuel cost policies Although

both of these policies are not effective separately their simultaneous effect is significant It can be

seen that their synergy appears in the higher levels of both policies

Figure 3 - Effect of implementing parking pricing and increasing fuel cost

24

73 Cordon pricing and increasing fuel cost

The effect of implementing cordon pricing and increasing fuel cost is shown in Figure 4 It can be

seen that these two policies have no interaction effect in decreasing car usage This result is verified

by this interactions insignificant marginal value in Table 6

Figure 4- Effect of implementing cordon pricing and increasing fuel cost

8 Conclusions

This study examined the role of TDM policies in individual mode choice for work trips in the city of

Tehran Five policies including increasing parking cost increasing fuel cost cordon pricing in the odd-

even zone of the city transit time reduction and transit access improvement were investigated Of

the five policies the former three were push policies and the latter two were pull policies The

design of experiments approach was used to design the questionnaire that would capture the stated

preferences of car commuters

A number of nested structures based on recognizing differences in the variances associated

with unobserved influences were examined and finally the MNL as a superior model for this study

was developed

25

The results show that for car users who faced TDM policies all of the five studied TDM

policies were factors in the mode choice process The model shows that in considering car usage

push policies play a main role and pull policies do not

Results also show that the interactions of policies are also significant The model shows that

the interaction of parking cost policy and fuel cost policy is significant in prompting car users to

consider other modes It also shows that the interaction of cordon pricing and increase in fuel cost is

effective to discourage motorcycle usage Furthermore the interaction of transit time reduction

policy and access improvements policy is significant in considering DampR as a transportation mode to

workplace

Pull policies in the study were expected to attract individuals to transit modes (pull policies

of this study are related to transit modes) This was verified by the model

Results show that although taxi usage is usually treated as an alternative for urban trips it

was not directly affected by the studied policies This may be due to its special function in Tehran as

a non-public and non-private mode

The results also show that except for the interaction of cordon pricing and increase in fuel

cost other interaction effects have significant marginal effects on mode choice

Results of the model show that to implement a single policy cordon pricing is the most

effective in decreasing car usage

This study assessed workplace commuters Thus future studies should explore the

individuals mode choice with other trip aims Additionally including more policies especially pull

policies in favor of non-transit modes is reasonable to rigorously assess the model

It can be seen that the studied modes are affected differently by the TDM policies and their

interactions The impact of the interaction of policies appeared in the consideration of three out of

six studied modes and also in two out of three developed graphs This issue needs more attention

within the transportation demand management context and further studies should be performed

26

Design of this study is based on the consideration of all two-way interactions with some

degrees of deficiency in D-efficient policy Focusing on the significant interactions of this study is

recommended to perform other studies with more D-efficient values

Further understanding of the impacts of TDM policies might be provided in future research

by focusing on the amount of their synergy in prohibiting car usage

9 References

Arentze TA Hofman F amp Timmermans HJP (2004) Predicting multi-faceted activityndashtravel adjustmentstrategies in response to possible congestion pricing scenarios using an Internet-based stated adaptationexperiment Transport Policy 11 p31ndash41

Burris M amp Patil S (2009) Estimating the Benefits of Managed Lanes Technical Report Texas TexasTransportation Institute University Transportation Center for Mobility

Cao X amp Mokhtarian PL (2005) How do individuals adapt their personal travel A conceptual exploration ofthe consideation of travel-related strategies Transport Policy 12(3) pp199-206

de Palma A amp Lindsey R (2001) Transportation Supply and congestion In International Encyclopedia of theSocial and Behavioral Sciences 1st ed Elsevier p15882ndash15888

Eriksson L Garvill J amp Nordlund AM (2008) Acceptability of single and combined transport policy measuresThe importance of environmental and policy specific beliefs Transportation Research Part A 42 p1117ndash1128

Eriksson L Nordlund AM amp Garvill J (2010) Expected car use reduction in response to structural traveldemand management measures Transportation Research Part F 13 p329ndash342

Golob TF amp Hensher DA (2007) The Trip chaining Activity of Sydney Residents A Cross-Section Assessmentby Age Group with a Focus on Seniors Journal of Transport Geography 15 p298ndash312

Hensher DA amp King J (2001) Parking demand and responsiveness to supply pricing and location in theSydney central business district Transportation Research A 35 pp177-96

Hensher DA amp Rose JM (2007) Development of commuter and non-commuter mode choice models for theassessment of new public transport infrastructure projects A case study Transportation Research Part A41 p428ndash443

Hensher DA Rose JM amp Greene WH (2005) Applied Choice Analysis A Primer New York CambridgeUniversity Press

Iranian Center of Statistics (ICS) (2009) Information of Iranian States [Online] Available at HYPERLINK httpwwwamarorgirUploadModulesContentsasset16tehrantehparthtmlhttpwwwamarorgirUploadModulesContentsasset16tehrantehparthtml [Accessed 13 November2009]

Kingham S Dickinson J amp Copsey S (2001) Travelling to work Will people move out of their cars TransportPolicy 8 pp151-60

Kuhfeld WF (2009) Marketing Research Methods in SAS Experimental Design Choice Conjoint andGraphical Techniques SAS 92 Edition NC USA SAS Institute Inc Cary Available on the Webwwwsupportsascomtechsuptnotetnote_stathtmlmarket

Litman T (2003) The online TDM Encyclopedia mobility management information gateway Transport Policy10 pp245-49

Litman T (2010) Online TDM Encyclopedia [Online] Available at HYPERLINK httpwwwvtpiorghttpwwwvtpiorg [Accessed 21 Augest 2010]

20

(000045) is greater than its effect on pulling them to Tel-taxi (000019) This is because of the

possibility of considering other non-car modes

Because consideration of this mode is a function of its operation travel time (Trip_time) appears as

a deterrent in this mode utility function Table 5 shows that individuals are more sensitive to the trip

time when using T_T mode versus taxi which is expected due to their relative costs

The greater the number of full time employees in a family (Nhempfull) the higher the probability of

considering T_T by its commuters which may be due to the higher income level of these

households This is verified by the greater likelihood of using T_T rather than taxis by such

commuters Individuals with higher levels of income who depend on their car during before or after

work time are less likely to use T_T Commuters with lower income levels who state that they use

their car for the sake of comfort (Comfortcar1) are less likely to use T_T which may be due to its

cost Although such individuals do not consider any other modes they specifically avoid T_T Greater

access to cars in a household leads to greater likelihood of T_T usage which could be due to the

higher income level of a household As mentioned before such individuals even avoid taxis

Females who drive to their workplace are more likely to use T_T It seems like this part of society

considers this mode when desiring to avoid the difficulties of driving Younger commuters are less

likely to use T_T and individuals between 30 and 39 years of age are specifically avoiding this mode

Results show that university graduated commuters are more likely to use this mode

