data needs for innovative modeling …onlinepubs.trb.org/.../nhts/workshop-dataneeds.pdfdraft...

Draft Workshop Resource Paper Submitted for Pre-Conference Review – Do not use for any other purpose

1

DATA NEEDS FOR INNOVATIVE MODELING WORKSHOP RESOURCE PAPER

National Household Travel Survey Conference Understanding our Nation’s Travel

Washington D.C., November 1-2, 2004 Konstadinos G. Goulias, University of California, Santa Barbara Mark Bradley, Bradley Research & Consulting Val Noronha, University of California, Santa Barbara Reg Golledge, University of California, Santa Barbara Peter S. Vovsha, PB Consult, Inc.

1. Introduction Many recent contributions to modeling travel behavior offer unique opportunities to expand the traditional modeling and simulation envelope and encompass a wide variety of dimensions of trip makers’ everyday life. Key aspects include behavioral dynamics of life cycle stage transitions, repetition and cycles in time allocation and travel over week long periods, day-to-day variation in activity participation and travel, and considerable progress in modeling episode duration of activity participation and travel as well as departure time choice and associated decision making. In this conference another resource paper (Pendyala and Bhat, 2004) provides a comprehensive overview of these new and (re)discovered directions of data collection, modeling, and behavioral simulation. Many examples of these more recently developed methods and models, however, do not address a few fundamental concerns of travel behavior resulting in data requirements that are limited. The scope of our paper and of the innovative modeling workshop is to take us further into the future of modeling and simulation to examine aspects of innovative modeling that go beyond traditional thinking. One intermediate objective is to challenge strategically selected tenets in modeling. The ultimate objective is to offer ideas for more advanced data collection strategies better suited for today’s needs in policy analysis. We also offer a few key points of discussion to determine specific directions in reforming and redesigning NHTS. As explained later in the paper and given the institutional barriers to redesigning NHTS we should at least expect openness in collecting data about behavioral aspects that are neglected and they are fundamental in addressing policy issues and designing policy actions that work. These new policy questions, as Garrett and Wachs, 1996, discuss in the context of a lawsuit against a regional planning agency in the Bay Area are increasingly outpacing data collection, modeling and simulation in practice creating yet another widening gap between policy and practice that parallels the gap between research and practice. Data collection is one area that helps to bridge these gaps.


2

Independently of the urgency and timeliness of data collection, modeling and simulation, one fundamental difference with past research initiatives is a widespread recognition of the need to also build “integrated” models. As a result forecasting models, in addition to long-term land use trends and air quality impacts, also address issues related to technology use and information provision to travelers in the short and medium terms. Similarly, policies are also targeting issues such as: increasing citizen participation, jurisdictional integration, decentralization, deregulation, privatization, environmental concerns, mobility costs, congestion management by population segments, and private infrastructure finance (van der Hoorn, 1997). These new policy initiatives place more complex issues in the domain of regional and national policy analysis and forecasting. They also amplify the need for methods that produce forecasts at the individual traveler and her/his household levels in addition to the traffic analysis zone or larger geographical area levels In addition to the long range planning activities and the typical traffic management activities, analysts and researchers in planning need to also evaluate the following:

• Travel demand and supply impacts of new technologies (e.g. ubiquitous mobile voice and data transmission),

• Traveler and transportation system manager information provision and use (e.g., on-board and off-board traveler information systems)

• Pricing and financing strategies (e.g., congestion pricing and tolling) • Combinations of transportation management actions and their impacts

(e.g, parking fee structures and city center restrictions, and • Assessment of combinations of environmental policy actions (e.g., carbon

taxes and information campaigns). The tools to perform all this need to also have forecasting capabilities that are more accurate and detailed in space and time (e.g., we are moving toward parcel by parcel analysis and separate analyses for different seasons of a year and days of the week to capture seasonal and within a week variations of travel). Echoing all this and in the context of the Dutch reality, Borgers, Hofman, and Timmermans (1997) have identified five information need domains the new envisioned policy analysis models will need to address and they are (in a modified format from the original list):

a) social and demographic trends that may produce a structural shift in the relationship between places and time allocation by individuals invalidating existing travel behavior model systems;

b) increasing scheduling and location flexibility and degrees of freedom for individuals in conducting their every day business leading to the need to consider additional choices (e.g., departure time from home, work at home, shopping by the internet, shifting activities to the weekend) in modeling travel behavior;


3

c) changing quality and price of transport modes based on market dynamics and not on external to the travel behavior policies (e.g., the effect of deregulation in public transport);

d) shifting of attitudes and potential cycles in the population outlook about modes; and

e) changing scales/jurisdictions (scale is the original term used to signify the different jurisdictions) – different policy actions in different sectors have direct and indirect effects on transportation and different policy actions in transportation have direct and indirect effects in the other sectors (typical example in the US is the welfare to work program).

In the next sections we define the data needs and model attributes required to design major improvements over the more traditional Systems Engineering simulation models. Each section defines a few key parameters for an improved model and its data needs that contain both the flexibility and the potential to fill the information needs of today’s policy initiatives. The review is not exhaustive, instead, it uses four aspects as case studies of gaps that showcase the need for change in data collection before NHTS and related initiatives become obsolete.

2. Defining the Background for Innovative Modeling As mentioned above an emerging framework in travel behavior loosely defined as dynamic and activity-based expanded the data collection to include time allocation in space and time. Planning and modeling have achieved tremendous progress toward a comprehensive approach to, in essence, build simulated worlds on computer enabling the study of policy scenarios. The emerging framework, although rich in its potential for scientific discovery, contains many gaps and it is incomplete. In this section, we first review four aspects that are: wayfinding and spatial cognition, behavioral processes and related decision making, telegeoinformatics, and sociodynamics. All four aspects are neglected in national transportation data collection activities and addressing them propose considerable challenges. Before moving to the review of these innovations it is worth mentioning the four dimensions emerging from policies and models. The first is the geographic space and its conditional continuity, the second is the temporal scale and calendar continuity, the third is interconnectedness of jurisdictions, and the fourth and most important is the set of relationships in social space for individuals and their communities. The first dimension, geographic space here is intended as the physical space in which human action occurs. This dimension has played important roles in transportation planning and modeling because the first preoccupation of the transportation system designers has been to move persons from one location to another. Initial applications considered the territory divided into large areas (traffic analysis zones), represented by a virtual center of gravity


4

(centroid), and connected by facilities (higher level highways). The centroids were connected to the higher level facilities using a virtual connector summarizing the characteristics of all the local roads within the zone. As computational power increased and the types of policies/strategies required higher resolution, the zone became smaller and smaller. Today, it is not unreasonable to expect software to handle zones that are as small as a parcel of land and transportation facilities that are as low in the hierarchy as a local road (the centroid becomes the housing unit and the centroid connector the driveway of the unit). As we will see later in this paper in analyzing behavior we are interested in understanding human action in many facets. For this reason in some applications geographic space needs to consider more than just physical features (Golledge and Stimpson, 1997, page 387) moving us into the notion of place and social space. The second dimension is time that is intended here as continuity of time, irreversibility of the temporal path, and the associated artificiality of the time period considered in many models. For example, models used in long range planning applications use typical days (e.g., a Summer day for air pollution). In many regional long-range models the unspoken assumption is that we target a typical work weekday in developing models to assess policies. Households and their members do not obey this strict definition of a typical weekday. They schedule activities following very different decision making horizons in allocating time spreading activities among many days including weekends, substituting out of home with in home activities on some days but doing exactly the opposite on others, allocate tasks within a household depending of emerging needs by adjusting and adapting to their environments, and use telecommunications only selectively. Obviously, taking into account these scheduling activities is by far more complex than allowed in existing transportation planning models. The third dimension is jurisdictions and their interconnectedness. The actions of each person are “regulated” by jurisdictions with different and overlapping domains such as federal agencies, state agencies, regional authorities, municipal governments, neighborhood associations, trade associations and societies, religious groups, and formal and informal networks of families and friends. In fact, the federal government defines many rules and regulations on environmental protection. These may end up being enforced by a local jurisdiction (e.g., a regional office of an agency within a city). On one hand, we have an organized way of governance that clearly defines jurisdictions and policy domains (e.g., tax collection in the US). On the other hand, however, the relationships among jurisdictions and decision making about allocation of resources does not follow always this orderly governance principle of hierarchy. A somewhat different and more “bottom up” relationship is found in the social network and for this reason requires a different dimension. The fourth and final dimension is social space and the relationships among persons within this space. For example, individuals from the same household


5

living in a neighborhood may change their daily time allocation patterns and location visits to accommodate and/or take advantage of changes in the neighborhood such as elimination of traffic and the creation of pedestrian zones. Depending on the effects of these changes on the pedestrian network we may also see a shift in the within the neighborhood social behavior. In contrast, increase in traffic to surrounding places may create outcry by other surrounding neighborhoods, thus, complicating the relationships among the residents. One important domain and entity within this social space is the household. This has been a very popular unit of analysis in transportation planning recognizing that strong relationships within a household can be used to capture behavioral variation (e.g., the simplest method is to use a household’s characteristics as explanatory variables in a regression model of travel behavior). In this way any changes in the household’s characteristics (e.g., change in the composition due to birth, death, or children leaving the nest or adults moving into the household) can be used to predict changes in travel behavior. New model systems, created to study this interaction within a household, are looking at the patterns of using time in a day and the changes across days and years. It is therefore very important in modeling and simulation as well as other types of policy analysis to incorporate in the models used for policy analysis not only the interactions described above but also interactions among these four fundamental dimensions. A typical example in long range planning is a simulation model that uses larger geographical areas (region, states, and countries) and addresses issues with horizons from 10 to 50 years. In many instances we may find that large geographic scale means also longer time frames applied to wider mosaics of social entities and including more diverse jurisdictions. On the other side of the spectrum issues that are relevant to smaller geographic scales are most likely to be accompanied by shorter term time frames applied to a few social entities that are relatively homogeneous and subject to the rule of a very few jurisdictions. This is one important organizing principle but also an indicator of the complex relationships we attempt to recreate in our computerized models for decision support. In developing the blueprints of these models one can choose from a variety of theories (e.g., neoclassical microeconomics) and conceptual representations of the real world that help us develop these models. At the heart of our understanding of how the world (as an organization, a household, or an individual human being) works are models of decision making and conceptual representations of relationships among entities making up this world. The next four sections illustrate exactly this in terms of human-environment interactions.

2.1 Human wayfinding and spatial cognition Finding one’s way to a novel or a familiar destination involves recognition of that destination when approached, selection of a mode of travel, developing a travel plan, and moving along a specific path or route. When the path is defined prior to initiating movement, the traveler is said to “navigate” along the path, following


6

a predetermined sequence of turns and route segments. If no specific path is determined a priori, then a traveler is said to be “wayfinding.” In this case, en-route decisions are made regarding which turns to take and which sequence of segments to traverse. When undertaking trips, a process of “piloting” or travel from landmark to landmark often occurs. Alternatively, a traveler might simply head out in the general direction of a destination relying on general knowledge of the layout of the extant geospatial environment to help find a successful route. Such travel may incorporate shortcuts or may be influenced by a variety of route selection criteria. In the former case, a process known as “dead reckoning” (or travel by a “homing vector” or “path integration”) may occur. Here the traveler constantly undertakes spatial updating of current location with regard to a home base or known destination. In the latter case, a specific trip may be chosen according to strategies such as minimal distance or time, least cost, minimizing turns, maximizing or minimizing traffic control apparatus such as traffic lights and stop signs, longest (or shortest) leg first, aesthetic quality, and so on. Thus, while a navigator would lay down a path to be followed using perhaps only one criteria (e.g., shortest path), a wayfinder may use a mix of strategies for different parts of a trip, or use one criterion on an outbound trip and another on an inbound trip. Environments are generally experienced either by travel or via a representation (e.g., map, sketch, video, movie, written description, set of slides). Both methods may achieve successful travel. But either can also produce misinformation or misunderstanding. Such effects are usually the result of the cognitive processes used in the travel activity. For example, changing route selection criteria on inbound and outbound trips may produce the frequently identified result that trip AB is perceived to be different from trip BA (i.e., shorter, longer, slower, faster, etc.). This effect can also be achieved given the physical configuration of the trip where, say, AB is uphill and BA is downhill. Trip making is bound up with signage. Signs can be deliberately manufactured (as in highway mileage signs) or can be idiosyncratic (as in the selection of specific environmental features that cue behavior). Practically speaking, most humans today are completely dependent on the signage they get from maps, from in-vehicle navigation systems from GPS readings, from AAA trip-tiks, or from commonly or idiosyncratically perceived landmarks. Determining what features and uses are important for travel has been assisted by research in spatial cognition. Cognitive processes include learning, thinking, memorizing, recalling, and internally manipulating sensed information. In other words, environmental information is sensed, coded, and stored in the brain wherein it is then recalled, manipulated in working memory, and used to determine behaviors. Few if any of most human’s movements are random. It thus becomes a significant practice to find the how, when, where, and why of travel.


7

Spatial cognition is that subset of cognition that deals with the spatial domain. In particular, emphasis has been placed on the cognitive mapping process (encoding), the cognitive map itself (the spatial information stored in the brain), and the recall and use process (decoding). Spatial cognition is sometimes referred to as an “internal geographic information system” since the processes, the functionalities, and the spatial products (i.e., external representations of stored information) are seen to be similar to those folded into GIS software. Discoveries and Current Understandings

1. Destination Choice and Route Selection are based on people’s cognitions of the environment and their preferences for travel mode and trip time.

2. People tend to perceptually overestimate shorter distances and underestimate longer distances.

3. Many trips are perceptually asymmetric—i.e., AB is perceived to have different lengths, travel time, or ease of movement than BA .

4. People’s environmental awareness is incomplete, fuzzy, and distorted. 5. People have difficulty integrating individual paths to make a network.

Network knowledge is sparse and error-ridden when compared to the real world.

6. Even without perfect environmental knowledge, successful trip making and goal achievement can occur, but evaluative criteria of individual travelers are usually those of “satisficing” rather than optimizing.

7. Criteria for route selection such as “shortest path” are important for commercial vehicles (e.g., trucks, express delivery, postal service, taxis) but do not dominate among individual travelers.

Addressing these aspects in data collection for behavioral models, however, also requires measurement of movements in time and space. For this reason the next section offers an overview of current state of the art in spatio-temporal measurement.


8

2.2 Measurement in time and space This section briefly reviews current technologies and future prospects of gathering high resolution time-space data to support advanced activity and related modeling. It begins with an introduction to telegeoinformatics, describes current technologies for tracking and other methods of behavioral data gathering, suggests technologies that are likely to become available in the future, and identifies challenges that lie ahead in realizing this vision. Telegeoinformatics Telegeoinformatics may be defined as the management (gathering, storage and analysis) and sharing of space-time couplings ( t,Λ ), where Λ is a location expression such as a street address or (x, y, [z]) coordinate set, representing the location of a person or object at time t. For a given object, this coupling over several observations forms a track, with routes, stops and delays; the aggregation of several tracks constitutes a data set documenting O-D tables and traffic density. Individual or multiple bundled tracks may be visualized in time-space prisms (Hagerstrand 1970, Miller 1991, Kwan 1998) or subjected to a variety of analyses addressing for example the mobility domain of individuals or groups, or averaging tracks to enhance map accuracy (Goodchild et al 1995, Noronha 2004). Attributes may be attached to any coupling, describing traffic congestion or environmental hazards such as black ice or fog; in this context the tracks could serve as space-time sampling probes of the physical and cultural environment. At the application end, a variety of spatial planning questions are addressed by travel data. Traditionally these have been public sector urban and transportation planning, and environmental impact studies. With GIS increasingly applied at finer scales in a broadening variety of fields, the potential for utilization of travel data is rapidly expanding. In the homeland security arena, there is a need to manage large unpredicted flows, e.g. consequent to an unforeseen event in a congested area. In commercial planning, retail location studies are shifting from static demographic calculations to dynamic travel activity data to predict store preference and shopper behavior. Potential technologies to measure behavior now The dominant technology for space-time data capture (more simply called tracking) is the Global Positioning System, GPS. The technology has experienced significant breakthroughs in price, accuracy and miniaturization over the last 20 years. The success story is well reflected in the class of objects tracked by GPS: from military vehicles in the 1970s to commercial trucks in the 1980s, private automobiles in the 1990s, and pedestrians in the current decade.