6 Marginal effects

To explore the effects of each policy on mode choice and to answer the second issue raised at the

beginning of this paper the marginal effects approach can be adopted Although the coefficients of

the models utility functions show the drivers behavior when facing one or more policies the

marginal effects of policies or their interactions may appropriately show the results of their

implementation More specifically the marginal effect for this study is interpreted as the change in

21

probability given a unit change in a variable ceteris paribus In this section the variable is defined as

a specific policy or a specific interaction of two policies Table 6 presents the marginal effects of the

studied policies and their interactions with mode choice The results are shown in the form of trip

percentages transferred away from the car to the studied modes and the probability-weighted

sample enumeration approach is adopted to find the values It is worth noting that this table is fully

compatible with Table 5 but the marginal effects that were less significant than 90 percent have

been removed

Table 6 - Marginal effects of policies (percent)

Tel-Taxi(TT)

Motorcycle(MC)

Drive ampRide(DampR)

Taxi (T)Walk ampRide(WampR)

Car (C)Mode

Variable-000088Cordon-000140Parking

-09069Access-0000001ParkampFuel

00040PT_TimeampAccess

Table 5 shows that a 1 Rial increase in cordon pricing causes a 000088 percent decrease in car

usage Furthermore a 1 Rial per hour increase in parking cost decreases the car usage probability to

00014 percent By assuming 8 hours for the average parking duration the daily marginal value of

parking cost converts to 000018 percent These values show that cordon pricing is more effective in

forcing individuals not to use their car than increasing parking cost with the same value Results also

show that a 1 Rial increase in parking cost and fuel cost simultaneously decreases the probability of

choosing car usage by 0000001 percent Table 5 also shows that a 1 minute decrease in transit

access time would result in a 09 percent increase in probability of choosing this mode It also shows

that each 1 minute decrease in transit travel time and transit access leads to a 0004 decrease in the

probability of choosing the DampR mode

22

The above results show that just one of the policy interactions is not significantly different from

none (ie CordonampFuel) and push policies have a significant role in impelling car users to consider

other modes

7 Planning

To demonstrate the application of the model some of its planning aspects for the city of Tehran are

presented here In fact the model is able to assess the effect of all studied variables which were

grouped into three tiers on commuters mode choice In this section we assess the effect of TDM

policies by adopting their pair-wise combinations for the city Table 6 shows that pull policies in this

study did not impose a major effect on car usage Because the average access time of the sample

was 11 minutes improving it by 50 percent (about 5 minutes) increases the probability of WampR to

509069100 which is equal to a 048 percent increase in considering this mode (and decrease in all

other modes) On the other hand decreasing transit time about 50 percent (about 19 minutes based

on sample average) and similarly improving access time decreases the probability of considering

DampR to 0004195100 which is equal to a 038 percent decrease in considering this mode (and

increase in all other modes) Therefore in this section we focused on the push policies which

appear in the utility function of car mode It is worth noting that in assessing each combination of

policies all other variables were assumed to be fixed as the current state

71 Parking pricing and cordon pricing

The average hourly parking charge of individuals in the sample is 71 Rials This charge is 3000 Rials

per hour in the curb lane on some streets about 2000 Rials per hour in parking lots and free in

alleys and other streets The average amount shows that most of the commuters benefited from

free parking spaces in alleys and non-controlled streets Figure 2 depicts the result of implementing

this policy and the cordon pricing policy simultaneously In line with marginal effects it can be seen

that in implementing each policy separately cordon pricing is more effective than parking pricing in

the planning range It is worth noting that commuters in this study are less sensitive to parking

23

pricing than cordon pricing One reason may be that they thought in any case they would find free

parking space Figure 2 shows that the effect of each policy depends on the level of the other policy

In other words the effect of the simultaneous implementation of two policies is not equal to the

sum of their separate individual effects

Figure 2- Effect of implementing parking pricing and cordon pricing

72 Parking pricing and increasing fuel cost

Figure 3 shows the simultaneous effects of parking pricing and increasing fuel cost policies Although

both of these policies are not effective separately their simultaneous effect is significant It can be

seen that their synergy appears in the higher levels of both policies

Figure 3 - Effect of implementing parking pricing and increasing fuel cost

24

73 Cordon pricing and increasing fuel cost

The effect of implementing cordon pricing and increasing fuel cost is shown in Figure 4 It can be

seen that these two policies have no interaction effect in decreasing car usage This result is verified

by this interactions insignificant marginal value in Table 6

Figure 4- Effect of implementing cordon pricing and increasing fuel cost

8 Conclusions

This study examined the role of TDM policies in individual mode choice for work trips in the city of

Tehran Five policies including increasing parking cost increasing fuel cost cordon pricing in the odd-

even zone of the city transit time reduction and transit access improvement were investigated Of

the five policies the former three were push policies and the latter two were pull policies The

design of experiments approach was used to design the questionnaire that would capture the stated

preferences of car commuters

A number of nested structures based on recognizing differences in the variances associated

with unobserved influences were examined and finally the MNL as a superior model for this study

was developed

25

The results show that for car users who faced TDM policies all of the five studied TDM

policies were factors in the mode choice process The model shows that in considering car usage

push policies play a main role and pull policies do not

Results also show that the interactions of policies are also significant The model shows that

the interaction of parking cost policy and fuel cost policy is significant in prompting car users to

consider other modes It also shows that the interaction of cordon pricing and increase in fuel cost is

effective to discourage motorcycle usage Furthermore the interaction of transit time reduction

policy and access improvements policy is significant in considering DampR as a transportation mode to

workplace

Pull policies in the study were expected to attract individuals to transit modes (pull policies

of this study are related to transit modes) This was verified by the model

Results show that although taxi usage is usually treated as an alternative for urban trips it

was not directly affected by the studied policies This may be due to its special function in Tehran as

a non-public and non-private mode

The results also show that except for the interaction of cordon pricing and increase in fuel

cost other interaction effects have significant marginal effects on mode choice

Results of the model show that to implement a single policy cordon pricing is the most

effective in decreasing car usage

This study assessed workplace commuters Thus future studies should explore the

individuals mode choice with other trip aims Additionally including more policies especially pull

policies in favor of non-transit modes is reasonable to rigorously assess the model

It can be seen that the studied modes are affected differently by the TDM policies and their

interactions The impact of the interaction of policies appeared in the consideration of three out of

six studied modes and also in two out of three developed graphs This issue needs more attention

within the transportation demand management context and further studies should be performed

26

Design of this study is based on the consideration of all two-way interactions with some

degrees of deficiency in D-efficient policy Focusing on the significant interactions of this study is

recommended to perform other studies with more D-efficient values

Further understanding of the impacts of TDM policies might be provided in future research

by focusing on the amount of their synergy in prohibiting car usage

9 References

Arentze TA Hofman F amp Timmermans HJP (2004) Predicting multi-faceted activityndashtravel adjustmentstrategies in response to possible congestion pricing scenarios using an Internet-based stated adaptationexperiment Transport Policy 11 p31ndash41

Burris M amp Patil S (2009) Estimating the Benefits of Managed Lanes Technical Report Texas TexasTransportation Institute University Transportation Center for Mobility

Cao X amp Mokhtarian PL (2005) How do individuals adapt their personal travel A conceptual exploration ofthe consideation of travel-related strategies Transport Policy 12(3) pp199-206

de Palma A amp Lindsey R (2001) Transportation Supply and congestion In International Encyclopedia of theSocial and Behavioral Sciences 1st ed Elsevier p15882ndash15888

Eriksson L Garvill J amp Nordlund AM (2008) Acceptability of single and combined transport policy measuresThe importance of environmental and policy specific beliefs Transportation Research Part A 42 p1117ndash1128

Eriksson L Nordlund AM amp Garvill J (2010) Expected car use reduction in response to structural traveldemand management measures Transportation Research Part F 13 p329ndash342

Golob TF amp Hensher DA (2007) The Trip chaining Activity of Sydney Residents A Cross-Section Assessmentby Age Group with a Focus on Seniors Journal of Transport Geography 15 p298ndash312

Hensher DA amp King J (2001) Parking demand and responsiveness to supply pricing and location in theSydney central business district Transportation Research A 35 pp177-96

Hensher DA amp Rose JM (2007) Development of commuter and non-commuter mode choice models for theassessment of new public transport infrastructure projects A case study Transportation Research Part A41 p428ndash443

Hensher DA Rose JM amp Greene WH (2005) Applied Choice Analysis A Primer New York CambridgeUniversity Press

Iranian Center of Statistics (ICS) (2009) Information of Iranian States [Online] Available at HYPERLINK httpwwwamarorgirUploadModulesContentsasset16tehrantehparthtmlhttpwwwamarorgirUploadModulesContentsasset16tehrantehparthtml [Accessed 13 November2009]

Kingham S Dickinson J amp Copsey S (2001) Travelling to work Will people move out of their cars TransportPolicy 8 pp151-60

Kuhfeld WF (2009) Marketing Research Methods in SAS Experimental Design Choice Conjoint andGraphical Techniques SAS 92 Edition NC USA SAS Institute Inc Cary Available on the Webwwwsupportsascomtechsuptnotetnote_stathtmlmarket

Litman T (2003) The online TDM Encyclopedia mobility management information gateway Transport Policy10 pp245-49

Litman T (2010) Online TDM Encyclopedia [Online] Available at HYPERLINK httpwwwvtpiorghttpwwwvtpiorg [Accessed 21 Augest 2010]

21

probability given a unit change in a variable ceteris paribus In this section the variable is defined as

a specific policy or a specific interaction of two policies Table 6 presents the marginal effects of the

studied policies and their interactions with mode choice The results are shown in the form of trip

percentages transferred away from the car to the studied modes and the probability-weighted

sample enumeration approach is adopted to find the values It is worth noting that this table is fully

compatible with Table 5 but the marginal effects that were less significant than 90 percent have

been removed

Table 6 - Marginal effects of policies (percent)

Tel-Taxi(TT)

Motorcycle(MC)

Drive ampRide(DampR)

Taxi (T)Walk ampRide(WampR)

Car (C)Mode

Variable-000088Cordon-000140Parking

-09069Access-0000001ParkampFuel

00040PT_TimeampAccess

Table 5 shows that a 1 Rial increase in cordon pricing causes a 000088 percent decrease in car

usage Furthermore a 1 Rial per hour increase in parking cost decreases the car usage probability to

00014 percent By assuming 8 hours for the average parking duration the daily marginal value of

parking cost converts to 000018 percent These values show that cordon pricing is more effective in

forcing individuals not to use their car than increasing parking cost with the same value Results also

show that a 1 Rial increase in parking cost and fuel cost simultaneously decreases the probability of

choosing car usage by 0000001 percent Table 5 also shows that a 1 minute decrease in transit

access time would result in a 09 percent increase in probability of choosing this mode It also shows

that each 1 minute decrease in transit travel time and transit access leads to a 0004 decrease in the

probability of choosing the DampR mode

22

The above results show that just one of the policy interactions is not significantly different from

none (ie CordonampFuel) and push policies have a significant role in impelling car users to consider

other modes

7 Planning

To demonstrate the application of the model some of its planning aspects for the city of Tehran are

presented here In fact the model is able to assess the effect of all studied variables which were

grouped into three tiers on commuters mode choice In this section we assess the effect of TDM

policies by adopting their pair-wise combinations for the city Table 6 shows that pull policies in this

study did not impose a major effect on car usage Because the average access time of the sample

was 11 minutes improving it by 50 percent (about 5 minutes) increases the probability of WampR to

509069100 which is equal to a 048 percent increase in considering this mode (and decrease in all

other modes) On the other hand decreasing transit time about 50 percent (about 19 minutes based

on sample average) and similarly improving access time decreases the probability of considering

DampR to 0004195100 which is equal to a 038 percent decrease in considering this mode (and

increase in all other modes) Therefore in this section we focused on the push policies which

appear in the utility function of car mode It is worth noting that in assessing each combination of

policies all other variables were assumed to be fixed as the current state

71 Parking pricing and cordon pricing

The average hourly parking charge of individuals in the sample is 71 Rials This charge is 3000 Rials

per hour in the curb lane on some streets about 2000 Rials per hour in parking lots and free in

alleys and other streets The average amount shows that most of the commuters benefited from

free parking spaces in alleys and non-controlled streets Figure 2 depicts the result of implementing

this policy and the cordon pricing policy simultaneously In line with marginal effects it can be seen

that in implementing each policy separately cordon pricing is more effective than parking pricing in

the planning range It is worth noting that commuters in this study are less sensitive to parking

23

pricing than cordon pricing One reason may be that they thought in any case they would find free

parking space Figure 2 shows that the effect of each policy depends on the level of the other policy

In other words the effect of the simultaneous implementation of two policies is not equal to the

sum of their separate individual effects

Figure 2- Effect of implementing parking pricing and cordon pricing

72 Parking pricing and increasing fuel cost

Figure 3 shows the simultaneous effects of parking pricing and increasing fuel cost policies Although

both of these policies are not effective separately their simultaneous effect is significant It can be

seen that their synergy appears in the higher levels of both policies

Figure 3 - Effect of implementing parking pricing and increasing fuel cost

24

73 Cordon pricing and increasing fuel cost

The effect of implementing cordon pricing and increasing fuel cost is shown in Figure 4 It can be

seen that these two policies have no interaction effect in decreasing car usage This result is verified

by this interactions insignificant marginal value in Table 6

Figure 4- Effect of implementing cordon pricing and increasing fuel cost

8 Conclusions

This study examined the role of TDM policies in individual mode choice for work trips in the city of