9

GPS is significant because more than any other technology, it has changed the spatial resolution of travel studies from the TAZ to the household, and the temporal resolution from the quarter-hour to the second. GPS is not yet as ubiquitous as say the wrist-watch, and the decision whether to carry one and to disclose data from the unit rests with the individual. This is a potential source of sampling bias (similar to the bias that afflicts traditional diary-based household travel surveys). There is often a need for more universal periodic sampling, which can take several forms of resolution in time and space, such as observing anonymous objects at a specific place over a continuing period of time (e.g. a loop detector or traffic camera), or a snapshot observation over a constantly changing spatial viewframe (imagery from mobile platforms such as helicopters and satellites), or tagging individual identifiable objects passively and involuntarily. The last item may seem in conflict with privacy rules, but the situation blurs when individuals give consent in return for a benefit. This issue is treated in detail below. Beyond the travel study arena, two rapidly growing areas of technological interest are prompting improvements in tracking. One is Location Based Services (LBS) and personal communications systems, which would for example allow a cell phone user to receive a text message offering a “10-minute special” while entering the vicinity of a fast food outlet or gadget store. The other is Intelligent Transportation Systems (ITS). In both cases, commercial possibilities motivate the tracking of individuals and vehicles. Private sector entities therefore promote technological innovation in location technology and are also likely to facilitate establishment of the communications infrastructure. One current challenge in the case of LBS is to track individuals indoors in shopping malls, where GPS is of limited use. Supplemental positioning technologies exist, from inertial measurement units (IMUs) to wireless proximity-based or ranging methods (e.g. PointLink 2002). In the U.S., cell phones are being required by the Federal Communications Commission to be capable of reporting their location correct to 100 m, using either GPS or ranging and triangulation from signal strength (FCC 1997). Cell phone location can also be determined by analyzing multi-path signatures (Mudge 2001). Motion detectors have recently been incorporated into cell phones (BBC 2004), primarily as a “cool” technological treat, but there is obvious potential for these to improve the devices’ self-locating ability in urban canyons, tunnels and indoors. Technologies of the future In Intelligent Transportation Systems (ITS), the Vehicle Infrastructure Integration (VII) initiative foresees automobiles in frequent or even constant two-way communication with Traffic Management Centers (TMCs) and thereby with other vehicles, via 5.9 MHz Dedicated Short-Range Communications (DSRC) roadside beacons. In the short term TMCs are expected to relay information to the vehicle


10

on congestion and signage; in the longer term, environment sensors in the vehicles will feed back data on black ice, fog and other road hazards. They will warn following vehicles when brakes are being applied, report problems with the vehicle’s mechanical systems, and summon emergency services if the driver becomes disabled. Further into the future, customized entertainment such as movies could be delivered to moving vehicles, and vehicles will eventually drive themselves using autonomous navigation supported by real time data on surrounding traffic (Wilson 2004). Note that this will involve location data at decimeter resolution, and decisecond frequency. In these futuristic scenarios, instantaneous space-time-stamped data are merely a footnote in a much broader system of data exchange. One could reasonably speculate that space-time data will become commercially available to select buyers, in the same way as credit records and personal profile data are now. Indeed, data volumes will be overwhelming and there will undoubtedly be a need for selectivity in the capture and transfer of data based on utility of content. While there will inevitably be changes in travel patterns resulting from the technological evolution, the future seems assured for data availability for travel surveys monitoring these changes. Unresolved issues A number of hurdles stand between the present and the future scenarios presented above. Currently spatial accuracy is a concern. In the longer term, cost, institutional barriers and privacy require special attention. There are two components to spatial accuracy: (a) the accuracy of the location sensor, and (b) the accuracy of the reference map. Map matching is the process of inferring a location on a map, given an observation from a mobile location sensor. If the map is accurate, map matching can improve the quality of the location observation. However, if the location sensor is inaccurate to the tune of 50 m, compounded by map inaccuracy of that order, an object may easily be placed on the wrong map segment and “correction” by map matching compounds the problem rather than solving it. Noronha and Goodchild (1999) report positional errors up to 200 m in commercial digital maps. Leading digital map offerings have improved considerably since that study, and the likelihood of mismatch is now much lower, but still an issue. Closely related to data quality is cost. GPS has become a popular consumer item over the last two years, available at sporting goods and office supply stores. Some of the lower-priced offerings employ low-precision processors and consequently deliver poorer quality data (based on informal experiments by the authors). Cost is also a significant impediment in installing the communications infrastructure for telegeoinformatics, which will probably require a combination of private sector funding in urban areas (where marketing possibilities are best) and public funding in remote areas (where the primary benefit is improved safety).


11

To the extent that the private sector leads the establishment of the infrastructure, data ownership issues arise. Currently we have no evidence that ATM and retail transactions are being reconstructed into personal time-space logs (except in criminal investigation), perhaps because data are fragmented among multiple custodians, suggesting that demand for integrating such data has not yet developed sufficient momentum to overcome the cost of doing so. This barrier will have to be crossed. Finally there is the issue of privacy. Currently, academic research involving Human Subjects is subject to strict controls to protect the rights of subjects (DHHS 2001). However, when details of the study are clarified to the subject in advance and the individual’s consent is secured, the researcher has latitude to investigate. Similarly, commercial credit databases flourish because the concern over privacy is eclipsed by the participant’s interest in the benefits of the deal; the individual therefore grants consent for release of financial information in return for the convenience of obtaining a credit card, and internet “cookies” are accepted in return for the convenience of online banking. When travel behavior data reaches the point of being a relatively small component in a wider, commercially motivated system of significant material benefits in return for information exchange, privacy concerns are likely to be overcome more easily. This does not trivialize the privacy issue or the real possibility of misuse of data.


12

2.3 Behavioral processes as opposed to outcomes only Travel surveys typically collect information on the outcomes of decision processes. Trip/activity diary surveys such as NTPS, for example, tell us about the locations people decide to visit, the things they decide to do there, and the modes they decide to use to get there. Such surveys tell us little or nothing about how the people came to those decisions. In addition to outcome data, types of information that might be useful, depending on the decision context, include:

• What other alternatives did a person consider? • What other alternatives were possible? • Was the decision planned in advance or made on the spur of the

moment? • If planned in advance, how far in advance, and did the plans change over

time? • Was the decision dependent on other decisions that were made? • Which decision factors were most important? • What information did the person have regarding those factors? • How and when did the person acquire that information? • What other information would have been useful? • Why did the person not have that information? • How would the person go about getting that information? • Had the person made the particular decision before? • If so, did the person tend to make the same choice each time or did it

vary? • If it varied, why would the person choose differently at different times? • Was the decision made jointly with other(s)? • If so, how did the different people enter into the decision? • Is so, was there a negotiation or a tradeoff of priorities? • Did other people (employers, etc.) indirectly influence the decision? • If so, what was their influence? • How did past experiences influence expectations regarding the decision? • Were there any recent major changes that influenced the decision? • Were there any expected future changes that influenced the decision? • What roles did variability, uncertainty and risk play? • Did the person have any strong attitudes about the choice alternatives? • If so, when and how did those attitudes develop? • Does the person have any strong unconscious feelings or associations

regarding the alternatives? • What are a persons’ attitudes and habits regarding health and exercise,

and how do these relate to observed travel and location choices?


13

• What are a person’s attitudes and preferences toward various aspects of neighborhood land use and housing, and how do these relate to past and present location and travel decisions?

Figure 1 provides a contrast of analytic methods based on outcome data only versus those based on data on both decision processes and outcomes. At the left are the key types of data and modeling assumptions used in methods based on choice outcomes. At the right are the additional types of data and assumptions that can be tested using survey and modeling methods based on both outcomes and processes. Note that these methods can still include any or all aspects of traditional models—there is no intention to reinvent the wheel. What makes process-based methods more “rounded” is their ability to simultaneously include or test other non-traditional aspects. The key differences are first introduced below, and then illustrated by means of some examples: The survey choice context: Outcome data are typically choices made in actual reported situations (revealed preferences) or else choice made in hypothetical situations (stated preferences) designed to mimic actual situations as closely as possible. Data on processes can also come from such survey contexts, or can come from more involved contexts such as longitudinal monitoring surveys or simulated choice environments. Deduction and induction: When working with outcome data, we typically deduce any factors that influenced the reported outcomes, in order to keep the models as objective as possible for forecasting. With stated preference surveys, the important factors are assumed beforehand and presented to respondents. Data on processes can also include respondents’ self-explication of aspects of their own behavior. Compensatory and non-compensatory choices: Analyses based on outcome data typically assume rational, compensatory utility maximization, and very rarely collect supplementary data that would be of use in identifying or confirming other types of decision processes. Surveys that also focus on process can ask supplementary questions to identify how actual or simulated choices were made. Dynamics and equilibrium: Except for dynamic models based on panel data, which have been few and far between, models based on outcome data are cross-sectional and assume that supply and demand are in equilibrium. Additional survey data on choice processes could add dynamic information into cross-sectional surveys (e.g. retrospective data), or could provide supplementary information within a panel survey in order to help explain the observe changes over time. Learning and information: Typical analyses based on outcome data assume that people have complete information and accurate perceptions about the


14

choice alternatives. Specific data about previous knowledge and perceptions would allow the use of more realistic assumptions. Quantitative and qualitative variables: Because outcome-based methods rely mainly on objective measures of choice factors, they tend to focus on the variables that are easiest to quantify. Other variables such as reliability, safety, comfort, etc. cannot be included unless we also have data on peoples’ perceptions of those variables on some type of qualitative or quantitative scale.

Figure 1: Schematic Comparison of Outcome-Based and Process-Based Methods

Can Consider

Learning and

Cognition

(Can) Assume Full, Accurate Information

Qualitative Variables and Measures

Quantitative

Variables and

Measures

Can Consider Dynamic State-Dependence (Can) Assume

Behaviour in

Equilibrium

Can Consider Non-compensatory Decision Rules

Models of

Compensatory

Utility Maximisation

Self-explication of

Behaviour

Deductive Models

of Behaviour

Simulated Choice Environments

Reported Choice

Situations

Methods

Based on

Outcomes and

Processes

Methods Based

on Outcomes


15

Information on behavioral processes should not be viewed as a substitute for the revealed preference and stated preference data we typically collect. In fact, it can be a complement to help us interpret the data on observed (or hypothetical) behavior. Just as revealed preference information is often collected in stated preference surveys, process data may be collected in revealed preference and stated preference surveys. The current relationship between SP and RP data collection provides a useful illustration: sometimes RP data is collected as an initial basis for customization in an SP survey, and sometimes SP data is collected as a follow-up survey for a subsample of an RP survey, but the two types of data are often collected and analyzed in a coordinated manner. Bradley (2004), defined process data as a combination of quantitative and qualitative information, collected systematically to reveal individual travel choice processes over time, noting that such data would be useful both in expanding the scope of knowledge for making policy decisions and in expanding the scope of models of household travel behavior. Many of the types of additional information on choice processes identified in that paper and in the discussion above can be collected as supplemental information during otherwise conventional RP and SP surveys. Further ideas along those lines are presented later in this paper (section ? (Vovsha’s section)). Further innovation may be possible, however, by using less conventional survey approaches. Bradley (2004) lists three types of approaches that meet all or most of the following criteria:

• Use a structured, systematic survey process. • Collect both quantitative and qualitative information. • Collect data on both decision outcomes and processes. • Collect data on both objective and subjective choice factors. • Become more feasible and/or attractive with recent advances in survey

technologies. “Intelligent Travel and Activity Diaries” One of the first approaches with these characteristics to be used in travel demand research was the ‘situational’ survey approach, reported by Brog (1982) and by Goulias, et al. (1998). This approach begins with a description of actual behaviour, similar to traditional household travel surveys, but then uses the reported behaviour as a basis for asking more in-depth, interactive questions that elicit perceptions and other subjective factors. Many applications of this technique have suggested that individual-specific subjective factors are much more important in driving choice processes than is generally realized. On the basis of such findings, Brog has developed the ‘individualised marketing’ approach, described in John and Brog (2001). Some of the pioneering research into the dynamic processes behind household travel behaviour was carried out at the Transport Studies Unit at Oxford in the 1970’s and 80’s (see Jones, et al., 1983). One particular outcome of that


16

research was The Household Activity and Travel Simulator (HATS) (Jones, 1979). This innovative survey method uses data from a typical travel and activity diary survey, and lays it out on a schematic color-coded board showing each household member’s trips and activities on a timeline. Important constraints can also be shown schematically. The household members are then asked how they would adjust their activity schedules as the result of specific changes such as shifts in work hours, parking charges, retail hours, transit services, road pricing, etc. With the advent of portable laptop computers for household interviews, a computerised version of this method (ATAQ) was created and tested in Adelaide, Australia (Jones, et al., 1987). A key feature of the HATS approach is that, like the best customised SP research, it uses a hypothetical framework structured around observed choices. Any stated changes can be recorded in this same framework, making the resulting data amenable for quantitative analysis. In contrast to most SP research, however, the survey structure is open-ended enough to allow for collection of a variety of additional data on the process that respondents go through while deciding on changes to their activity patterns, including which alternatives are considered and then rejected. Unfortunately, all of the surveys done this approach have been quite small-scale and exploratory in nature, so that its wider potential in providing new types of data for modeling has never been fully tested. It has, however, inspired a number of related survey approaches, including the CUPIG survey method which is focused more specifically on the allocation and use of cars within households (Lee-Gosselin, 1990). As personal computers have become lighter and less expensive, it has become possible to put hardware in respondents’ homes and have them complete computerised versions of travel and activity diaries, as opposed to completing diaries by hand. One example of this approach is the SMASH method from Eindhoven University (Ettema, et al. 1995), one of the data sources used to create the ALBATROSS activity-based model system. Another example is the CHASE method (Doherty and Miller, 2000). In these methods, the respondent can fill in planned activities before the diary days, and then adjust those plans over time. When the respondent eventually fills in the actual schedule for the diary day, some of the travel and activities will have been pre-planned and done according to plan, some will have been pre-planned but adjusted at the last moment, while others will not have been pre-planned at all. The computer algorithm is clever enough to identify each type of situation and ask appropriate questions regarding the decision process that led to the schedule adjustment, new trip/activity or cancelled trip/activity. From this type of data, Doherty (2003) was able to show that our typical classification of activities by purpose is often not adequate to predict which activities will be given priority in the activity scheduling process. On the basis of such analyses, we may be able to identify a few key questions regarding timing and location flexibility that should be added to the telephone retrieval stage of activity diary surveys.