Tehran Five policies including increasing parking cost increasing fuel cost cordon pricing in the odd-

even zone of the city transit time reduction and transit access improvement were investigated Of

the five policies the former three were push policies and the latter two were pull policies The

design of experiments approach was used to design the questionnaire that would capture the stated

preferences of car commuters

A number of nested structures based on recognizing differences in the variances associated

with unobserved influences were examined and finally the MNL as a superior model for this study

was developed

25

The results show that for car users who faced TDM policies all of the five studied TDM

policies were factors in the mode choice process The model shows that in considering car usage

push policies play a main role and pull policies do not

Results also show that the interactions of policies are also significant The model shows that

the interaction of parking cost policy and fuel cost policy is significant in prompting car users to

consider other modes It also shows that the interaction of cordon pricing and increase in fuel cost is

effective to discourage motorcycle usage Furthermore the interaction of transit time reduction

policy and access improvements policy is significant in considering DampR as a transportation mode to

workplace

Pull policies in the study were expected to attract individuals to transit modes (pull policies

of this study are related to transit modes) This was verified by the model

Results show that although taxi usage is usually treated as an alternative for urban trips it

was not directly affected by the studied policies This may be due to its special function in Tehran as

a non-public and non-private mode

The results also show that except for the interaction of cordon pricing and increase in fuel

cost other interaction effects have significant marginal effects on mode choice

Results of the model show that to implement a single policy cordon pricing is the most

effective in decreasing car usage

This study assessed workplace commuters Thus future studies should explore the

individuals mode choice with other trip aims Additionally including more policies especially pull

policies in favor of non-transit modes is reasonable to rigorously assess the model

It can be seen that the studied modes are affected differently by the TDM policies and their

interactions The impact of the interaction of policies appeared in the consideration of three out of

six studied modes and also in two out of three developed graphs This issue needs more attention

within the transportation demand management context and further studies should be performed

26

Design of this study is based on the consideration of all two-way interactions with some

degrees of deficiency in D-efficient policy Focusing on the significant interactions of this study is

recommended to perform other studies with more D-efficient values

Further understanding of the impacts of TDM policies might be provided in future research

by focusing on the amount of their synergy in prohibiting car usage

9 References

Arentze TA Hofman F amp Timmermans HJP (2004) Predicting multi-faceted activityndashtravel adjustmentstrategies in response to possible congestion pricing scenarios using an Internet-based stated adaptationexperiment Transport Policy 11 p31ndash41

Burris M amp Patil S (2009) Estimating the Benefits of Managed Lanes Technical Report Texas TexasTransportation Institute University Transportation Center for Mobility

Cao X amp Mokhtarian PL (2005) How do individuals adapt their personal travel A conceptual exploration ofthe consideation of travel-related strategies Transport Policy 12(3) pp199-206

de Palma A amp Lindsey R (2001) Transportation Supply and congestion In International Encyclopedia of theSocial and Behavioral Sciences 1st ed Elsevier p15882ndash15888

Eriksson L Garvill J amp Nordlund AM (2008) Acceptability of single and combined transport policy measuresThe importance of environmental and policy specific beliefs Transportation Research Part A 42 p1117ndash1128

Eriksson L Nordlund AM amp Garvill J (2010) Expected car use reduction in response to structural traveldemand management measures Transportation Research Part F 13 p329ndash342

Golob TF amp Hensher DA (2007) The Trip chaining Activity of Sydney Residents A Cross-Section Assessmentby Age Group with a Focus on Seniors Journal of Transport Geography 15 p298ndash312

Hensher DA amp King J (2001) Parking demand and responsiveness to supply pricing and location in theSydney central business district Transportation Research A 35 pp177-96

Hensher DA amp Rose JM (2007) Development of commuter and non-commuter mode choice models for theassessment of new public transport infrastructure projects A case study Transportation Research Part A41 p428ndash443

Hensher DA Rose JM amp Greene WH (2005) Applied Choice Analysis A Primer New York CambridgeUniversity Press

Iranian Center of Statistics (ICS) (2009) Information of Iranian States [Online] Available at HYPERLINK httpwwwamarorgirUploadModulesContentsasset16tehrantehparthtmlhttpwwwamarorgirUploadModulesContentsasset16tehrantehparthtml [Accessed 13 November2009]

Kingham S Dickinson J amp Copsey S (2001) Travelling to work Will people move out of their cars TransportPolicy 8 pp151-60

Kuhfeld WF (2009) Marketing Research Methods in SAS Experimental Design Choice Conjoint andGraphical Techniques SAS 92 Edition NC USA SAS Institute Inc Cary Available on the Webwwwsupportsascomtechsuptnotetnote_stathtmlmarket

Litman T (2003) The online TDM Encyclopedia mobility management information gateway Transport Policy10 pp245-49

Litman T (2010) Online TDM Encyclopedia [Online] Available at HYPERLINK httpwwwvtpiorghttpwwwvtpiorg [Accessed 21 Augest 2010]

22

The above results show that just one of the policy interactions is not significantly different from

none (ie CordonampFuel) and push policies have a significant role in impelling car users to consider

other modes

7 Planning

To demonstrate the application of the model some of its planning aspects for the city of Tehran are

presented here In fact the model is able to assess the effect of all studied variables which were

grouped into three tiers on commuters mode choice In this section we assess the effect of TDM

policies by adopting their pair-wise combinations for the city Table 6 shows that pull policies in this

study did not impose a major effect on car usage Because the average access time of the sample

was 11 minutes improving it by 50 percent (about 5 minutes) increases the probability of WampR to

509069100 which is equal to a 048 percent increase in considering this mode (and decrease in all

other modes) On the other hand decreasing transit time about 50 percent (about 19 minutes based

on sample average) and similarly improving access time decreases the probability of considering

DampR to 0004195100 which is equal to a 038 percent decrease in considering this mode (and

increase in all other modes) Therefore in this section we focused on the push policies which

appear in the utility function of car mode It is worth noting that in assessing each combination of

policies all other variables were assumed to be fixed as the current state

71 Parking pricing and cordon pricing

The average hourly parking charge of individuals in the sample is 71 Rials This charge is 3000 Rials

per hour in the curb lane on some streets about 2000 Rials per hour in parking lots and free in

alleys and other streets The average amount shows that most of the commuters benefited from

free parking spaces in alleys and non-controlled streets Figure 2 depicts the result of implementing

this policy and the cordon pricing policy simultaneously In line with marginal effects it can be seen

that in implementing each policy separately cordon pricing is more effective than parking pricing in

the planning range It is worth noting that commuters in this study are less sensitive to parking

23

pricing than cordon pricing One reason may be that they thought in any case they would find free

parking space Figure 2 shows that the effect of each policy depends on the level of the other policy

In other words the effect of the simultaneous implementation of two policies is not equal to the

sum of their separate individual effects

Figure 2- Effect of implementing parking pricing and cordon pricing

72 Parking pricing and increasing fuel cost

Figure 3 shows the simultaneous effects of parking pricing and increasing fuel cost policies Although

both of these policies are not effective separately their simultaneous effect is significant It can be

seen that their synergy appears in the higher levels of both policies

Figure 3 - Effect of implementing parking pricing and increasing fuel cost

24

73 Cordon pricing and increasing fuel cost

The effect of implementing cordon pricing and increasing fuel cost is shown in Figure 4 It can be

seen that these two policies have no interaction effect in decreasing car usage This result is verified

by this interactions insignificant marginal value in Table 6

Figure 4- Effect of implementing cordon pricing and increasing fuel cost

8 Conclusions

This study examined the role of TDM policies in individual mode choice for work trips in the city of

Tehran Five policies including increasing parking cost increasing fuel cost cordon pricing in the odd-

even zone of the city transit time reduction and transit access improvement were investigated Of

the five policies the former three were push policies and the latter two were pull policies The

design of experiments approach was used to design the questionnaire that would capture the stated

preferences of car commuters

A number of nested structures based on recognizing differences in the variances associated

with unobserved influences were examined and finally the MNL as a superior model for this study

was developed

25

The results show that for car users who faced TDM policies all of the five studied TDM

policies were factors in the mode choice process The model shows that in considering car usage

push policies play a main role and pull policies do not

Results also show that the interactions of policies are also significant The model shows that

the interaction of parking cost policy and fuel cost policy is significant in prompting car users to

consider other modes It also shows that the interaction of cordon pricing and increase in fuel cost is

effective to discourage motorcycle usage Furthermore the interaction of transit time reduction

policy and access improvements policy is significant in considering DampR as a transportation mode to

workplace

Pull policies in the study were expected to attract individuals to transit modes (pull policies

of this study are related to transit modes) This was verified by the model

Results show that although taxi usage is usually treated as an alternative for urban trips it

was not directly affected by the studied policies This may be due to its special function in Tehran as

a non-public and non-private mode

The results also show that except for the interaction of cordon pricing and increase in fuel

cost other interaction effects have significant marginal effects on mode choice

Results of the model show that to implement a single policy cordon pricing is the most

effective in decreasing car usage

This study assessed workplace commuters Thus future studies should explore the

individuals mode choice with other trip aims Additionally including more policies especially pull

policies in favor of non-transit modes is reasonable to rigorously assess the model

It can be seen that the studied modes are affected differently by the TDM policies and their

interactions The impact of the interaction of policies appeared in the consideration of three out of

six studied modes and also in two out of three developed graphs This issue needs more attention

within the transportation demand management context and further studies should be performed

26

Design of this study is based on the consideration of all two-way interactions with some

degrees of deficiency in D-efficient policy Focusing on the significant interactions of this study is

recommended to perform other studies with more D-efficient values

Further understanding of the impacts of TDM policies might be provided in future research

by focusing on the amount of their synergy in prohibiting car usage

9 References

Arentze TA Hofman F amp Timmermans HJP (2004) Predicting multi-faceted activityndashtravel adjustmentstrategies in response to possible congestion pricing scenarios using an Internet-based stated adaptationexperiment Transport Policy 11 p31ndash41

Burris M amp Patil S (2009) Estimating the Benefits of Managed Lanes Technical Report Texas TexasTransportation Institute University Transportation Center for Mobility

Cao X amp Mokhtarian PL (2005) How do individuals adapt their personal travel A conceptual exploration ofthe consideation of travel-related strategies Transport Policy 12(3) pp199-206

de Palma A amp Lindsey R (2001) Transportation Supply and congestion In International Encyclopedia of theSocial and Behavioral Sciences 1st ed Elsevier p15882ndash15888

Eriksson L Garvill J amp Nordlund AM (2008) Acceptability of single and combined transport policy measuresThe importance of environmental and policy specific beliefs Transportation Research Part A 42 p1117ndash1128

Eriksson L Nordlund AM amp Garvill J (2010) Expected car use reduction in response to structural traveldemand management measures Transportation Research Part F 13 p329ndash342

Golob TF amp Hensher DA (2007) The Trip chaining Activity of Sydney Residents A Cross-Section Assessmentby Age Group with a Focus on Seniors Journal of Transport Geography 15 p298ndash312

Hensher DA amp King J (2001) Parking demand and responsiveness to supply pricing and location in theSydney central business district Transportation Research A 35 pp177-96

Hensher DA amp Rose JM (2007) Development of commuter and non-commuter mode choice models for theassessment of new public transport infrastructure projects A case study Transportation Research Part A41 p428ndash443

Hensher DA Rose JM amp Greene WH (2005) Applied Choice Analysis A Primer New York CambridgeUniversity Press

Iranian Center of Statistics (ICS) (2009) Information of Iranian States [Online] Available at HYPERLINK httpwwwamarorgirUploadModulesContentsasset16tehrantehparthtmlhttpwwwamarorgirUploadModulesContentsasset16tehrantehparthtml [Accessed 13 November2009]

Kingham S Dickinson J amp Copsey S (2001) Travelling to work Will people move out of their cars TransportPolicy 8 pp151-60

Kuhfeld WF (2009) Marketing Research Methods in SAS Experimental Design Choice Conjoint andGraphical Techniques SAS 92 Edition NC USA SAS Institute Inc Cary Available on the Webwwwsupportsascomtechsuptnotetnote_stathtmlmarket

Litman T (2003) The online TDM Encyclopedia mobility management information gateway Transport Policy10 pp245-49

Litman T (2010) Online TDM Encyclopedia [Online] Available at HYPERLINK httpwwwvtpiorghttpwwwvtpiorg [Accessed 21 Augest 2010]

23

pricing than cordon pricing One reason may be that they thought in any case they would find free

parking space Figure 2 shows that the effect of each policy depends on the level of the other policy

In other words the effect of the simultaneous implementation of two policies is not equal to the

sum of their separate individual effects

Figure 2- Effect of implementing parking pricing and cordon pricing

72 Parking pricing and increasing fuel cost

Figure 3 shows the simultaneous effects of parking pricing and increasing fuel cost policies Although

both of these policies are not effective separately their simultaneous effect is significant It can be

seen that their synergy appears in the higher levels of both policies

Figure 3 - Effect of implementing parking pricing and increasing fuel cost

24

73 Cordon pricing and increasing fuel cost

The effect of implementing cordon pricing and increasing fuel cost is shown in Figure 4 It can be

seen that these two policies have no interaction effect in decreasing car usage This result is verified

by this interactions insignificant marginal value in Table 6

Figure 4- Effect of implementing cordon pricing and increasing fuel cost

8 Conclusions

This study examined the role of TDM policies in individual mode choice for work trips in the city of