17

“Information Acceleration” Another important example is the Information Acceleration (IA) approach, developed at the MIT Sloan School of Management. Compared to most SP-type approaches, this method moves more fully from a hypothetical choice environment to a simulated choice environment. The emphasis is on the information search process in contexts when people are faced with new and unfamiliar choice alternatives. In order to investigate the purchase decision process for electric vehicles (Urban, et al., 1995), respondents were allowed to request various types of information about the vehicles, including mock-ups of television commercials, newspaper and magazine reviews, word-of-mouth opinions, and dealer sales pitches. A test ride in a driving simulator was also available. By systematically varying the content of each of these simulated information sources, the importance of each piece of information in determining the simulated purchase could be measured. The information search process itself could also be recorded and analysed, a possibility lacking in stated preference experiments where the information given to each respondent is pre-determined. Another interesting application of IA was carried out by Walker (1994) to look at the influence of automated traveler information systems (ATIS) on route choice and departure time decisions. Information about highway congestion levels and travel times was available to respondents in the form of mock-ups of real-time telephone messages and in-vehicle message. General information about the ATIS system and how to use it was available as mock-ups of newspaper articles, printed brochures, and word-of-mouth. Perceptions and intended decisions were measured both before and after the information search process. Then, the respondent entered a travel simulator, in which he or she experienced the ‘actual’ simulated congestion and travel times resulting from the decision. In contrast to infrequent decisions such as vehicle purchases, traveler information systems can be used on a repeated basis. So, in the ATIS context, it was possible to apply the IA approach in an iterative framework and study the learning process by measuring post-travel changes in perceptions and attitudes, as well as their influence on the next round of information search and decision-making. The IA approach requires a large investment of time from both the survey designers and the respondents. Due to its expense and its usefulness in the context of new product development, the method has been used mostly for corporate market research, where most of the results have been kept proprietary. Nevertheless, Urban, et al. (1996) were able to provide an overview of various applications, as well as the results of tests of the validity of the quantitative results. This is one of the few survey techniques that has been able to provide qualitative insights about dynamic search and decision processes while also providing useful quantitative forecasting models. As computer and Internet technology for creating simulated choice environments becomes easier and


18

easier to use, the feasibility and affordability of this type of survey method should continue to improve. “Intensive Market Monitoring” Another area of growing interest is in studies that look specifically at the effect of marketing policies to influence travel behaviour. As opposed to ‘hard’ policies to change infrastructure or service levels, these studies concentrate on ‘soft’ policies that use information and marketing techniques to influence peoples’ awareness and perceptions of the existing options. Jones and Sloman (2003) describe a pair of European Commission projects to look at such policies. The INPHORMM project uses a conceptualization of behavioural change based on a process of ‘five A’s’: (1) awareness of a problem, (2) acceptance of a need for change, (3) a change in attitudes toward choice alternatives, (4) action to initiate a change, and (5) assimilation of this new behaviour into everyday life. A follow-on project, TAPESTRY, extends the framework to look more at longer-term dynamics. The last three ‘A’s’ are divided into separate elements highlighting the dynamic components: (a) change in attitudes becomes change in perceptions and (re-)evaluation of the options, (b) action to initiate a change becomes making a new decision and then trying out the new decision in terms of experimental behaviour, and (c) assimilation of the new behaviour becomes longer term adoption of habitual behavior, as well as feeding back as learning to influence awareness, acceptance and attitudes. After developing a useful conceptual framework, the EC studies depended on various site-specific monitoring studies and opinion and attitude surveys to gauge the effects of specific policy measures. When attempting to translate these results solely to the UK context, Jones and Sloman raise some crucial issues regarding the results: Are they transferable across regions? Are the policies synergistic or redundant when applied in combination? The authors conclude that, to answer such questions, “our conceptual models of travel behaviour need to be expanded to recognize more fully these various subjective elements of travel decision making”, and that “one of the key limitations from a research perspective is lack of data. … In particular, stated preference exercises need to be more sophisticated, in at least two respects: in the treatment of information deficiency and uncertainty, …, and in their recognition of respondent’s interest in, or willingness to consider, a change in behaviour”. The IA approach and the HATS/ATAQ/CATS approach described above are extensions of SP survey methods in these directions, while the SMASH/CHASE family of approaches forms the basis of a similar extension of RP survey methods. Of course, whenever possible it is desirable to use an actual choice environment rather than a simulated one. An example of such a study is described by Sheehan (1999). In this study, a “quasi-longitudinal survey was administered over a four month period to assess dynamics in an individual’s learning and valuing response to the CarLink car sharing innovation over time”. Different


19

educational media were distributed to respondents at three points in time, with an identical questionnaire on perceptions and attitudes completed at recruitment and after each of the three educational media, followed by final focus groups and a stated intentions questionnaire regarding use of the car sharing system. The same questionnaires were administered to a control group who did not receive the educational media. In this study, the analysis was limited to straightforward binary hypothesis testing and descriptive analyses. The study framework, however, could easily be extended to include predictive modeling of both the stated intention to use the car share system and the actual (RP) decision whether or not to use the system, as influenced by receiving various combinations of the information sources. In other words, such experimental policy introductions provide an ideal opportunity to carry out a ‘real world’ version of the Information Acceleration market simulation method. Others While the focus above is on a few specific lines of research, a wide variety of additional studies have been done that have yielded types of data that would also be very useful in a more comprehensive behavioral process data framework. These include:

• Studies of choice set consideration and formation (Fiorenzo-Catalano, et al. 2003)

• Mode choice segmentation methods based on attitudinal statements (Outwater, et al. 2003)

• Relationships between vehicle choice, attitudes, and personality traits (Collantes and Mokhtarian, 2002)

• Direct and innovative questioning on the desirability of travel (Handy, et al. 2003)

Data collection issues Asking questions about behavioral processes, some of them requiring respondents to think in depth about their own past or habitual behavior, brings up a new set of methodological issues in data collection. Bradley (2004) discusses a number of issues in some detail. Several of the issues have to do with the fact that some respondents may not be aware of the true motivating factors behind their behavior, particularly behavior that is largely unconscious or has become habitual. The best approach is to rely too much on respondents’ self-knowledge, but also not to place too much blind faith in our own deductive powers, as we often do at present.


20

2.4 Travel behavior in its sociodynamic context From another related viewpoint to behavioral processes, the idea of constraints in the movement of persons was taken a step further by the time geography school in Lund. In that framework, the movement of persons among locations can be viewed as their movement in space and time. Movement in time is viewed as the one way (irreversible) movement in the path while space is viewed as a three dimensional domain. Practical calculations of the extreme points in such a domain use existing data in surveys (Pendyala, 2003). Most important, however, in this framework is the idea of a project that according to Golledge and Stimpson, (1997) is a set of linked tasks that are undertaken somewhere at some time within a constraining environment (pages 268-269). This idea of the project underlies one of the most exciting developments in travel behavior – the activity-based approaches to travel demand analysis and forecasting that are methods of modeling time allocation. In this context, Chapin’s research (1974), providing one of the first comprehensive studies about time allocated to activity in space and time, is also credited for motivating the foundations of activity-based approaches to travel demand analysis. His focus has been on the propensity of individuals to participate in activities and travel linking their patterns to urban planning. In about the same period Becker also developed his theory of time allocation from a household production viewpoint (Becker, 1976) applying economic theory in a non-marketing sector and demonstrating the possibility of formulating time allocation models using economics reasoning (i.e., activity choice). In parallel, another approach was developing in geography and Hagerstrand’s seminal publication on time space geography (1970) presents the foundations of the approach. It provides the third base about constraints in human paths in time and space for a variety of planning horizons. Cullen and Dobson in two papers in the mid-1970s as reviewed by Arentze and Timmermans (2000) and Golledge and Stimpson (1997) appear to be the first researchers attempting to bridge the gap between the motivational (Chapin) approach to activity participation and the constraints (Hagerstrand) approach by creating a model that depicts a routine and deliberated approach to activity analysis. Most subsequent contributions to the activity-based approach emerge in one way or another from these initial frameworks with important operational improvements (for reviews see Kitamura, 1988, Bhat and Koppelman, 1999, Arentze and Timmermans, 2000, and McNally, 2000). The basic ingredients of an activity based approach for travel demand analysis (Jones, Koppelman, and Orfeuil, 1990 and Arentze and Timmermans, 2000) are:

a) explicit treatment of travel as derived demand (Manheim, 1979), i.e., participation in activities such as work, shop, and leisure motivate travel but travel could also be an activity as well (e.g., taking a drive). These activities are viewed as episodes (starting time, duration, and ending time) and they are arranged in a sequence forming a pattern of behavior that


21

can be distinguished from other patterns (a sequence of activities in a chain of episodes). In addition, these events are not independent and their interdependency is accounted for in the theoretical framework;

b) the household is considered to be the fundamental social unit (decision making unit) and the interactions among household members are explicitly modeled to capture task allocation and roles within the household, relationships and change in these relationships as households move along their life cycles and the individual’s commitments and constraints change and these are depicted in the activity-based model; and

c) explicit consideration of constraints by the spatial, temporal, and social dimensions of the environment is given. These constraints can be explicit models of time-space prisms or reflections of these constraints in the form of model parameters and/or rules in a production system format (Arentze and Timmermans, 2000).

The input to these models are the typical regional model data of social, economic, and demographic information of potential travelers and land use information to create schedules followed by people in their everyday life. The output are detailed lists of activities pursued, times spent in each activity, and travel information from activity to activity (including travel time, mode used, and so forth). This output is very much like a “day-timer” or “calendar” for each person in a given region. An output like this can easily contain the more traditional four-step model Origin-Destination matrices at different times of a day and the predicted “volumes” of individuals at specific locations in the city (Kuhnau and Goulias, 2003). Activity-based model systems that aim at becoming operational follow the three original traditions that are: a) microeconomics Homo Economicus vision based on Becker’s formulation (Jara-Diaz, 1998). b) production system/computational process modeling following Newel and Simon (1972) such as (Kitamura and Fujii, 1998, Arentze and Timmermans, 2000); and c) statistical pattern recognition and transition probability approaches to create synthetic schedules (Ma, 1997) or to improve existing four step models (Kuhnau and Goulias, 2003). Many planning and modeling applications, however, aim at forecasting. Inherent in forecasting are the time changes in the behavior of individuals and their households and their response to policy actions. Consideration of behavioral dynamics At the heart of behavioral change are questions about the process followed in shifting from a given pattern of behavior to another. In addition to measuring change and the relationships among behavioral indicators that change in their values over time, we are also interested in the timing, sequencing, and staging of these changes. Moreover and most important, we are interested in the triggers that may accelerate desirable or delay undesirable changes and the identification of social and demographic segments that may follow one time path versus


22

another in systematic patterns (e.g., some triggers may work better for some population segments). Knowledge about all this is required to design policies but it is also required to design better forecasting tools. Developments in exploring behavioral dynamics and advancing models for them made some progress in a few arenas. First, in the data collection arena with panel surveys, repeated observation of the same persons over time that are now giving us a considerable history in developing new ideas about data collection but also about data analysis (Golob, Kitamura, and Long, 1997, Goulias and Kim, 2003). In the same arena we also find interactive and laboratory data collection techniques (Doherty, 2003, Golledge 2002) that allow a more in-depth examination of behavioral tasks followed by decision makers as reviewed in the previous section. The second arena is in the development of microeconomic dynamic formulations for travel behavior (Supernak, 1989, Goodwin, 1998) that challenge conventional assumptions and offer alternative formulations. The third arena, is in the behavior from a developmental viewpoint as a single stochastic process (Kitamura 2000), a staged development process (Goulias, 1999), or as the outcome from multiple processes operating at different levels (Goulias, 2002). Experimentation with new theories from psychology emphasizing development dynamics is a potential fourth area that is just beginning to emerge (Goulias, 2003). The examples of studies in the previous section focus more on the paths of persons in space and time within a somewhat short time horizon such a day, week, or maybe a month. The consideration of behavioral processes and their dynamics has expanded the temporal horizons to a few years. In contrast, regional simulation models are very often designed for long range plans spanning 25 years or even longer time horizons. Within these longer horizons, changes in the spatial distribution of activity locations and residences (land use) are substantial, changes in the demographic composition and spatial distribution of demographic segments are also substantial, and changes in travel patterns, transport facilities, and quality of service offered can be extreme. Past approaches in modeling and simulating the relationship among land use, demographics, and travel in a region attempted to disengage travel from the other two treating them as mutually exogenous. As interactions among them became more interesting and pressing, due to urban sprawl and suburban congestion, increasing attention was paid to their complex interdependencies. This led to a variety of attempts to develop “integrated model systems” that enable the study of scenarios of change and mutual influence between land use and travel. An earlier review of these models with heavy emphasis on discrete choice models can be found in Anas (1982). Miller (2003) and Waddell and Ulfarsson (2003) twenty years later provide two comprehensive reviews of models that have integrated many aspects in the interdependent triad of demographics-travel-land use models. Both reviews trace the history of some of the most notable developments and both link these models to the activity-based approach above. Both reviews also agree that a microeconomic and/or macroeconomic approach to modeling land and transportation interactions are


23

not sufficient and more detailed simulation of the individuals and their organizations “acting” in a time-space domain need to be simulated in order to obtain the required output for informed decision making. They also introduce the idea of simulating interactive agents in a dynamic environment of other agents (multi-agent simulation). Creation of integrated systems is further complicated by the emergence of an entire infrastructural system as another layer of human activity - telecommunication. A critical gap in developing these integrated models that consider the relationship between land use and travel behavior are data that contain land use variables considered by trip makers. These variables can be the usual location descriptors that we find in most land use models but also other perceptions of quality of life.


24

3. An Example of Data Needs to Move us Beyond Conventional Thinking In this section and before defining data needs that can be used as a step toward the more advanced ideas presented above we use an example of current forward thinking practice as an example in activity-based modeling and simulation. In the section we describe design principles and components for the new generation of travel demand modeling system. All these proposed modeling components can be practically implemented within a reasonable time and budget framework and without excessive data requirements or risk of failure. These components have been carefully selected from an already implemented exhaustive list of successful practical experience in the US and worldwide.

3.1 Basic Features of the New Generation of Applied Travel Demand Models The new generation of applied regional travel demand models in the US currently includes the San-Francisco County Transportation Authority (SFCTA) model [Bradley, et al., 2001; Jonnalaggada, et al., 2001], Portland METRO model [Bowman et al., 1998; Bradley, et al., 1998; Bradley, et al. 1999], New York Metropolitan Transportation Council (NYMTC) model [Vovsha, et al., 2002; Peterson, et al., 2002], Mid-Ohio Regional Planning Commission (MORPC) model [Vovsha, et al., 2004a, 2004b; Vovsha & Bradley, 2004], and Atlanta Regional Commission (ARC) [PBConsult, 2004]. Comparing to the conventional 4-step model the new generation of models is characterized by the following three positive features:

1. Tour-based structure where the tour that is a closed chain of trips starting and ending at the base location (home or workplace) is used as the base unit of modeling travel instead of the elemental trip; this structure preserves a consistency across trips included into the same tour, by such travel dimensions as destination, mode, and time of day (TOD). In particular, the whole spectrum of travel dimensions (mode, destination and TOD) related to non-home-based trips can be properly linked to the relevant home-based trips.

2. Activity-based platform, that implies that modeled travel be derived within a general framework of the daily activities undertaken by households and persons including in-home activities, intra-household interactions, time allocation to activities, and many other aspects pertinent to activity analyses, but typically missing in the conventional travel demand models.

3. Micro-simulation modeling techniques that are applied at the fully-disaggregate level of persons and households, which convert activity and travel related choices from fractional-probability model outcomes into a series of “crisp” decisions among the discrete choices; this method of


25

model implementation results in realistic model outcomes, with output files that look very much like a real travel/activity survey data. Explicitly modeling a full list of households rather than zonal strata of households with identical attributes, avoids numerous aggregation biases that arise in the conventional modeling framework. It allows for more realistic and consistent linkage across travel choices made by the individual in a course of a day. Micro-simulation technique greatly reduces a size of the model intermediate outcomes to store and handle and, as a result, opens a way to apply a much higher level of segmentation (number of trip purposes, modes, household and person attributes) as well as of a spatial resolution (number of transport analysis zones and access sub-zones).

These three features constitute a fundamental core of the approach and are already incorporated in the first new-generation models developed for the Portland METRO, SFCTA, and NYMTC. Explicitly modeling a full list of individual households and persons in the region at the level of travel details close to the reality does not mean that the model system design is intended to pinpoint each individual behavior in each target year. In line with the basic paradigm of the disaggregate demand models this level of details is only a way to better predict aggregate travel statistics that are in focus of interest of transport planners. However, a better aggregate prediction of transportation flows is ensured by means of a more realistic and consistent representation of the underlying individual travel choices. Although at the individual level, even a perfect behavioral simulation cannot predict travel choices exactly because of the inherent random variability of individual, multiple individual simulations are summed to a reasonable prediction of the aggregate travel statistics of interest. In the same way as the probability theory is unable to predict a particular outcome of one toss of a coin, simulation gives almost exact prediction of the aggregate outcome of one thousand tosses. The right unbiased core probability model (50/50 in the coin tossing case) serves as a useful tool for aggregate predictions, not for individual outcomes. However, a quality of aggregate predictions is a direct function of the statistical quality of the individual core model. Thus, application of the individual simulations at the detailed level helps to properly capture numerous internal sensitivities of various population groups to changing travel conditions or land-use developments that would be otherwise inevitably lost in aggregation biases.

3.2 Advanced Features Added Recently Several new features and enhancements were incorporated in the recently completed Columbus (MORPC) model as well as in the Atlanta (ARC) model currently being developed. They reflect the growing body of research on activity-based modeling and micro-simulation for the last years. Two important and inter-related aspects have been frequently in the focus of research – intra-household


26

interactions and time-use framework that proved to be of critical importance for describing and modeling individual activity and travel behavior. Comparing to the previous model design, the new structures of MORPC and ARC represent two significant steps further in a better and more realistic description of travel behavior along these two lines:

• Explicit modeling of intra-household interactions and joint travel that is of crucial importance for realistic modeling of the individual decisions made in the household framework and in particular for choice of the high occupancy vehicle (HOV) as travel mode. The original concept of a “full individual daily pattern” that constituted a core of the previously proposed activity-based model systems has been extended in the MORPC and ARC systems to incorporate various intra-household impacts of different household members on each other, joint participation in activities and travel, and intra-household allocation mechanisms for maintenance activities.

• Enhanced temporal resolution (of 1 hour or less) with explicit tracking of available time windows for generation and scheduling of tours instead of the 4-5 broad time-of-day periods applied in most of the conventional and also activity-based models previously developed. The time-of-day choice model adopted for MORPC and ARC with further enhancements is essentially a continuous duration model transformed into a discrete choice form. The enhanced temporal resolution opens a way to explicitly control the person time windows left after scheduling of each tour and use the residual time window as an important explanatory variable for generation and scheduling of the subsequent tours.

The proposed enhancements are not just technical. They represent reflections on the natural and logical “evolution” of the model system structures in certain conceptual directions some of which are already in the form of operational structures while some others will be explored in the future.

3.3 Correspondence of Model Components to a Conventional Model Table 1 below summarizes the main model components and their correspondence between conventional 4-step and activity-based / tour-based model structures. It can be seen that the new model structure essentially covers all dimensions of the conventional 4-step model. However, the new structure adds numerous additional details and benefits to the modeling procedure. Table 1. Correspondence of the model components

Activity / tour based Conventional 4-step Tour level Trip level

Advantage of activity / tour based model

Trip production

Daily activity-travel pattern in terms of

Stop frequency by activity type

Inter-relations and trade-offs across generated tours


27

tours and trips during the day Trip attraction Primary

destination zonal choice size variables

Stop-location zonal size variables

Distinction between primary and secondary activities (stops on the way)

Trip distribution

Primary destination choice

Stop-location choice

Linkage of non-home-based trip origins and destinations to home-based trips

Modal split Entire-tour mode combination

Trip mode choice

Controlled consistency of mode choices across all trips in the tour

Trip time of day (TOD) factoring

Tour scheduling Arrival / departure time choice

Controlled consistency of TOD choices across all tours and trips; impact of TOD changes for one segment (say, trips to work) on the other segments (trips from work home)

Trip Assignment

Dynamic micro-simulation of individual travel patterns

Trip assignment

Impact of network conditions on activity pattern formation, scheduling, and implementation

Conventional trip production models by trip purpose are replaced with a comprehensive daily activity-travel pattern model that takes into account numerous inter-relations, trade-offs, and substitution effects between different tours and activities implemented by a person in the course of a day. Contrary to the conventional trip-production models, the daily activity pattern model is sensitive to the travel environment and accessibility, thus ensuring proper accounting for “induced demand”. Trip attraction models take a form of location-choice zonal size variables. In the activity-based modeling framework, these size variables are segmented by activity type and primary / secondary role in the tour structure. In particular, primary destinations are associated with major attractions, while secondary stops are more associated with favorable travel conditions. Conventional models fail to incorporate such a distinction. Trip distribution technique has undergone a significant transformation. One of the crucial deficiencies of the conventional 4-step models relates to the independent modeling of non-home-based trips while in the tour-based framework, locations of origins and destinations of non-home-based trips are properly linked to the location of the correspondent home-based trips in the same tour.


28

Another important and beneficial modification of the modeling technique relates to the consistent mode choice across all trips in the tour. In particular, the most important mode-related decision whether to use a private car or public / non-motorized mode is properly modeled at the level of entire tour while in the conventional 4-step framework it is modeled for each trip separately, thus creating numerous illogical mode combinations. One of the major advantages of the activity-based tour-based approach over the conventional 4-step approach is a full consistency of the time-of-day choices across different tours in the course of the day and different trips in the same tour. This makes the model system sensitive to policy measures and changes in any particular time period and track impacts of these changes to all other dimensions of travel demand. For example, if congestion pricing is applied in the AM period, it would change the departure time for outbound commuting trips; then it would have a certain reflection on the inbound (reverse) commuting trips; finally and through rescheduling of the reverse commuting trips it would have an impact on the post-work activity and trip generation as well, etc. Conventional models are unable to track all these effects and normally have a limited and unrealistic sensitivity of the time-of-day factors to policy measures within a particular period of a day only.

The network assignment procedures are currently the same for 4-step and activity-based tour-based models. It is based on the limitations of the available software packages for large-scale network simulations like TransCAD, TP+, or EM/2 that operate with aggregate zone-to-zone traffic flows. For this reason, the output of an activity-based tour-based micro-simulation model is converted into conventional trip-table format before assignment. As soon as micro-simulation assignment procedures become available and effective for handling regional networks, the additional advantage of the activity-based tour-based micro-simulation models will take place since these models provide direct input to the traffic micro-simulation models. Additional step further is expected in this field in view of microscopic network modeling of tours and entire daily patterns rather than elemental trips (the TRANSIMS project). This would allow for tracking impacts of traffic conditions and unexpected delays on scheduling / re-scheduling of activities.

3.4 Further Conceptual Directions

In the most general way these conceptual directions can be classified as the following “lines of integrity” in modeling various travel-related multidimensional choices:

• “Intra-person integrity” of each modeled individual daily activity and travel pattern in a sense that all modeled activity episodes, their durations, locations, and travel tours associated with visiting out-of-home activities are consistent and feasible within the person time-space constraints.


29

• “Inter-person intra-household integrity” that means that daily patterns of different household members are properly coordinated in view of participation in joint activities, joint travel arrangements as well as intra-household mechanism for allocation of maintenance activities, allocation of cars to the household members, etc.

Intra-person integrity is associated with a proper conditioning in sequence of choices related to each individual from the top-level choice related to the daily activity pattern type to the lower-level choice related to details of each activity episode. Intra-person integrity was in the core of the original concept of the daily activity pattern choice model [Bowman & Ben-Akiva, 1999; Bowman & Ben-Akiva, 2001; Bhat & Singh, 2000]. The major breakthrough that made this approach operational was the integrative formulation of the daily pattern in terms of a number and structure of travel tours rather than elemental episodes that provides the necessary input to the subsequent set of travel models. The number of observed individual daily activity patterns and structural complexity of the choice model in combination with a huge number of possible activity location alternatives make it impossible to model all dimensions in one choice structure. Thus, various hierarchical structures were proposed that resulted in a cascade of conditional choice models. This inevitable decomposition leads to two different structural lines within the intra-person integrity framework:

• “Downward Intra-person integrity” that means that all lower-level decisions in the choice hierarchy should be properly conditional upon the upper-level decisions and take into account a gradually narrowed scope of lower-level choice alternatives as the upper-level choices progress.

• “Upward Intra-person integrity” that means that when modeling upper-level choices the composite measure of quality of the lower-level choices associated with each upper-level alternative should be properly taken into account

Downward Intra-person integrity is not an automatic property of hierarchical cascades of choice models, especially if different activity dimensions such number of tours/activities, their location, and timing are considered. For example, first activity-based models for Portland METRO, SFCTA, and NYMTC had independent-by-tours mode, destination, and TOD choice models that could produce conflicting choices for different tours made by the same person. Downward intra-person integrity is ensured by a proper sequencing of models and tracking all important variables from choice to choice that accurately describe the feasible scope left for each subsequent choice and prevent conflicting choices for the same individual. It has recently been recognized that time-use approach provides an operational framework for downward intra-person integrity because time serves as an ultimate and constrained resource for any type of activity. From this point of view, it proved to be more convenient to generate tours/activities and schedule them


30

according to a certain hierarchy using residual time windows left after scheduling previously generated tours as variables explaining generation of the subsequent tours. Further research is needed to better understanding the interrelationship between activity generation and scheduling stages and their positioning in the model system hierarchy. Similar relationships should be further explored between such dimensions as activity locations/durations and tour configuration in terms of a distribution of activity episodes by tours. Also possible substitution between in-home and out-of-home (travel) activities can be considered as a part of the downward intra-person integrity issue. Upward intra-person integrity is important to prevent illogically bad choices made at the upper levels of the choice hierarchy that may result in impasse at the lower level (for example, if a worker who has three non-work tours in addition to the work tour has been assigned a work schedule from 7:00AM to 22:00 PM) as well as it is crucial for the model system sensitivity to travel environment from the upper-level activity generation choices. Conventional fractional-probability models use the log-sum (expected maximum utility over the lower-level choices) technique to “inform” the upper-level choices about what can happen down the hierarchy. This technique can be used in the micro-simulation framework as well, however it is extremely intensive computationally when it comes to calculation of tour mode choice log-sums for destination choice (takes more than 60% of running time of the model system) and is not realistic at all when full destination choice log-sums (across all destinations and TOD periods) are considered as variables for daily activity pattern model. One possible solution that is currently explored is to exploit the overall iterative framework of the model application and use generated lower-level outcomes from the previous iteration as variables in the upper-level choices at the next iteration. This approach can be interpreted as “learning process”. Time-use framework also can be affectively used in this iterative procedure. Instead of feeding-back computationally intensive but actually quite abstract log-sums contracted over multiple choice dimensions a simple variable representing total travel time spent by individual to realize the activity pattern in time and space, can be fed-back and considered at the next iteration for a choice of the new daily pattern. To make the upper-level choice sensitive to the total expected travel time a continuous time allocation model (with travel budget as input variable) can be applied first and then daily pattern type and the subsequent chain of choices can be made conditional upon the expected time allocation. With this actually very simple technically approach, the whole model chain will be sensitive to network improvements since these improvements are finally expressed in time savings. Inter-person Intra-household integrity principle includes numerous ways to incorporate intra-household interactions in a travel demand model, either explicitly or implicitly:

• Explicit joint or at least coordinated modeling of daily activity pattern types (or related activity-travel characteristics) for several household members.


31

Most frequently, time allocation units are used for modeling and the Structural Equation System is employed [Golob & McNally, 1997; Fujii et al, 1999; Meka eat al, 2002; Simma & Axhausen, 2001]. The proposed approach, used in the MORPC and ARC system, however, is based on a linked set of discrete choice models [Vovsha et al, 2004a]

• Explicit modeling of joint activity and travel. This component has been modeled in terms of either episode generation or time allocation between individual and joint activities [Ettema et al, 2004; Gliebe & Koppelman, 2002; Scott & Kanaroglou, 2002]. Explicit modeling of joint tours has been incorporated into the MORPC and ARC regional travel demand models [Vovsha et al, 2003].

• Explicit modeling of within-household allocation of maintenance activities to household members [Borgers et al, 2002; Srinivasan & Bhat, 2004]. The corresponding component has also been included and successfully tried in the MORPC modeling system [Vovsha et al, 2004b].

• Explicit allocation of cars to household members that accounts for actual availability of a car for a particular person’s travel tour [Wen & Koppelman, 1999; 2000; Miller et al, 2003]. This model component is reserved for future model development.

3.5 Outline of the Core Model Structure The current generation of activity-based regional travel demand models of which the MORPC and ARC model systems are the most advanced representatives, is based on a sequence of discrete choice models applied in a micro-simulation fashion. In the model system design, there always has been a question, what is the better behavioral unit that represents a decision maker for trip (or tour) generation stage – household or person. Conventional travel demand models are mostly household-based (i.e. applied at the entire-household level though any person-related characteristics can be incorporated) while the contemporary activity-based models tend to be person-based (i.e. applied at the individual person level though any household characteristics can be incorporated). The choice of the decision-making unit (household or person) is less crucial if simple statistical models are applied that link person/household characteristics to the number of generated trips/tours (like conventional regression models for trip production). Conventional trip production models based on limited market segmentation produce very similar results for both strategies (household-based and person-based), being aggregated at the zonal level, while the model outcomes at the individual level are not analyzed. Micro-simulation modeling implies more detailed segmentation by household and person types, and is much more sensitive to the choice of the decision-making unit. Additionally, since ensuring consistency at the individual level is one of the main challenges of micro-simulation modeling, it is important to find a right balance and linkage between household and person dimensions. The micro-simulation technique


32

allows this to be resolved by using the household for some choice dimensions and person for other dimensions. Micro-simulation also allows for explicit incorporation of intra-household interactions of various types. Many travel-related decisions are made within the complicated framework of the entire-household decision-making process, where each person’s preferences are intertwined and consolidated with those of all household members. As a result some activities are shared among several household members; some other ones are generated at the entire-household level but allocated to particular members to implement; while other activities have a purely individual character. In the design and development of the MORPC and ARC modeling system, the following three-part segmentation of household and person activities is used:

• Individual activities. Corresponding tours are generated and scheduled at the person level (with possible inclusion of the household variables, but without direct coordination of choices). The frequency of these activities is modeled for each person either as a part of the daily activity/travel pattern (as currently proposed), or by means of the frequency choice model.

• Allocated activities. Activities are generated at the entire-household level because they reflect the collective household needs. However, they are implemented and scheduled individually. Thus, an activity (or tour) frequency model is used for the household, followed by an intra-household allocation model that household members as alternatives.

• Joint activities. Corresponding tours are generated at the entire-household level and also implemented by several household members traveling together (and frequently sharing the same activity). A tour-frequency model is used for the household, followed by a person participation model that is applied for each generated tour and considers possible travel parties (subsets of the household members) as alternatives.

The activity types and trip purposes are grouped into three main segments:

• Mandatory activities (including going to work, university, or school).

• Maintenance activities (including shopping, banking, visiting doctor, etc).

• Discretionary activities (including social and recreational activities, eating out, etc).

Table 2 summarizes the main assumptions made regarding the possible combinations of activity types and settings. Only five out of the nine possible combinations are allowed, which greatly simplifies the modeling system, while preserving behavioral realism and covering most of the observed cases.


33

Table 2. Modeled Activity-Travel Segments Activity Type / Travel Purpose

Individual Setting

Allocation Setting

Joint Setting

Mandatory X Maintenance X X Discretionary X X Travel for mandatory activities is always assumed to have an individual character. Frequency of these activities, location, and scheduling are modeled for separately for each person. While household-composition variables are used in the utility functions for these individual activities, there is no explicit linkage across all choices made by different individuals with the notable exception of staying at home together or having a non-mandatory travel day together. This assumption is based on the fact that most of the mandatory activities have fixed frequencies and schedules defined exogenously to the household activity framework; however, a realistic activity-based model should be sensitive to the fact that unscheduled at-home activity (child at home sick) will negatively impact the frequency of other mandatory travel. Maintenance activities may be either allocated or joint. It is assumed that the maintenance function is inherently household-based, even if it is implemented individually or related to a need of a particular household member, like visiting doctor. Even in these cases, maintenance activities are characterized by a significant degree of intra-household coordination, substitution, and possibly sharing. Discretionary activities may be either individual or joint. It is assumed that these activities are not allocated to household members since they do not directly relate to household needs. Thus, these activities are either planned and implemented together by several household members or are planned and implemented individually. It is assumed that all else being equal, there is a predetermined structure of preferences in the activity generation and scheduling procedure along both dimensions (activity type and setting). Mandatory activities take precedence over maintenance activities, while maintenance activities take precedence over discretionary activities. Joint activities are considered superior to allocated activities, while allocated activities are in turn considered superior to individual activities. Combination of these two preference principles yields the following order of generation and scheduling activities that serves as the main modeling skeleton for the model system design:

1. Individual mandatory activities, 2. Joint maintenance activities, 3. Joint discretionary activities, 4. Allocated maintenance activities,


34

5. Individual discretionary activities. In the MORPC and ARC model system the stages 2, 3 and 4 are partially combined. The household generation of all joint tours may be done in one simultaneous choice structure and a further combination of joint and allocated tour generation stages is considered. The person participation models for joint and allocated tours, however, are still de-composed into stages 2-4 and implemented sequentially. The intra-household interactions and enhanced time-of-day (TOD) resolution have lead to several important re-arrangements in the day-level hierarchy of choices stemming from:

• Daily activity pattern types for all household members are modeled in a coherent way by means of explicit linkages across household members. These linkages account for the most important features of the daily patterns that have principal impact on the entire-day level choice (go to work or school, stay at home, or have a day-off for a major out-of-home non-mandatory activity). This requires de-composition of the individual daily pattern into several parts and modeling the first part (pattern type) for all household members taking into account interactions between them before going into pattern details for each person.