Tehran Five policies including increasing parking cost increasing fuel cost cordon pricing in the odd-

even zone of the city transit time reduction and transit access improvement were investigated Of

the five policies the former three were push policies and the latter two were pull policies The

design of experiments approach was used to design the questionnaire that would capture the stated

preferences of car commuters

A number of nested structures based on recognizing differences in the variances associated

with unobserved influences were examined and finally the MNL as a superior model for this study

was developed

25

The results show that for car users who faced TDM policies all of the five studied TDM

policies were factors in the mode choice process The model shows that in considering car usage

push policies play a main role and pull policies do not

Results also show that the interactions of policies are also significant The model shows that

the interaction of parking cost policy and fuel cost policy is significant in prompting car users to

consider other modes It also shows that the interaction of cordon pricing and increase in fuel cost is

effective to discourage motorcycle usage Furthermore the interaction of transit time reduction

policy and access improvements policy is significant in considering DampR as a transportation mode to

workplace

Pull policies in the study were expected to attract individuals to transit modes (pull policies

of this study are related to transit modes) This was verified by the model

Results show that although taxi usage is usually treated as an alternative for urban trips it

was not directly affected by the studied policies This may be due to its special function in Tehran as

a non-public and non-private mode

The results also show that except for the interaction of cordon pricing and increase in fuel

cost other interaction effects have significant marginal effects on mode choice

Results of the model show that to implement a single policy cordon pricing is the most

effective in decreasing car usage

This study assessed workplace commuters Thus future studies should explore the

individuals mode choice with other trip aims Additionally including more policies especially pull

policies in favor of non-transit modes is reasonable to rigorously assess the model

It can be seen that the studied modes are affected differently by the TDM policies and their

interactions The impact of the interaction of policies appeared in the consideration of three out of

six studied modes and also in two out of three developed graphs This issue needs more attention

within the transportation demand management context and further studies should be performed

26

Design of this study is based on the consideration of all two-way interactions with some

degrees of deficiency in D-efficient policy Focusing on the significant interactions of this study is

recommended to perform other studies with more D-efficient values

Further understanding of the impacts of TDM policies might be provided in future research

by focusing on the amount of their synergy in prohibiting car usage

9 References

Arentze TA Hofman F amp Timmermans HJP (2004) Predicting multi-faceted activityndashtravel adjustmentstrategies in response to possible congestion pricing scenarios using an Internet-based stated adaptationexperiment Transport Policy 11 p31ndash41

Burris M amp Patil S (2009) Estimating the Benefits of Managed Lanes Technical Report Texas TexasTransportation Institute University Transportation Center for Mobility

Cao X amp Mokhtarian PL (2005) How do individuals adapt their personal travel A conceptual exploration ofthe consideation of travel-related strategies Transport Policy 12(3) pp199-206

de Palma A amp Lindsey R (2001) Transportation Supply and congestion In International Encyclopedia of theSocial and Behavioral Sciences 1st ed Elsevier p15882ndash15888

Eriksson L Garvill J amp Nordlund AM (2008) Acceptability of single and combined transport policy measuresThe importance of environmental and policy specific beliefs Transportation Research Part A 42 p1117ndash1128

Eriksson L Nordlund AM amp Garvill J (2010) Expected car use reduction in response to structural traveldemand management measures Transportation Research Part F 13 p329ndash342

Golob TF amp Hensher DA (2007) The Trip chaining Activity of Sydney Residents A Cross-Section Assessmentby Age Group with a Focus on Seniors Journal of Transport Geography 15 p298ndash312

Hensher DA amp King J (2001) Parking demand and responsiveness to supply pricing and location in theSydney central business district Transportation Research A 35 pp177-96

Hensher DA amp Rose JM (2007) Development of commuter and non-commuter mode choice models for theassessment of new public transport infrastructure projects A case study Transportation Research Part A41 p428ndash443

Hensher DA Rose JM amp Greene WH (2005) Applied Choice Analysis A Primer New York CambridgeUniversity Press

Iranian Center of Statistics (ICS) (2009) Information of Iranian States [Online] Available at HYPERLINK httpwwwamarorgirUploadModulesContentsasset16tehrantehparthtmlhttpwwwamarorgirUploadModulesContentsasset16tehrantehparthtml [Accessed 13 November2009]

Kingham S Dickinson J amp Copsey S (2001) Travelling to work Will people move out of their cars TransportPolicy 8 pp151-60

Kuhfeld WF (2009) Marketing Research Methods in SAS Experimental Design Choice Conjoint andGraphical Techniques SAS 92 Edition NC USA SAS Institute Inc Cary Available on the Webwwwsupportsascomtechsuptnotetnote_stathtmlmarket

Litman T (2003) The online TDM Encyclopedia mobility management information gateway Transport Policy10 pp245-49

Litman T (2010) Online TDM Encyclopedia [Online] Available at HYPERLINK httpwwwvtpiorghttpwwwvtpiorg [Accessed 21 Augest 2010]

24

73 Cordon pricing and increasing fuel cost

The effect of implementing cordon pricing and increasing fuel cost is shown in Figure 4 It can be

seen that these two policies have no interaction effect in decreasing car usage This result is verified

by this interactions insignificant marginal value in Table 6

Figure 4- Effect of implementing cordon pricing and increasing fuel cost

8 Conclusions

This study examined the role of TDM policies in individual mode choice for work trips in the city of

Tehran Five policies including increasing parking cost increasing fuel cost cordon pricing in the odd-

even zone of the city transit time reduction and transit access improvement were investigated Of

the five policies the former three were push policies and the latter two were pull policies The

design of experiments approach was used to design the questionnaire that would capture the stated

preferences of car commuters

A number of nested structures based on recognizing differences in the variances associated

with unobserved influences were examined and finally the MNL as a superior model for this study

was developed

25

The results show that for car users who faced TDM policies all of the five studied TDM

policies were factors in the mode choice process The model shows that in considering car usage

push policies play a main role and pull policies do not

Results also show that the interactions of policies are also significant The model shows that

the interaction of parking cost policy and fuel cost policy is significant in prompting car users to

consider other modes It also shows that the interaction of cordon pricing and increase in fuel cost is

effective to discourage motorcycle usage Furthermore the interaction of transit time reduction

policy and access improvements policy is significant in considering DampR as a transportation mode to

workplace

Pull policies in the study were expected to attract individuals to transit modes (pull policies

of this study are related to transit modes) This was verified by the model

Results show that although taxi usage is usually treated as an alternative for urban trips it

was not directly affected by the studied policies This may be due to its special function in Tehran as

a non-public and non-private mode

The results also show that except for the interaction of cordon pricing and increase in fuel

cost other interaction effects have significant marginal effects on mode choice

Results of the model show that to implement a single policy cordon pricing is the most

effective in decreasing car usage

This study assessed workplace commuters Thus future studies should explore the

individuals mode choice with other trip aims Additionally including more policies especially pull

policies in favor of non-transit modes is reasonable to rigorously assess the model

It can be seen that the studied modes are affected differently by the TDM policies and their

interactions The impact of the interaction of policies appeared in the consideration of three out of

six studied modes and also in two out of three developed graphs This issue needs more attention

within the transportation demand management context and further studies should be performed