• Various episodic intra-household interactions in a form of joint or allocated activities are modeled explicitly. Explicit modeling of joint and allocated activities requires an entire-household formulation of the tour-generation model. Then, participation in the generated joint tours is modeled for each person as well as allocated tours are assigned to persons. This structure requires a further de-composition of the individual daily pattern into several successive stages. Essentially, in this structure, important aspects of the individual daily pattern emerge as the result of the numerous intra-household participation and allocation mechanisms, and the individual incorporates them, along with their work activity (if applicable) and individual discretionary activities into a complete pattern.

• Enhanced resolution of the time-of-day (TOD) choice model allows for explicit tracking of time-use attributes (time windows available for implementing activities and travel tours) for each person at each stage of the tour generation and scheduling procedure. In particular, it has proven to be beneficial to model time-of-day choice for mandatory activities (that normally take the biggest share of the daily time budget) first and them condition the further generation and scheduling of non-mandatory activities on the size of the residual time windows left after the mandatory activities have been scheduled. This requires a certain re-arrangement of the choice hierarchy with modeling time-of-day choice for mandatory tours earlier in the model stream.


35

Figure 2 presents the latest modified model system hierarchy adopted for the ARC project.

Person Pattern Type--primary activities--at-home or on-tour--work/school destination--work/school time period

Household Activities--joint tours --number and purpose --participation by HH subsets--maintenance activities --number --allocation to individuals

Person Pattern--extra stops--secondary tours--at-home maintenance

Tour--detailed purpose--time periods--destination and mode

Stop--Purpose--location--trip mode & departure time

--one per person--conditioned by pattern type of higher priority persons

--one per person--conditioned by pattern of higher priority persons

--one set of tours per person--conditioned by stops of higher priority tours

--one set of stops per tour--locations conditioned by stop priority--mode & time conditioned by temporal sequence

--one set per household

Figure 2. Current Proposed Household Activity and Travel Model System for ARC


36

3.7 Ability of NHTS to Provide the Right Data at the Right Time for Developing the Components in Forecasting This section represents a pragmatic view on improvements that can be made in NHTS and other large-scale household travel surveys in a short term in order to support development, estimation, and application of the new generation of travel demand models. There are several directions in which the static models based on the cross-sectional data can be improved with no explicit dynamics in either model estimation and data structures or model application. These improvements are comparatively simple and can be implemented in practice in a short term. The proposed improvements do not represent a digression from the long-term direction associated with transition from the current outcome-based to the future process-based data and model structures. It rather offers a useful intermediate “stop” on this way with taking practical advantages of what can be already done today or in the near future. For a long period of time, the structure of household travel surveys was limited by the inevitable consideration of the subsequent development of conventional 4-step models. One if the most important deficiencies of 4-step models is the matrix structure of the trip distribution and modal split sub-models that severely limited the model segmentation and a number of explanatory variables that could be used. The surveys were actually much “richer” than the models and it was not clear for what purposes the surveys should be complicated further. Shift to the micro-simulation modeling paradigm has lifted this technical limitation and opened the door to get out of the “loop”. Now we can apply more complicated models with unlimited segmentation and set of variables; thus came appetite for better data. There are at least three pragmatic and short-term directions in which travel demand models and the corresponding surveys can be significantly improved:

1. Better understanding and modeling of causal linkages across various dimensions of travel behavior.

2. Widening the range of explanatory variables used in models and collected in surveys.

3. Adding attitudinal and SP extensions to conventional RP surveys Each of these points is described below in details. These three directions are not independent and actually are closely intertwined. Most of the attitudinal and SP questions are actually intended to provide answers on the causal mechanism. Many newly proposed variables are actually derived from the causal linkages assumed, etc. These directions are distinguished for mostly explanation purposes.


37

Causal Linkages Focusing on causality represents a constructive intermediate stage between old-fashioned outcome-based approach and new process-based approach. The difference between outcome-based, cause-based, and process-based approaches can be illustrated by the following example of location choice for shopping. The conventional outcome-based approach would try to explain the chosen location by means of the location characteristics (size in terms of retail employment, distance from home, accessibility by different modes) and person/household characteristics (person type, gender, age, car ownership, presence of children, etc) in a single choice framework. All location, person, and household attributes would probably be blindly blended in the utility function and also all other locations (zones) in the region would be considered as available alternatives. The cause-based approach would be focused on formation of the available choice set under the given conditions of the person that are considered as prior in the causal chain and prove that these conditions indeed were fixed in the decision making at the time of the making the modeled decision (available time window, car availability, usual spatial “domain” of the person including location near the residential place, locations near the workplace / school, and locations for usual major shopping). Then formulation of a choice model would take a maximum advantage of the causal/conditional variables along with the conventional variables. The cause-based approach is oriented to a proper sequencing and conditioning of decision making steps in an overall static environment. The decision-process-based approach would be focused on both causal and chronological aspects of the decision making associated with the modeled event. Ideally, this would include a historical sequence of preliminary decisions regarding the time and location for the modeled shopping activity including probably numerous corrections and adjustments until the final decision was made and the corresponding activity was implemented. The described three approaches are not actually alternative. It can be easily seen that they are sequentially inclusive. All factors, variables, and observed statistics pertinent to the conventional outcome-based approach are still relevant for the cause-based approach as well as causality is still a part of the decision-making screening. However, in addition to “What” happens as a result of the combination of explanatory variables, the cause-based approach offers insights into the “Why” sequence of decisions and events that led to the modeled “What”. The decision-process-based approach makes additional step further in mapping the whole “How” chronology of the decision-making that was built up around the


38

modeled event. The modeling complexity and amount of information needed to these approaches grows exponentially from “What” to “Why” and then to “How”. Chronological peculiarities of individual decision-making are less important for large-scale models and frequently lead to complicated multi-stage procedures with numerous feedbacks that are difficult to convert into operational models. Understanding of casual linkages is a simpler task though it is a limited view on travel behavior. It may significantly improve the structure of the travel model system and sequencing of the modeled choices and associated decision-making steps. The cause-based approach to surveys pragmatically serves the existing static structure of choice models and helps improving it. It is not a substitution to a full-fledged process-based approach; it is a simplification that is practically helpful in a short term. It may also be helpful in a long term as well since the knowledge and understanding acquired in causal analysis may be of great value for the subsequent process-based analysis. There is no deviation from the main line of research there, just and attempt to formulate a constructive first step that would already bring some fruit in mostly practical, but also theoretical terms. There are many possible ways the travel-related decisions could be sequenced. For example, what are the causal relationships between departure time scheduling, trip chain complexity, and mode choice? Data on choice outcomes alone does not appear sufficient to provide much guidance on such questions. Introducing causality and proper sequencing in a static framework requires addition of specific questions to the household surveys that would refer to the order and conditionality of decisions as well as the formation of the choice set. In particular, for each visited activity location and the corresponding choice of destination, mode, and TOD, the following set of questions can be added to either RP or SP surveys:

• Was this activity preliminary scheduled or undertaken as a result of occasionally saved time in the course of the day?

• Were the destination, mode, and TOD choices made simultaneously or was there a certain order of conditional choices? Which of these choices are usual and stable over time and which are subject to change?

• If the actually chosen alternative was not available, what would be the second-best choice?

• Is there any predetermined area from which the locations choice was made (like shopping on the some shopping street in the town or visiting the closes cinema theatre) or the choice of location was based on some unique properties of the location not associated with any area around (like visiting Madison Square Garden or Carnegie Hall in NY)?


39

Screening up sequentiality and causal linkages across choices cannot be done by standard quantitative questions. This is one of the aspects where qualitative information is as important as quantitative measures. Additional “causal” data would complement estimation techniques by offering new covariates to help explain the differences that are identified across the population. For example, in hierarchical choice models of the Generalized Extreme Value class (multinomial logit, nested logit, cross-nested logit) that constitute the analytical core of the most of applied models, covariates actually explain sequencing of choice dimensions. Introducing causalty into the modeling framework should naturally reduce a tendency of using simplified models of compensatory utility maximization and work in favor of more elaborate decision making chains with partially non-compensatory rules (eliminations). In particular, the most mathematically elegant and complete theory that is based on non-compensatory rules is the Elimination-by-Aspects model proposed by A.Tversky in the 70th. This model can be estimated based on a pure static outcome-based survey. The Elimination-by-Aspects model can capture elimination thresholds from the cross-sectional comparison to the same extent as a Generalized Extreme Value model can infer sequence of choices through the correlation structure of random utility disturbances. In both cases, it would be interesting to see how process-based information could improve the model structure in an explicit way.

Explanatory variables Travel demand modeling for a long period of time has been heavily influenced by practical limitations of the 4-step modeling paradigm. The 4-step framework where each variable or travel segment would eventually require an additional multiplication of full origin-destination matrices, pushed the practical modelers and researches who supported model development to economize on explanatory variables as much as possible. This finally had led to an almost universal culture that was expressed in a limited set of variables like household size, number of workers, car ownership, income group, etc that are indeed important for travel behavior but are not nearly exhaustive as determinants of travel behavior on the side of households and persons. In the same way, attraction of different locations was measured by a limited number of zonal employment variables stratified by 3-4 major branches like industrial employment, office employment, commercial employment, etc. In a similar way, level of service variables by different travel modes were limited to average time and cost components that could be skimmed by the existing network simulation procedures. New modeling frameworks open a constructive way to add variables and explanatory power to travel models. A lot of improvements can be done within a


40

static decision-making framework just by adding right variables and providing them in the surveys. In many cases in the past, modelers resorted to mysterious decision-making processes in order to explain why the model was not able to explain the observed variability by more than 20-30% (usual case with disaggregate models) while the model had only 5-6 explanatory variables and 3-4 travel segments. Below is a list of traditionally used variables and new variables that could add significant explanatory power to such important travel models as mode and destination choice (trip distribution) taken as examples:

• Mode choice o Traditional variables:

Average travel time and cost, Number of transfers Household car ownership / sufficiency Household income Person age and driver license possession Area-type constants

o New variables: Travel time uncertainty (probability of delays) Reliability in terms of transit schedule adherence Parking constraints, search, and conditions Individual parking cost including free parking and discounted

parking eligibility Driving conditions / road type Probability of having a seat for transit Probability of having a parking place for auto and P&R Commercial and information services on transit stations and

P&R lots Frequency and location of stops on the way to and from the

primary destination Individual car availability for the person and given travel tour

taking into account broken cars (on one hand) and rented cars (on the other hand)

Joint travel arrangements with the other household members Individual GIS-based walk time and pedestrian conditions for

transit and non-motorized modes Road and personal safety / crime rate / public image

associated with the area of transit station / line Person-type, gender, age, and income specific time and cost

perceptions (VOT) Non-linear effects corresponding to marginal impacts of time,

cost, and other variables as functions of trip length Comfort and convenience in transit cars / possibility of

reading / using laptop / air conditioning • Destination choice


41

o Traditional variables: Mode-choice log-sum or particular time/cost/distance

variables Zone attraction variable based on the employment

/enrollment mix o New variables:

Bottleneck facilities (river-crossings, bridges, tunnels) Statutory borders (states, counties, municipalities, school

districts) Social frictions (income incompatibility, social / ethnic

clusters) Special sensitivity to transit accessible destinations of non-

driving population (children under 16, zero car households) Household composition and activity patterns that limit spatial

domain of activity (for example, having a preschool or school child at home)

Individual attraction characteristics and special trip generators that take into account size / profile of the individual attraction (we are going into more and more detail on the household / person side but still have terrible aggregate zonal attraction variables based on 3-4 crude employment variables

Cognitive maps based on the spatial domain of the household and person with the pivot points corresponding to most frequently visited usual locations (residential, work, school, usual shopping malls).

All new variables listed above have been already included in different research and modeling frameworks and contexts as well as the ways to quantify them and collect the corresponding data have been proposed. What is needed is to move these research achievements into practice of travel surveys and models. In particular, widening the range of explanatory variables should eventually allow for a full exclusion of flat mode-choice constants and distribution K-factors that dominate the current models and “explain” most of the observed variability. One of the most important and interesting general questions in the context of collecting data and modeling is the “nature vs. nurture” question. In simple words, would it be possible to explain most of the travel behavior by means of static cross-sectional analysis or it is essential to explicitly model dynamic aspects like path / state dependence, transition rules, transaction cost, etc?. Living aside the land-use development aspect where dynamic approach is obviously essential, and assuming that land use is fixed, we should take a close look on choices that we model and find statistical evidence of dynamic effects (like published works on “Hysteresis” that was found between public transit development and car ownership growth).


42

One of approaches to explore this effect is to use a variable of “years of living in this residential place” as a proxy for dynamic / learning / information-search cycle of the households. If, for example, there is evidence that “newcomers” are characterized by a higher average distance for non-mandatory purposes, it would definitely be an indication on a kind of “searching” and “learning” process that has not yet been settled in. If there is no statistical impact of this variable, we may conclude that learning and adaptation processes are quite quick and not that important for long-term planning and modeling. This observation, however, does not substitute for exploration of dynamics in a sense of long-term trends in travel behavior. Travel behavior obviously undergoes a significant evolution that is not captured by static travel demand models. There have been only several attempts to capture long-term trends in VOT estimates with the corresponding consequence for the choice model coefficients. Another interesting example of research in this direction was an attempt to build a trip distribution (spatial interaction) model that used a relative impedance measure (travel time scaled by the average regional travel time). This model has a self-calibration property in both estimation and application. This model is able to capture impact of urban sprawl on tolerance to longer travel times.

Attitudinal and SP extensions to the conventional RP surveys A conventional RP household survey still represents the major source of information for estimation of a travel demand model. By virtue of this survey it includes a detailed household and person information as well as a full description of the actual daily activity-travel patterns of all household members. It constitutes an ideal basis for additional attitudinal and SP type questions that would be put in the actual context. It is much better than a sharp-pointed standing-alone SP survey where normally one of the trips / activities is taken out of the daily pattern context and then different questions about hypothetical alternatives are pivoted off the observed choice. Attitudinal and SP extensions in the overall framework of a conventional RP survey can be thought of as a limited Information Acceleration technique. As long as the simulated choice environment is realistic and detailed enough to allow the respondent to follow what feels like a natural, unforced choice process, the resulting simulated choices will also be realistic. However, addition of attitudinal and SP questions to the household survey represents a practical problem since the existing household surveys are already at the border of lengths and complexity that can be tolerated by the interviewed persons. Thus, it is important to make these extensions as easy, natural, and not-time-taking as possible. This can be achieved by pre-prepared sets of answers from which the interviewed person could pick up one or use an open question if needed. These extensions are not intended to replace SP surveys.


43

They cannot include extensive SP games with multiple hypothetical alternatives considered unless there is an agreement with the interviewed household about extra time for the SP survey. They are mostly intended to better understanding of the actually made choices, their sequencing, and the way how choice sets were formed. There are several examples of extensions of this sort that could be added to the conventional household surveys:

• For mode and location choices, there can be a question if the mode/location was usual or occasional

• For mode and location choices, there can be a pre-prepared set of answers on question “why”. For transit mode choice it could include answers like “auto was not available”, “travel time is better”, “parking is a problem”, etc. For auto choice it could include answers like “had to drive a kid on the way”, “transit was not available”, “poor transit service”, For location choice for non-mandatory activity it is important to distinguish between choice of “the closest location for this activity” and choice of “special pre-planned location for a particular activity”

• For departure and arrival time choices, there can be a pre-prepared set of answers on question how the schedule was actually built like “Usual schedule for this activity”, “Planned in advance”, “Occurred in the course of the day because of necessity”, “Was added in the course of a day because of the saved time”. It may also be possible to ask respondents to order activities in the schedule by the schedule priority. It may be possible to ask respondents regarding mandatory activities, if it was any schedule adjustment in order to accommodate some other activities in the schedule.


44

4. Toward A New NHTS At first glance it may appear that asking NHTS to provide data for modeling and simulation activities as described above is out of place. NHTS is designed to provide a national data collection activity for transportation planning at the national level and helps some of the statewide and regional activities. However, planning in the US is mostly dictated by cyclical legislation that very often is haphazard. Continuity in planning directions is ensured by a general public agency culture instead of a true strategic plan. The institutional backbone of this mentality is in the programmatic component of regions and states. Figure 3 shows continuity in plans in two of the key dimensions of time and space. Local plans and special studies feed information and policies into the regional plans. These regional plans (comprehensive and mostly advisory plans in the Eastern US) became the true roadmap to the future of large geographical areas that can contain any number of residents (small areas of 50,000 residents and larger that include millions of residents). On top of these regional plans and based on the most recent legislation (TEA-21 and its many extensions) we have the statewide plans that most recently are strategic plans containing a large public involvement component. Many of the mode-specific projects that appear in the cycles of new national legislation and in appropriations emerge from local and regional planning that eventually get inserted into the statewide plans and through “negotiation” are subsequently inserted in the programmatic lists at the state and national levels. Data collection about travel behavior, however, is not based on the same continuity as programming and project selection. All three geographical levels (local, regional, and statewide) contain in one way or another data collection activities that are rarely linked to NHTS (unless a region purchases an add-on component). They are also rarely linked to other data collection activities such as the Census. The result is a tremendous amount of wasted resources and possible duplication because of the limited ability to expand and extrapolate smaller scale focused studies to larger geographical scales. In fact, these data collection investments are not even related to the Highway Performance Monitoring System that provides one of the most popular sources of model validation 9traffic counts and traffic counts by vehicle types on selected highways). A similar limitation is observed along the time scale. Figure 2 shows the episodic character of NHTS in a time-space continuum of planning and behavior as illustrated expensively in part 2. Many of the processes described in earlier sections are longitudinal processes. Since NHTS is a cross-sectional survey, many events are simply unobserved and as a result we have extremely limited ability to detect traces of behavioral processes in time and space and triggers of behavioral change even at the national level.


45


46

Typical regional planning models and usual 4-step applications Focused studies on a specific issue

Statewide planning in 15 to 25 year horizons

Nationwide “plans” most often dictated by national legislation – should be longer term but usually 5-6 years cycles

Figure 3 Planning Scales and the NHTS

Time Horizon

Geography

Local & Specialized

National

Statewide

Regional

NHTS episodic data collection


47

Figure 4 below shows how both descriptive analysis and predictive modeling are likely to evolve in the future (from left to right), as well as what types of data will be needed to support these innovations. At the left is the current “state of the practice” with analyses generally limited to trends in aggregate trip patterns, and modeling typically limited to aggregate trip-based models. The current “state of the practice” in travel/activity diary surveys is sufficient for these purposes. One step to the right is the current “state of the art”, with analyses of detailed tour-based and full-day travel patterns at the individual level, and microsimulation models of activities and travel. As described in Part 2 there is still a good deal of uncertainty regarding the causal relationships driving activity scheduling, particularly with regard to constraints and flexibility. A further step to the right is an area that will be the focus of much of the research in the coming years—analyses of the interrelationships between land use/location decisions and travel decisions, eventually reflected in integrated dynamic land use and activity model systems. This type of analysis will depend on more longitudinal data becoming available through panel surveys and/or through retrospective surveys. It will also depend on being able to link respondent data to detailed parcel-level land use databases, suggesting that survey efforts should be focused on areas where such databases already exist. Another policy and research area of growing interest and importance is that of the relationships between public health and safety and land use and transportation behavior. To the present data, most analysis along these lines has been rather aggregate and there has been very little modeling done. In the future, with adequate data, we can expect the descriptive analysis to become much more detailed and informative for policy, and for predictive models of air pollution and other health risks and accident exposures to become more accurate and useful. The use of automated GPS data capture in such surveys will be vital to obtain accurate objective pictures of lifestyle, movement and exercise patterns. Longer term methodological innovations may be focused on the more “psychological” aspects of transportation behavior, such as formation of attitudes, learning, the role of key formative experiences and the effects of normative and altruistic motives on choices. Here, the types of survey methods envisioned could range across a broad spectrum, including the “process data” survey methods discussed in Part 2. A less innovative approach (but perhaps more feasible and useful for NTPS), would be to add sets of questions to a standard core survey—for example, different subsamples could receive supplemental surveys


48

measuring their attitudes, experiences, and perceptions in specific travel-related areas, and the cross-sectional shifts in such factors across time would provide a key indicator of changing travel markets, regionally and/or nationally. Descriptive Analysis

Aggregate

trip patterns

Detailed individual

travel patterns

(Re)location and activity

patterns over time

Factors affecting

health and safety

Attitudes and

information about travel alternatives

Needs for Survey Data

Current

diary-type surveys

are sufficient

Need more

data on constraints

and flexibility

Need more longitudinal and

retrospective data, and more accurate land

use data

Need more data on

exercise and lifestyle,

supported by more GPS

data

Need more data

on attitudes, perceptions and effects of past experiences

Predictive Modeling

Aggregate trip-based

model systems

Activity

schedule model

systems

Dynamic land

use and activity model systems

Models of air

quality, accidents and

exposure

Models

incorporating learning and information

search

Figure 4 Evolution of Descriptive Analyses and Predictive Modeling Although the review and critique presented in the previous three sections offers evidence and examples of “failures” in current data collection to address policy and research questions, it did not offer concrete suggestions on a new NHTS design. The task of redesigning NHTS to target even the most focused aspect of new data collection is not feasible within a resource paper for a workshop. This task is not even feasible within the entire workshop. Instead, our hope is that new directions have now been defined along which new data collection can start. In addition, existing data collection activities such as NHTS can be modified in their margins so that an opportunity is given for innovative modeling to be developed further. For example, in the area of data and information needed we can identify a battery of questions that are repeated in all related surveys to use as anchors in “synthetic” population generation exercises. In the area of alternate survey designs we can envision an NHTS as a conglomerate of satellite survey designs that allow breadth and depth of knowledge targeted (Figure 5). The core component can be an enriched NHTS and the satellite surveys will procure the more in-depth knowledge about spatial aspects of travel behavior, behavioral processes, and longer term decisions and triggers of change. This needs to be carefully considered in terms of alternate design advantages and disadvantages,


49

target elimination of duplication and waste of resources but as one would expect it would also require additional institutional and overhead burden.

-

Longitudinal National NHTS

Component

Interface with ACS

& Census in General

Attitudes & Opinions

Link to Public Involvement

for States

Expanded Regional Add-ons Spatial

Measures

Expanded Regional Add-ons

Longitudinal

Spatial Perception

and Cognition

Link to Pilots

Activity-Based

Questions

Traditional NHTS

Figure 5 A Possible New Design for NHTS


50

Our workshop’s charge then should be to:

1. Identify in the universe of innovative modeling the most promising ideas for each of the short, medium, and long terms

2. Develop a hierarchy of data needs in each of these three time frames and models/processes

3. Identify core essential elements to use in expanding the core data collection

4. Identify core essential elements that go into each of the satellite surveys 5. Develop guidelines for interfaces with other major national data collection

efforts such as ACS/Census and HPMS 6. Develop guidelines for state and regional modules of data collection that

can enhance and complement NHTS.


51

Bibliography and References

1. Allam B. S. and K. G. Goulias (1999) Dynamic emergency evacuation management system using GIS and spatio-temporal models of behavior. Transportation Research Record, 1660, pp. 92-99.

2. Anas, A. (1982) Residential Location Markets and Urban Transportation. Academic Press, New York, NY.

3. Arentze T. and H. Timmermans (2000) Albatross A learning Based Transportation Oriented Simulation System. European Institute of Retailing and Services Studies, Technical University of Eindhoven, Eindhoven, NL.

4. Avineri E. and Y. Prashker (2003) Sensitivity to uncertainty: The need for a paradigm shift. CD-TRB ROM Proceedings, Paper presented at the 82nd Annual Transportation Research Board Meeting, January 12-16, 2003, Washington D.C.

5. Barret C., K. Berkbigler, L. Smith, V. Loose, R. Beckman, J. Davis, D. Roberts, and M. Williams (1995) An operational description of TRANSIMS. LA-UR-95-2393, Los Alamos, NM.

6. BBC 2004. Nokia unveils mid-air messaging. 2004 June 2. British Broadcasting Corporation.

7. Becker G. S. (1976) The Economic Approach to Human Behavior. The University of Chicago Press, Chicago, IL.

8. Ben-Akiva M.E. and S.R. Lerman (1985) Discrete Choice Analysis: Theory and Application to Travel Demand. MIT Press, Cambridge, MA.

9. Ben-Akiva M.E. and T. Morikawa (1989) Estimation of Mode Switching Models from Revealed Preferences and Stated Intentions. Paper presented at the International Conference on Dynamic Travel Behavior at Kyoto University Hall, July Kyoto, JP.

10. Ben-Akiva M.E., J. Walker, A. T. Bernardino, D.A. Gopinath, T. Morikawa, A. Polydoropoulou (2002) Integration of Choice and Latent Variable Models. In In Perceptual Motion: Travel behavior Research Opportunities and Application Challenges (ed H.S. Mahmassani), Pergamon, Amsterdam, NL.

11. Bhat C, Singh S. (2000) A Comprehensive Daily Activity-Travel Generation Model System for Workers, Transportation Research A, 34(A), 1-22.

12. Bhat C. and F.S. Koppelman (1999) Activity-based modeling of travel demand. In Handbook of transportation science (ed. R.W. hall), Kulwer, Boston, MA. Pp. 35-61.

13. Bhat C. R. (2000) Flexible model structures for discrete choice analysis. In Handbook of Transport Modelling (D.A. Hensher and K.J. Button editors) Pergamon, Amsterdam, NL. pp. 71-89.

14. Bhat C. R. (2003) Random Utility-Based Discrete Choice Models for Travel Demand Analysis. Chapter 10 in Transportation Systems Planning: Methods and Applications. Edited by K.G. Goulias, CRC Press, Boca Raton, FL, pp. 10-1 to 10-30.


52

15. Bockenholt U. (2002) Comparison and Choice: Analyzing Discrete Preference Data by latent Class Scaling Models. In Applied Latent Class Analysis (J.A. Hagenaars and A. L. McCutcheon). Cambridhe University Press, Cambridge, UK. Pp. 163-182.

16. Borgers A, Hofman F, Timmermans H (2002) Conditional Choice Modeling of Time Allocation among Spouses in Transport Settings, European Journal of Transport and Infrastructure Research, 2, 5-17.

17. Borgers A.W.J., F. Hofman, and H.J.P. Timmermans (1997) Activity-Based Modelling: prospects. In Activity-Based Approaches to Travel Analysis (eds D.F. Ettema and H.J.P. Timmermans). Pergamon, Oxfoord, UK.pp.339-351.

18. Bowman J, Ben-Akiva M (1999) The Day Activity Schedule Approach to Travel Demand Analysis. Paper presented at the 78th Annual Meeting of the Transportation Research Board, Washington D.C.

19. Bowman J, Ben-Akiva M (2001) Activity-Based Disaggregate Travel Demand Model System with Activity Schedules, Transportation Research A, 35A, 1-28.

20. Bowman J, Bradley M, Shiftan Y, Lawton K, Ben-Akiva M. (1998) Demonstration of an Activity Based Model System for Portland, Paper presented at the 8th World Conference on Transport Research, Antwerp, Belgium.

21. Bowman, J.L. "The day activity schedule approach to travel demand analysis," Ph.D. Dissertation, Department of Civil and Environmental Engineering, Massachusetts Institute of Technology, 1998.

22. Bradley M, Bowman J, Lawton K. (1999) A Comparison of Sample Enumeration and Stochastic Microsimulation for Application of Tour-Based and Activity-Based Travel Demand Models, Paper presented at the European Transport Conference, Cambridge, UK

23. Bradley M, Outwater M, Jonnalagadda N, Ruiter E (2001) Estimation of an Activity-Based Micro-Simulation Model for San Francisco. Paper presented at the 80th Annual Meeting of the Transportation Research Board, Washington D.C.

24. Bradley, M.A. (1988). “Realism and Adaptation in Designing Hypothetical Travel Choice Contexts”. Journal of Transport Economics and Policy, v. 22(1).

25. Bradley, M.A. (2004). “Process Data for Understanding and Modeling Travel Behavior”. Resource Paper for the Seventh International Conference on Travel Survey Methods, Costa Rica.

26. Bradley, M.A., Bowman, J.L., Shiftan, Y., Lawton, K. and Ben-Akiva, M.E.(1998) “A System of activity-Based Models for Portland, Oregon”. Report prepared for the Federal Highway Administration Travel Model Improvement Program. Washington, D.C.

27. Brög W. and E. Erl (1989) Interactive measurement methods - Theoretical bases and practical applications. Transportation Research Record, 765.


53

28. Brog, W. and E. Erl. (1980). “Interactive Measurement Methods: Theoretical Bases and Practical Applications”. TRB Transportation Research Record No. 765. Washington, DC.

29. Chandrasekharan B. and K.G. Goulias (1999) Exploratory longitudinal analysis of solo and joint trip making in the Puget Sound transportation panel. Transportation Research Record, 1676, pp. 77-85.

30. Chapin F. S. Jr.(1974) Human Activity Patterns in the City: Things people do in time and space. Wiley, New York, NY.

31. Cohen, G., I. Salomon, and P. Nijkamp (2002) Information-communications technologies (ICT) and transport: does knowledge underpin policy? Telecommunications Policy 26, pp 31-52.

32. Collantes, Gustavo and Patricia L. Mokhtarian (2002). “How Much Do I Travel?: Modeling Subjective Assessments of Personal Mobility”. Transportation Research Board 2002 Annual Meeting CD-ROM.

33. Creighton R. L. (1970) Urban Transportation Planning. University of Illiniois Press, Urbana, IL.

34. DHHS 2001: Code of Federal Regulations, Title 45: Public Welfare, Part 46: Protection of Human Subjects. US Department of Health and Human Services

35. Dillman D.A. (2000) Mail and Internet Surveys: the Tailored Design Method. Second Edition. Wiley, New York, NY.

36. Doherty S. (2003) Interactive methods for Activity Scheduling Processes. Chapter 7 in Transportation Systems Planning: Methods and Applications. Edited by K.G. Goulias, CRC Press, Boca Raton, FL, pp. 7-1 to 7-25.

37. Doherty S. T., N. Noel, M. Lee-Gosselin, C. Sirois, and M. Ueno (2001) Moving beyond observed outcomes: Global positioning systems and interactive computer-based travel behavior surveys. Transportation Research Circular,E-C026, March 2001, Transportation Research Board, Washington DC.

38. Doherty, S.T. and E.J. Miller (2000). “A Computerised Household Activity Scheduling Survey”. Transportation. 27(1): 75-97.

39. Doherty, Sean. T. (2003). “Should We Abandon Activity Type Analysis?” 10th International Conference on Travel Behavior Research. http://www.ivt.baum.ethz.ch/allgemein/iatbr2003.html

40. Ettema D, Schwanen T, Timmeramns H (2004) The Effect of Locational Factors on Task and Time Allocation in Households. Paper presented at the 83nd Annual Meeting of the Transportation Research Board, Washington, D.C.

41. Ettema, D., A. Borgers and H. Timmermans (1995) “SMASH (Simulation Model of Activity Scheduling Heuristics): Empirical Test and Simulation Issues”. Paper presented at the Conference "Activity Based Approaches: Activity Scheduling and the Analysis of Activity Patterns" Eindhoven, The Netherlands, May 1995.

42. Faivre d'Arcier B. (2000). " Hypothetical Situations : The Attempt To Find New Behavioral Hypotheses ", in Transport Surveys : Raising the


54

Standard, Transportation Research Circular E-C008, II-L 1-18, TRB, Washington, DC

43. FCC 1997. Commission's Rules To Ensure Compatibility with Enhanced 911 Emergency Calling Systems, CC Docket No. 94-102. Federal Communications Commission.

44. Fiorenzo-Catalano, S., S. Hoogendoorn-Lanser and R. van Nes. (2003). “Choice Set Composition Modeling in Multimodal Traveling”. 10th International Conference on Travel Behavior Research. http://www.ivt.baum.ethz.ch/allgemein/iatbr2003.html

45. Fujii S, Kitamura R, Kishizawa K. (1999) Analysis of Individuals’ Joint Activity Engagement Using a Model System of Activity-Travel Behavior and Time Use, Transportation Research Record 1676, 11-19.

46. Garling T, M. Kwan, and R. Golledge (1994) Computational-process modeling of household travel activity scheduling. Transportation Research Part B, 25, pp355-364.

47. Garling T. (1998) Behavioural assumptions overlooked in travel-choice modeling. In Travel Behaviour Research: Updating the state of play (eds. Juan de Dios Ortuzar, David Hensher, and Sergio Jara-Diaz), Pergamon, Amsterdam, pp. 3-18.