26

Design of this study is based on the consideration of all two-way interactions with some

degrees of deficiency in D-efficient policy Focusing on the significant interactions of this study is

recommended to perform other studies with more D-efficient values

Further understanding of the impacts of TDM policies might be provided in future research

by focusing on the amount of their synergy in prohibiting car usage

9 References

Arentze TA Hofman F amp Timmermans HJP (2004) Predicting multi-faceted activityndashtravel adjustmentstrategies in response to possible congestion pricing scenarios using an Internet-based stated adaptationexperiment Transport Policy 11 p31ndash41

Burris M amp Patil S (2009) Estimating the Benefits of Managed Lanes Technical Report Texas TexasTransportation Institute University Transportation Center for Mobility

Cao X amp Mokhtarian PL (2005) How do individuals adapt their personal travel A conceptual exploration ofthe consideation of travel-related strategies Transport Policy 12(3) pp199-206

de Palma A amp Lindsey R (2001) Transportation Supply and congestion In International Encyclopedia of theSocial and Behavioral Sciences 1st ed Elsevier p15882ndash15888

Eriksson L Garvill J amp Nordlund AM (2008) Acceptability of single and combined transport policy measuresThe importance of environmental and policy specific beliefs Transportation Research Part A 42 p1117ndash1128

Eriksson L Nordlund AM amp Garvill J (2010) Expected car use reduction in response to structural traveldemand management measures Transportation Research Part F 13 p329ndash342

Golob TF amp Hensher DA (2007) The Trip chaining Activity of Sydney Residents A Cross-Section Assessmentby Age Group with a Focus on Seniors Journal of Transport Geography 15 p298ndash312

Hensher DA amp King J (2001) Parking demand and responsiveness to supply pricing and location in theSydney central business district Transportation Research A 35 pp177-96

Hensher DA amp Rose JM (2007) Development of commuter and non-commuter mode choice models for theassessment of new public transport infrastructure projects A case study Transportation Research Part A41 p428ndash443

Hensher DA Rose JM amp Greene WH (2005) Applied Choice Analysis A Primer New York CambridgeUniversity Press

Iranian Center of Statistics (ICS) (2009) Information of Iranian States [Online] Available at HYPERLINK httpwwwamarorgirUploadModulesContentsasset16tehrantehparthtmlhttpwwwamarorgirUploadModulesContentsasset16tehrantehparthtml [Accessed 13 November2009]

Kingham S Dickinson J amp Copsey S (2001) Travelling to work Will people move out of their cars TransportPolicy 8 pp151-60

Kuhfeld WF (2009) Marketing Research Methods in SAS Experimental Design Choice Conjoint andGraphical Techniques SAS 92 Edition NC USA SAS Institute Inc Cary Available on the Webwwwsupportsascomtechsuptnotetnote_stathtmlmarket

Litman T (2003) The online TDM Encyclopedia mobility management information gateway Transport Policy10 pp245-49

Litman T (2010) Online TDM Encyclopedia [Online] Available at HYPERLINK httpwwwvtpiorghttpwwwvtpiorg [Accessed 21 Augest 2010]

25

The results show that for car users who faced TDM policies all of the five studied TDM

policies were factors in the mode choice process The model shows that in considering car usage

push policies play a main role and pull policies do not

Results also show that the interactions of policies are also significant The model shows that

the interaction of parking cost policy and fuel cost policy is significant in prompting car users to

consider other modes It also shows that the interaction of cordon pricing and increase in fuel cost is

effective to discourage motorcycle usage Furthermore the interaction of transit time reduction

policy and access improvements policy is significant in considering DampR as a transportation mode to

workplace

Pull policies in the study were expected to attract individuals to transit modes (pull policies

of this study are related to transit modes) This was verified by the model

Results show that although taxi usage is usually treated as an alternative for urban trips it

was not directly affected by the studied policies This may be due to its special function in Tehran as

a non-public and non-private mode

The results also show that except for the interaction of cordon pricing and increase in fuel

cost other interaction effects have significant marginal effects on mode choice

Results of the model show that to implement a single policy cordon pricing is the most

effective in decreasing car usage

This study assessed workplace commuters Thus future studies should explore the

individuals mode choice with other trip aims Additionally including more policies especially pull

policies in favor of non-transit modes is reasonable to rigorously assess the model

It can be seen that the studied modes are affected differently by the TDM policies and their

interactions The impact of the interaction of policies appeared in the consideration of three out of

six studied modes and also in two out of three developed graphs This issue needs more attention

within the transportation demand management context and further studies should be performed

26

Design of this study is based on the consideration of all two-way interactions with some

degrees of deficiency in D-efficient policy Focusing on the significant interactions of this study is

recommended to perform other studies with more D-efficient values

Further understanding of the impacts of TDM policies might be provided in future research

by focusing on the amount of their synergy in prohibiting car usage

9 References

Arentze TA Hofman F amp Timmermans HJP (2004) Predicting multi-faceted activityndashtravel adjustmentstrategies in response to possible congestion pricing scenarios using an Internet-based stated adaptationexperiment Transport Policy 11 p31ndash41

Burris M amp Patil S (2009) Estimating the Benefits of Managed Lanes Technical Report Texas TexasTransportation Institute University Transportation Center for Mobility

Cao X amp Mokhtarian PL (2005) How do individuals adapt their personal travel A conceptual exploration ofthe consideation of travel-related strategies Transport Policy 12(3) pp199-206

de Palma A amp Lindsey R (2001) Transportation Supply and congestion In International Encyclopedia of theSocial and Behavioral Sciences 1st ed Elsevier p15882ndash15888

Eriksson L Garvill J amp Nordlund AM (2008) Acceptability of single and combined transport policy measuresThe importance of environmental and policy specific beliefs Transportation Research Part A 42 p1117ndash1128

Eriksson L Nordlund AM amp Garvill J (2010) Expected car use reduction in response to structural traveldemand management measures Transportation Research Part F 13 p329ndash342

Golob TF amp Hensher DA (2007) The Trip chaining Activity of Sydney Residents A Cross-Section Assessmentby Age Group with a Focus on Seniors Journal of Transport Geography 15 p298ndash312

Hensher DA amp King J (2001) Parking demand and responsiveness to supply pricing and location in theSydney central business district Transportation Research A 35 pp177-96

Hensher DA amp Rose JM (2007) Development of commuter and non-commuter mode choice models for theassessment of new public transport infrastructure projects A case study Transportation Research Part A41 p428ndash443

Hensher DA Rose JM amp Greene WH (2005) Applied Choice Analysis A Primer New York CambridgeUniversity Press

Iranian Center of Statistics (ICS) (2009) Information of Iranian States [Online] Available at HYPERLINK httpwwwamarorgirUploadModulesContentsasset16tehrantehparthtmlhttpwwwamarorgirUploadModulesContentsasset16tehrantehparthtml [Accessed 13 November2009]

Kingham S Dickinson J amp Copsey S (2001) Travelling to work Will people move out of their cars TransportPolicy 8 pp151-60