48. Garling T., T. Laitila, K. Westin (1998) Theoretical Foundations of Travel Choice Modeling: An Introduction. In Theoretical Foundations of Travel Choice Modeling (eds Garling T., T. Laitila, K. Westin), Pergamon, pp1-30.

49. Garrett M. and M. Wachs (1996) Transportation Planning on Trial. The Clean Air Act and Travel Forecasting. Sage Publications. Thousand Oaks, CA.

50. Gliebe J, Koppelman F (2002) A Model of Joint Activity Participations between Household Members, Transportation, 29, 49-72.

51. Gliebe J. P. and F. S. Koppelman (2002) A model of joint activity participation between household members. Transportation 29(1). Pp 49-72.

52. Golledge R. (2002) Dynamics and ITS: Behavioral Responses to Information Available from ATIS. In In Perpetual Motion, Travel Behavior Research Opportunities and Application Challenges (editor by H. Mahmassani). Pergamon, Amsterdam, NL, pp 81-126.

53. Golledge R.G. and R. J. Stimson (1997) Spatial Behavior: A Geographic Perspective. The Guilford Press, New York, NY.

54. Golledge, R. G., Marston, J. R., & Costanzo, C. M. (1997). Attitudes of visually impaired persons toward the use of public transportation. Journal of Visual Impairment & Blindness, 91(5), 446-459.

55. Golledge, R.G., and T. Garling (2003) Spatial Behavior in Transportation Modeling and Planning. In Transportation Systems Planning: Methods and Applications. (editor Goulias, K.G). CRC Press, Boca Raton, FL, pp 3.1-3.27.

56. Golob T, McNally M. (1997) A Model of Activity Participation and Travel Interactions between Household Heads, Transportation Research B, 31B(3), 177-194.


55

57. Golob T. F. and M.G. McNally (1996) A model of household interactions in activity participation and the derived demand for travel. Paper presented at the 75th Annual Transportation Research Board Meeting, Washington, D.C.

58. Golob T.F, R. Kitamura, and L. Long (1997) Panels for Transportation Planning: Methods and Applications. Kluwer, Boston, MA.

59. Golob, T.F. (2001), Travelbehaviour.com: Activity Approaches to Modeling the Effects of Information Technology on Personal Travel Behaviour, in Travel Behavior Research, The Leading Edge, D. Hensher, Editor. 2001, Elsevier Science/ Pergamon: Kidlington, Oxford, pp 145-184.

60. Goodchild MF, Cova T, Ehlschlaeger C. 1995. Mean objects: extending the concept of central tendency to complex spatial objects in GIS. Proceedings GIS/LIS '95, Nashville, TN, pp. 354–364.

61. Goodman L.A. (2002) Latent Class Analysis: the Empirical Study of Latent Types, latent Variables, and Latent Structures. In Applied Latent Class Analysis (J.A. Hagenaars and A. L. McCutcheon). Cambridhe University Press, Cambridge, UK. Pp. 3-55.

62. Goodwin, P. B. (1998) The end of equilibrium, In: Garling T. et al. (Eds), Theoretical Foundations of Travel Choice Modelling, Pergamon, Oxford. pp103-132.

63. Goulias K. G. (2001) A Longitudinal integrated forecasting environment (LIFE) for activity and travel forecasting (2001). In Ecosystems and Sustainable Development III (Ed. Y. Villacampa, C.A. Brebbia, and J-L. Uso), WIT Press, Southampton, UK, pp.811-820.

64. Goulias K. G. (2002) Multilevel analysis of daily time use and time allocation to activity types accounting for complex covariance structures using correlated random effects. Transportation, Volume 29(1), pp. 31-48.

65. Goulias K.G. (2003) Transportation Systems Planning, Chapter 1 in Transportation Systems Planning: Methods and Applications. Edited by K.G. Goulias, CRC Press, Boca Raton, FL, pp. 1-1 to 1-45.

66. Goulias K.G. and T. Kim (2003) Analysis of the Puget Sound Transportation Panel Survey Database in Waves 1-9 Draft Final Report. Submitted to the Puget Sound Regional Council. Seattle, WA.

67. Goulias K.G., T. Kim, and O. Pribyl (2003) A longitudinal analysis of awareness and use for advanced traveler information systems. Paper to be presented at the European Commission Workshop on Behavioural Responses to ITS – April 1-3, 2003, Eindhoven, the Netherlands.

68. Goulias K.G., T. Litzinger, J. Nelson, and V. Chalamgari (1993) A study of emission control strategies for Pennsylvania: Emission reductions from mobile Sources, cost effectiveness, and economic impacts. Final report to the Low Emissions Vehicle Commission. PTI 9403. The Pennsylvania Transportation Institute, University Park, PA.

69. Goulias, K. G., and D. Szekeres (1994). Centre Region Transportation Demand Management Plan. Final Draft Report. Prepared for the Centre Regional Planning Commission. University Park, PA.


56

70. Goulias, K.G. (1999) Longitudinal analysis of activity and travel pattern dynamics using generalized mixed Markov latent class models. Transportation Research, 33B, 535-557.

71. Goulias, K.G. (2001). “On the Role of Qualitative Methods in Travel Surveys”. Workshop Summary for workshop B5, 6th International Conference on Travel Survey Methods, Kruger National Park, South Africa.

72. Goulias, K.G. and R. Kitamura (1992) Travel Demand Analysis with Dynamic Microsimulation. Transportation Research Record, 1357, 8-18.

73. Goulias, K.G., W. Brog and E. Erl. (1998). “Perceptions in Mode Use Choice Using the Situational Approach: A Trip by Trip Multivariate Analysis for Public Transportation”. Transportation Research Board 1998 Annual Meeting. Preprint 981355. Washington, DC.

74. Haab T.C. and R.L. Hicks (1997) Accounting for choice set endogeneity in random utility models of recreation demand. Journal of Environmental Economics and Management 34. pp 127-147.

75. Hagerstrand T. (1970) What about people in regional science? Papers of the Regional Science association, 10, pp 7-21.

76. Handy, Susan, Lisa Weston and Patricia Mokhtarian. (2003). “Driving by Choice

77. Hayes-Roth B. and F. Hayes-Roth (1979) A cognitive model of planning. Cognitive Science 3, pp. 275-310.

78. Horowitz J. L. (1991) Modeling the choice of choice set in discrete-choice random-utility models. Environment and Planning A, 23, pp 1237-1246.

79. Horowitz J.L. and J.J. Louviere (1995) What is the role of consideration sets in choice modeling? International Journal of Research in Marketing 12. pp 39-54.

80. http://tmip.fhwa.dot.gov/clearinghouse/docs/abtf/kurani.pdf 81. http://www.dpi.wa.gov.au/travelsmart/pdfs/pubtransnz.pdf 82. Humphrey T.F. (1990) Suburban congestion: Recommendations for

transportation and land use responses. Transportation, 16, 221-240. 83. Hutchinson B. G. (1974) Principles of Urban Transport Systems Planning.

Scripta, wasghington, D.C. 84. Jara-Diaz S. R. (1998) Time and income in travel choice: Towards a

microeconomic activity-based theoretical framework. In Theoretical Foundations of Travel Choice Modeling (eds Garling T., T. Laitila, K. Westin), Pergamon, pp51-73.

85. JHK & Associates, Clough, Harbour & Associates, Pennsylvania Transportation Institute, and Bogart Engineering (1996) Scranton/Wilkes-Barre Area Strategic Deployment Plan. Final Report. Prepared for Pennsylvania Department of Transportation District 4-0. August 1996, Berlin, CT. pages 143.

86. John, G. and W. Brog. (2001). “Individualised Marketing – the Perth Success Story”. Presented at the Conference on Marketing Public Transport, Auckland, NZ.


57

87. Jones P. (1990) Developments in Dynamic and Activity-Based Approaches to Travel Analysis. A compendium of papers from the 1989 Oxford Conference. Avebury, UK.

88. Jones P., F. Koppelman, and J Orfeuil (1990) Activity analysis: State-of-the-art and future directions. In Developments in Dynamic and Activity-Based Approaches to Travel Analysis. A compendium of papers from the 1989 Oxford Conference (ed. P. Jones). Avebury, UK. Pp. 34-55.

89. Jones, P. M. and L. Sloman (2003). “Encouraging Behavioral Change Through Marketing and Management: What Can Be Achieved?”. 10th International Conference on Travel Behavior Research. http://www.ivt.baum.ethz.ch/allgemein/iatbr2003.html

90. Jones, P.M. (1979). “HATS: A Technique for Investigating Household Decisions” Environment and Planning A. Vol. 11(1).

91. Jones, P.M., M. Bradley and E. Ampt (1987). “Forecasting Household response to Policy Measures Using Computerised Activity-Based Stated Preference Techniques”. Proceedings of The 1987 Conference on Travel Behavior, Aix-en-Provence, October 1987.

92. Jones, P.M., M.C. Dix, M.I. Clarke and I.G. Heggie (1983). Understanding Travel Behaviour. Aldershot, UK: Gower Publishing.

93. Jonnalagadda N, Freedman J, Davidson W, Hunt J (2001) Development of a Micro-Simulation Activity-based Model for San Francisco – Destination and Mode Choice Models. Paper presented at the 80th Annual Meeting of the Transportation Research Board, Washington DC.

94. Kahneman, D., and A. Tversky (1979) Prospect Theory: An Analysis of Decisions under Risk, Econometrica Vol. 47, No. 2, pp. 263-291.

95. Kasten, R.A. and F.J. Sammartino (1990) A method for simulating the distribution of combined federal taxes using census, tax return, and expenditure microdata. In G.H. Lewis and R.C.

96. Kenworthy J. R. and F. B. Laube (1999) Patterns of automobile dependence in cities: an international overview of key physical and economic dimensions with some implications for urban policy. Transportation Research, Part A, 33, 691-723.

97. Kharoufeh J. P. and K. G. Goulias (2002) Nonparametric identification of daily activity durations using Kernel density estimators. Transportation Research, Part B Methodological, Volume 36, pp. 59-82.

98. Kitamura R (2000) Longitudinal methods. In Handbook of Transport Modelling (eds. D.A. Hensher and K.J. Button). Pergamon, Ansterdam, NL. pp. 113-128.

99. Kitamura R. (1988) An evaluation of activity-based travel analysis. Transportation 15, 9-34.

100. Kitamura R. and S. Fujii (1998) Two computational process models of activity-travel choice. In Theoretical Foundations of Travel Choice Modeling (eds Garling T., T. Laitila, K. Westin), Pergamon, pp 251-279.

101. Kitamura R., E.I. Pas, C.V. Lula, T.K. Lawton, P.E. Benson (1994) The sequenced activity mobility simulator (SAMS): An integrated approach to modeling transportation, land use and air quality. Mimeo.


58

102. Koppelman F.S. and V. Sethi (2000) Closed-form Discrete-choice Models. In Handbook of Transport Modelling (D.A. Hensher and K.J. Button editors) Pergamon, Amsterdam, NL.pp. 211-225.

103. Krizek, K.J., and A. Johnson.(2003) Mapping of the terrain of information and communications technology (ICT and household travel, Transportation Research Board annual meeting CD-ROM, Washington, DC., January 2003.

104. Kuhnau J. and K.G. Goulias (2003) Centre SIM: First-generation Model Design, Pragmatic Implementation, and Scenarios, Chapter 15 in Transportation Systems Planning: Methods and Applications. Edited by K.G. Goulias, CRC Press, Boca Raton, FL, pp. 16-1 to 16-14.

105. Kurani, K. and M. Lee-Gosselin. (1997). “Synthesis of Past Activity Analysis Applications”. Proceedings of the TMIP Activity-Based Forecasting Conference. New Orleans, LA.

106. Kwan M. (1995) GISICAS: An activity-based spatial decision support system for ATIS. Paper presented at the conference AActivity Based Approaches: Activity Scheduling and the Analysis of Activity Patterns.@ Eindhoven, The Netherlands, May 25-29, 1995.

107. Kwan, M-P. 1998. Space-time and integral measures of individual accessibility: A comparative analysis using a point-based network. Geographical Analysis, 30:191-216

108. Law A. M. and W.D. Kelton (1991) Simulation Modeling and Analysis. McGraw Hill, New York, NY.

109. Lawrence M.F. and M. Tegenfeldt (1997) The use of economic and demographic forecasts by metropolitan planning organizations. Paper presented at the 76th Annual Transportation Research Board Meeting. Washington D.C.

110. Lee, M. and K.G. Goulias (1997) Accessibility indicators for transportation planning using GIS. Paper presented at the 76th Annual Transportation Research Board Meeting. Washington D.C.

111. Lee-Gosselin, M. (1990) “The Dynamics of Car Use Patterns Under Different Scenarios: A Gaming Approach”. In Developments in Dynamic and Activity-Based Approaches to Travel Analysis. P.Jones (ed). Gower Press. 1990.

112. Lee-Gosselin, M. (1996) “Scope and Potential of Interactive Stated Response Data Collection Methods”, in Conference on Household Travel Surveys: New Concepts and Research Needs. Transportation Research Board Conference Proceedings 10. National Academy Press. Washington, D.C. 1996.

113. Levine (1997) The geography of time. 114. Linzie M. (2000) Future of international activities. Transportation in the

New Millenium. State of the Art and Future Directions. Transportation Research Board, Washington D.C.

115. Litman T. (2001) You can get there from here. Evaluating transportation choice. . Paper 01-3035. Presented at the Transportation Research


59

Board 80th Annual Meeting and included in the CD ROM proceedings, Washington, D.C.

116. Loikanen H.A. (1982) Housing demand and intra-urban mobility decisions: a search approach. Mimeo.

117. Lomborg B. (2001) The Skeptical Environmnetalist: Measuring the Real State of the World. Cambridge University Press, Cambridge, UK..

118. Loudon W.R. and D.A. Dagang (1994) Evaluating the effects of transportation control measures. In Transportation Planning and Air Quality II (eds. T.F. Wholley). American Society of Civil Engineers, New York, NY.

119. Loudon W.R., J.L. Henneman, L.I. Hartnett, and M.J. Lawlor (1997) Integrating transportation and land use planning: Addressing the requirements of federal legislation and rule making. Paper presented at the 76th Transportation Research Board Meeting, Washington, D.C.

120. Louviere J.J., D. A. Hensher, and J. D. Swait (2000) Stated Choice Methods: Analysis and Application. Cambridge University Press, Cambridge, UK.

121. Louviere, J.L., D.A. Hensher and J.D. Swait (2000). Stated Choice Methods: Analysis and Application. Cambridge University Press.

122. Lowry, I. S. (1988) Planning the urban sprawl. In A Look Ahead: Year 2020., (275-312), Transportation Research Board, Washington D.C..

123. Ma J. (1997) An Activity-Based and Micro-Simulated Travel Forecasting System: A Pragmatic Synthetic Scheduling Approach. Unpublished Ph.D. Thesis. Department of Civil and Environmental Engineering. The Pennsylvania State University, University Park, PA.

124. Ma J. And K.G. Goulias (1996) Multivariate marginal frequency analysis of activity and travel patterns in the first four waves of the Puget Sound Transportation Panel. Transportation Research Record (forthcoming).

125. Mackett R.L. (1990) Exploratory analysis of long term travel demand using micro-analytical simulation. In P.M. Jones (ed.), New Developments in Dynamic and Activity-Based Approaches to Travel Analysis, (384-405), Gower Publishing, Aldershot, England.

126. Mackett, R.L. (1985) Micro analytic simulation of locational and travel behaviour. In Proceedings of PTRC Summer Annual Meeting, Seminar L, Transportation Planning Methods, PTRC, London, England, 175-188.

127. Madder G.H. and O.M. Bevilacqua (1989) Electric vehicle commercialization. In Alternative Transportation Fuels. An Environmental and Energy Solution (D. Sperling ed.). Quorum Books. New York.

128. Mahmassani H.S. and R. Herman (1990) Interactive experiments for the study of tripmaker behaviour dynamics in congested commuting systems, in Developments in Dynamic and Activity-Based Approaches to Travel Analysis. A compendium of papers from the 1989 Oxford Conference. Avebury, UK.