Kuhfeld WF (2009) Marketing Research Methods in SAS Experimental Design Choice Conjoint andGraphical Techniques SAS 92 Edition NC USA SAS Institute Inc Cary Available on the Webwwwsupportsascomtechsuptnotetnote_stathtmlmarket

Litman T (2003) The online TDM Encyclopedia mobility management information gateway Transport Policy10 pp245-49

Litman T (2010) Online TDM Encyclopedia [Online] Available at HYPERLINK httpwwwvtpiorghttpwwwvtpiorg [Accessed 21 Augest 2010]

26

Design of this study is based on the consideration of all two-way interactions with some

degrees of deficiency in D-efficient policy Focusing on the significant interactions of this study is

recommended to perform other studies with more D-efficient values

Further understanding of the impacts of TDM policies might be provided in future research

by focusing on the amount of their synergy in prohibiting car usage

9 References

Arentze TA Hofman F amp Timmermans HJP (2004) Predicting multi-faceted activityndashtravel adjustmentstrategies in response to possible congestion pricing scenarios using an Internet-based stated adaptationexperiment Transport Policy 11 p31ndash41

Burris M amp Patil S (2009) Estimating the Benefits of Managed Lanes Technical Report Texas TexasTransportation Institute University Transportation Center for Mobility

Cao X amp Mokhtarian PL (2005) How do individuals adapt their personal travel A conceptual exploration ofthe consideation of travel-related strategies Transport Policy 12(3) pp199-206

de Palma A amp Lindsey R (2001) Transportation Supply and congestion In International Encyclopedia of theSocial and Behavioral Sciences 1st ed Elsevier p15882ndash15888

Eriksson L Garvill J amp Nordlund AM (2008) Acceptability of single and combined transport policy measuresThe importance of environmental and policy specific beliefs Transportation Research Part A 42 p1117ndash1128

Eriksson L Nordlund AM amp Garvill J (2010) Expected car use reduction in response to structural traveldemand management measures Transportation Research Part F 13 p329ndash342

Golob TF amp Hensher DA (2007) The Trip chaining Activity of Sydney Residents A Cross-Section Assessmentby Age Group with a Focus on Seniors Journal of Transport Geography 15 p298ndash312

Hensher DA amp King J (2001) Parking demand and responsiveness to supply pricing and location in theSydney central business district Transportation Research A 35 pp177-96

Hensher DA amp Rose JM (2007) Development of commuter and non-commuter mode choice models for theassessment of new public transport infrastructure projects A case study Transportation Research Part A41 p428ndash443

Hensher DA Rose JM amp Greene WH (2005) Applied Choice Analysis A Primer New York CambridgeUniversity Press

Iranian Center of Statistics (ICS) (2009) Information of Iranian States [Online] Available at HYPERLINK httpwwwamarorgirUploadModulesContentsasset16tehrantehparthtmlhttpwwwamarorgirUploadModulesContentsasset16tehrantehparthtml [Accessed 13 November2009]

Kingham S Dickinson J amp Copsey S (2001) Travelling to work Will people move out of their cars TransportPolicy 8 pp151-60

Kuhfeld WF (2009) Marketing Research Methods in SAS Experimental Design Choice Conjoint andGraphical Techniques SAS 92 Edition NC USA SAS Institute Inc Cary Available on the Webwwwsupportsascomtechsuptnotetnote_stathtmlmarket

Litman T (2003) The online TDM Encyclopedia mobility management information gateway Transport Policy10 pp245-49

Litman T (2010) Online TDM Encyclopedia [Online] Available at HYPERLINK httpwwwvtpiorghttpwwwvtpiorg [Accessed 21 Augest 2010]

27

Loukopoulos P (2005) Future urban sustainable mobility Implementing and understanding the impacts ofpolicies designed to reduce private automobile usage Doctoral dissertation Gothenburg Sweden GotegorgUniversity

Loukopoulos P et al (2004) Car-user response to travel demand management measures goal setting andchoice of adaptation alternatives Transportation Research Part D 9 pp263-80

Louviere JJ Hensher DA amp Swait J (2000) Stated Choice Methods Analysis and Application CambridgeUK Cambridge University Press

Louw E amp Maat K (1999) Enschede measures in a package Built Environment 25 pp118-28Mackett RL (2001) Policies to attract drivers out of their cars for short trips Transport Policy 8 pp295-306Marshall S amp Banister D (2000) Travel reduction strategies intentions and outcomes Transportation

Research Part A 34 pp321-38Marshall S Banister D amp McLellan A (1997) A strategic assessment of travel trends and travel reduction

strategies Innovation The European Journal of the Social Sciences 10 pp289-304May AD Jopson AF amp Matthews B (2003) Research challanges in urban transport policy Transport Policy

10 pp157-64May AD Kelly C amp Shepherd S (2006) The principles of integration in urban transport strategies Transport

Policy 13 p319ndash327Meyer MD (1999) Demand management as an element of transportation policy using carrots and sticks to

influence travel behavior Transportation Research A 33 pp575-99Mogas J Riera P amp Bennett J (2006) A comparison of contingent valuation and choice modeling with

second-order intaractions Journal of Forest Economics 12 pp5-30OFallon C Sullivan C amp Hensher DA (2004) Constraints Affecting Mode Choices by Morning Car

Commuters Transport Policy 11 p17ndash29Parkhurst G (2000) Influence of bus-based park and ride facilities on users_ car traffic Transport Policy 7(2)

p159ndash172Pendyala RM Kitamura R Chen C amp Pas EI (1997) An activity based micro-simulation analysis of

transportation control measures Transport Policy 4 pp183-92Rose JM amp Bliemer MCJ (2009) Constructing effcient stated choice experimental design Transport

Reviews 29(5) pp587-617Steg L amp Vlek C (1997) The role of problem awareness in willingness-to-change car use and in evaluating

relevant policy measures In Traffic and transport psychology Theory and application (Vaya TRampWC Ed)Amsterdam Pergamon pp465-75

Stradling SG Meadow ML amp Beatty S (2000) Helping drivers out of their cars Integrating transport policyand social psychology for sustainable change Transport Policy 7 pp207-15

Thorpe N Hills P amp Jaensirisak S (2000) Public attitudes to TDM measures a comparative study TransportPolicy 7 pp243-57

Vieira J Moura F amp Viegas JM (2007) Transport policy and environmental impacts The importance ofmulti-instrumentality in policy integration Transport Policy 14 p421ndash432

Visser J amp Van der Mede P (1986) The effects of parking measures on traffic congestion In 1986 PTRCSummer Annual Meeting Brighton England 1986

Washbrook K Haider W amp Jaccard M (2006) Estimating commuter mode choice Adiscrete choice analysisof the impact of road pricing and parking charges Transportation 33 pp621-39

Zareii H (2003) Evaluation of Transportation demand management measures effectiveness in the critical airpollution days MSc Thesis Tehran Sharif University of Technology