129. Mahmassani H.S. and R.-C. Jou (1998) Bounder rationality in commuter decision dynamics:Incorporating trip chaing in departure time and route


60

switching decisions, in Theoretical Foundations of Travel Choice Modeling (eds Garling T., T. Laitila, K. Westin), Pergamon.

130. Manheim 91979) Transportation systems book. 131. Marker J. T. and Goulias K.G. (2000) Framework for the analysis of

grocery teleshopping. Transportation Research Record, 1725, pp.1-8. 132. McFadden D. (1998) Measuring Willingness-to-pay for Transportation

Improvements. In Theoretical Foundations of Travel Choice Modeling (eds Garling T., T. Laitila, K. Westin), Pergamon, pp339-364..

133. McNally M. G. (2000) the Activity-based Approach. In Handbook of Transport Modelling (eds. D.A. Hensher and K.J. Button). Pergamon, Ansterdam, NL. pp. 113-128.

134. Meka S, Pendyala R, Kumara M. (2002) A Structural Equations Analysis of Within-Household Activity and Time Allocation between Two Adults. Paper presented at the 81st Annual Meeting of Transportation Research Board, Washington, D.C.

135. Meyer M. D. and E. J. Miller (2001) Urban Transportation Planning. Second Edition. McGrawHill, boston, MA.

136. Miller E, M. Roorda, J.A. Carrasco (2003) A Tour-Based Model of Travel Mode Choice. Paper presented at the 10th International Conference on Travel Behavior Research, Lucerne.

137. Miller E. J. (2003) Land Use: Transportation Modeling. Chapter 5 in Transportation Systems Planning: Methods and Applications. Edited by K.G. Goulias, CRC Press, Boca Raton, FL, pp. 5-1 to 5-24.

138. Miller, HJ. 1991 Modelling accessibility using space-time prism concepts within geographical information systems. International Journal of Geographical Information Systems, 5:287-302

139. Mokhtarian, P.L. (1990) A Typology of Relationships Between Telecommunications and Transportation. Transportation Research A, 24(3): pp 231-242.

140. Mudge, RR. 1999. The market for advanced traffic management systems. ITS America Symposium “Show Me the Market,” Arlington, Virginia.

141. Newell A. and H. A. Simon (1972) Human Problem Solving. Prentice Hall. Englewood Cliffs, NJ

142. Niemeier D.A. (2003) Mobile Source Emissions: An Overview of the Regulatory and Modeling Framework. Chapter 13 in Transportation Systems Planning: Methods and Applications. Edited by K.G. Goulias, CRC Press, Boca Raton, FL, pp. 13-1 to 13-28.

143. Noronha V, 2004. Centerlines and other transportation data from GPS. TRB Annual Meeting, workshop on “Using GPS for GIS data collection.”

144. Noronha V, Goodchild MF. 2000. Map accuracy and location expression in transportation - reality and prospects. Transportation Research C Special Issue on GIS-T 8: 53–69.

145. or Necessity?”. 10th International Conference on Travel Behavior Research. http://www.ivt.baum.ethz.ch/allgemein/iatbr2003.html

146. Ortuzar J. de D. and L.G. Willumsen (2001) Modelling Transport (Third Edition). Wiley, New York, NY.


61

147. Outwater, M., S. Castleberry, Y. Shiftan, M. Ben-Akiva, Y. Shuang Zhou and A. Kuppam (2003). “Use of Structural Equations Modeling for an Attitudinal Market Segmentation Approach to Mode Choice and Ridership Forecasting”. 10th International Conference on Travel Behavior Research. http://www.ivt.baum.ethz.ch/allgemein/iatbr2003.html

148. Paaswell R.E., N. Rouphail, and T.C. Sutaria, Editors (1992) Site impact traffic assessment. Problems and solutions. ASCE, New York, NY.

149. Patten, M. L. and K. G. Goulias (2001) Test Plan: Motorist Survey - Evaluation of the Pennsylvania Turnpike Advanced Travelers Information System (ATIS) Project, Phase III PTI-2001-23-I. April 2001. University Park, PA.

150. Patten, M.L., M.P. Hallinan, O. Pribyl, and K.G. Goulias (2003) Evaluation of the Smartraveler advanced traveler information system in the Philadelphia metropolitan area. Technical memorandum. PTI 2003-33. March 2003. University Park, PA.

151. Payne J. W., J. R. Bettman, and E. J. Johnson (1993) The Adaptive Decision Maker. Cambridge University Press. Cambridge, UK.

152. PBConsult (2004). “Progress Report for the Year 2003: Modeling Task 2 – Activity/Tour-Based Models”. Regional Transportation Plan Major Update Project for the Atlanta Regional Commission, January, 2004.

153. Pendyala R. (2003) Time use and travel behavior in space and time. Chapter 2 in Transportation Systems Planning: Methods and Applications. Edited by K.G. Goulias, CRC Press, Boca Raton, FL, pp. 2-1 to 2-37.

154. Pendyala R. and C. Bhat, 2004 (2004) 155. Pendyala R., R. Kitamura, and D.V.G. P. Reddy (1995) A rule-based

activity-travel scheduling algorithm integrating neural networks of behavioral adaptation. Paper presented at the conference AActivity Based Approaches: Activity Scheduling and the Analysis of Activity Patterns.@ Eindhoven, The Netherlands, May 25-29.

156. Petersen, E., P. Vovsha and R, Donnelly (2002) “Managing Competition in Micro-simulation”. Paper presented at the 81st Annual Meeting of the Transportation Research Board, Washington, DC.

157. Pribyl O. and K. G. Goulias (2003) On the application of Adaptive Neuro-fuzzy Inference System (ANFIS) to analyze travel behavior. Paper presented at the 82nd Transportation Research Board Meeting and included in the CD ROM proceedings and accepted for publication in the Transportation Research Record, Washington D.C., January 2003.

158. Richardson A. (1982) Search models and choice set generation. Transportation Research Part A, 16A(5-6), pp 403-416.

159. Roberts P.O. (1973) Resource paper for Demand Forecasting for Long-range and Contemporary Options. In Urban Travel Demand Forecasting. Proceedings of a Conference at Williamsburg, Virginia. Special Report 143, Highway Research Board, Washington DC.

160. Rubinstein A. (1998) Modeling Bounded rationality. The MIT Press, Cambridge, MA.


62

161. Salomon, I., (1986) Telecommunication and travel relationships: A review. Transportation Research, Part A, 20(3), pp 223-238.

162. Savage L. J. (1954) The Foundations of Statistics. Reprinted version in 1972 by Dover Publications, New York, NY.

163. Savas E.S. (1983) Systems analysis and urban policy. In Systems Analaysis in Urban Policy-Making and Planning. (eds M. Batty and B. Hutchinson). NATO Conference Series. Serie II: System Science. Volume 12. Plenum Press, New York, NY.

164. Scott D, Kanaroglou P (2002) An Activity-Episode Generation Model that Captures Interaction between Household Heads: Development and Empirical Analysis, Transportation Research B, 36B, 875-896.

165. Shaheen, Susan (1999). Dynamics in Behavioral Adaptation to a Transportation Innovation: A Case Study of CarLink—A Smart Carsharing System. UCD-ITS-RR-99-16. Davis, California. October, 232 pp.

166. Simma A, Axhausen K. W. (2001) Within-Household Allocation of Travel – the Case of Upper Austria, Transportation Research Record 1752, 69-75.

167. Simon H. A. (1983) Alternate Visions of Rationality. In Reason in Human Affairs (H.A. Simon editor) Stanford University Press, Stanford, CA. pp.3-35.

168. Simon H. A. (1997) Administrative Behavior, Fourth Edition. The Free Press, New York, NY.

169. Southworth F. (2003) Freight transportation planning: Models and methods. In Transportation Systems Planning: Methods and Applications. (editor Goulias, K.G). CRC Press, Boca Raton, FL, pp 4.1-4.29.

170. Srinivasan S, Bhat C (2004) Modeling the Generation and Allocation of Shopping Activities in a Household. Paper presented at the 83rd Annual Meeting of the Transportation Research Board, Washington, D.C.

171. Stopher and Meyburg (1976) Behavioral Travel-Demand Models. Lexington Books, Lexington, MA.

172. Stopher P. R. (1994) Predicting TCM responses with urban travel demand models. In Transportation Planning and Air Quality II (eds. T.F. Wholley). American Society of Civil Engineers, New York, NY.

173. Sundararajan A. and K. G. Goulias (2002) Demographic Microsimulation with DEMOS 2000: Design, Validation, and Forecasting, Chapter 14 in Transportation Systems Planning: Methods and Applications. Edited by K.G. Goulias, CRC Press, Boca Raton, FL, pp. 14-1 to 14-23.

174. Supernak (1990) A dynamic interplay of activities and travel: Analysis of time of day utility profiles. In In Developments in Dynamic and Activity-Based Approaches to Travel Analysis. A compendium of papers from the 1989 Oxford Conference (ed. P. Jones). Avebury, UK. Pp. 99-122.

175. Swait J. and M. Ben-Akiva (1987a) Incorporating random constraints in discrete models of choice set generation. Transportation Research Part B, 21B(2), pp. 91-102.

176. Swait J. and M. Ben-Akiva (1987b) Empirical test of a constrained choice discrete model: Mode choice in Sao Paolo, Brazil. Transportation Research Part B, 21B(2), pp. 103-115.


63

177. Teodorovic, D., and Vukadinovic, K. (1998) Traffic Control and Transport Planning: A Fuzzy Sets and Neural Networks Approach. Kluwer, Boston, MA.

178. Thill J. (1992) Choice set formation for destination choice modeling. Progress in Human Geography 16, 3, pp 361-382.

179. Tiezzi E. (2003) The End of Time. WIT Press, Southampton, UK. 180. Train K. E. (2003) Discrete Choice Methods with Simulation. Cambridge

University Press. Cambridge, UK. 181. Train K. E. (2003) Discrete Choice Methods with Simulation. Cambridge

University Press, Cambridge, UK. 182. Transportation Research Board (1999) Transportation, Energy, and

Environment. Policies to promote sustainability. Transportation Research Circular 492. TRB. Washington D.C.

183. Transportation Research Board (2002) Surface Transportation Environmental Research: A Long-Term Strategy. Transportation Research Board, Washington, D.C.

184. Tversky (1969) Intransitivity of preferences. Psychological Review. 76. pp 31-48.

185. Tversky (1972) Elimination by aspects: A Theory of choice. Psychological Review 79. pp 281-299

186. Tversky, A., and Kahneman, D.(1992), Advances in Prospect Theory: Cumulative Representation of Uncertainty.Journal of Risk and Uncertainty Vol. 9, pp. 195-230.

187. Urban, Glen L, B.D. Weinberg and J.R. Hauser (1995). “Premarket Forecasting of Really New Products”. Journal Of Marketing 59(1).

188. Urban, Glen L, J.R. Hauser..W.J. Qualls, B.D. Weinberg, J.D. Bohlmann and R.A. Chicos. “Validation and Lessons from the Field: Applications of Information Acceleration”. MIT Sloan School of Management. WP # 3882.

189. van der Hoorn T. (1997) Practitioner=s Future Needs. Paper presented at the conference ATransport Surveys, Raising the Standard.@ Grainau, Germany, May 24-30.

190. Van der Waerden, P., H. Timmermans, and A. Borgers (2003). “The Influence of Key Events and Critical Incidents on Transport Mode Switching Behaviour: A Descriptive Analysis”. 10th International Conference on Travel Behavior Research. http://www.ivt.baum.ethz.ch/allgemein/iatbr2003.html

191. Van Wissen L. (1989) A model of household interactions in activity patterns. Paper presented at the International Conference on Dynamic Travel behavior Analysis. July 18-19, Kyoto University Hall, Kyoto, Japan.

192. Vause M. (1997) A Rule-Based Model of Activity Scheduling Behaviour. In Activity-Based Approaches to Travel Analysis (eds D.F. Ettema and H.J.P. Timmermans). Pergamon, Oxfoord, UK.pp.73-88..

193. Vovsha P, Bradley M. (2004) A Hybrid Discrete Choice Departure Time and Duration Model for Scheduling Travel Tours. Paper presented at the 83rd Annual Meeting of the Transportation Research Board, Washington, D.C.


64

194. Vovsha P, Petersen E, Donnelly R (2002) Micro-Simulation in Travel Demand Modeling: Lessons Learned from the New York Best Practice Model, Transportation Research Record 1805, 68-77

195. Vovsha P, Petersen E, Donnelly R. (2003) Experience, Tendencies, and new Dimensions in the Application of Activity-Based Demand Modeling Systems for Metropolitan Planning Organizations in the United States. Paper presented at the 10th International Conference on Travel Behavior Research, Lucerne.

196. Vovsha P, Petersen E, Donnelly R. (2003) Explicit Modeling of Joint Travel by Household Members: Statistical Evidence and Applied Approach. Transportation Research Record 1831, 1-10.

197. Vovsha P, Petersen E, Donnelly R. (2004a) Impact of Intra-Household Interactions on Individual Daily Activity-Travel Patterns. Paper presented at the 83rd Annual Meeting of the Transportation Research Board, Washington, D.C.

198. Vovsha P, Petersen E, Donnelly R. (2004b) A Model for Allocation of Maintenance Activities to the Household Members. Paper presented at the 83rd Annual Meeting of the Transportation Research Board, Washington, D.C.

199. Waddell P. and G. F. Ulfarsson (2003) Dynamic Simulation of Real Estate Development and Land Prices within an Integrated Land Use and Transportation Model System. Presented at the 82nd Annual Meeting of the Transportation Research Board, January 12-16, 2003, Washington, D.C. (also available in http://www.urbansim.org/papers/ - accessed April 2003)

200. Walker, Joan L. (1994). “Modeling traveler response to traveler information systems: Laboratory simulation of information searches using multimedia technology.”. M.S. Dissertation, MIT Dept. of Civil and Environmental Engineering. Cambridge, MA.

201. Wegener M. and F. Fηrst (1999) TRANSLAND Integration of Transport and Land Use Planning, Deliverable D2a Land-Use Transport Interaction: State of the Art. Final Draft, Institute of Spatial Planning, University of Dortmund, Dortmund, Germany.

202. Weiland R.J. and L.B. Purser (2000) Intelligent Transportation Systems. In Transportation in the New Millennium. State of the Art and Future Directions. Perspectives from Transportation Research Board Standing Committees. Transportation Research Board. National Research Council. The National Academies, Washington, D.C. Pages 6 (2000). (Also in http://nationalacademies.org/trb/).

203. Wen C, Koppelman F. (1999) Integrated Model System of Stop Generation and Tour Formation for Analysis of Activity and Travel Patterns, Transportation Research Record 1676, 136-144

204. Wen C, Koppelman F. (2000) A Conceptual and Methodological Framework for the Generation of Activity-Travel Patterns, Transportation, 27(1), 5-23.


65

205. Williams H. C. W. L. and J. D. Ortuzar (1982) Behavioral theories of dispersion and the mis-specification of travel demand models. Transportation Research B, 16B(3), pp.167-219.

206. Wilson F.D. (1979) Residential consumption, economic opportunity, and race. Academic Press, New York, NY.

207. Wilson, C. 2004. Personal communication. 208. Wolf J., R. Guensler, S. Washington, and L. Frank. (2001) Use of

Electronic travel diaries and vehicle instrumentation packages in the year 2000. Atlanta Regional Household Travel Survey. Transportation Research Circular,E-C026, March 2001, Transportation Research Board, Washington DC.

209. World Bank (1996) Sustainable Transport: Priorities for Policy Reform. Washington, D.C..

210. Zhang J, Fujiwara A (2004) Representing Heterogeneous Intra-Household Interactions in the Context of Time Allocation. Paper presented at the 83rd Annual Meeting of the Transportation Research Board, Washington, D.C.

211. Zhang J, Timmeramns H, Borgers A (2004) Model Structure Kernel for Household Task and Time Allocation Incorporating Household Interaction and Inter-Activity Dependency. Paper presented at the 83rd Annual Meeting of the Transportation Research Board, Washington, D.C.

212. Zhang J, Timmermans H, Borgers A (2002) A Utility-Maximizing Model of Household Time Use for Independent, Shared and Allocated Activities Incorporating Group Decision Mechanism, Transportation Research Record 1807, 1-8.

data needs for innovative modeling …onlinepubs.trb.org/.../nhts/workshop-dataneeds.pdfdraft...

Documents