understanding customer preferences for airport access ... · rail services between heathrow airport...

13
David Jevons, Oxera, Park Central 40/41 Park End Street, Oxford OX1 1JD, UK Tel: 44 (0)1865 253097 E-mail: [email protected] 226 © HENRY STEWART PUBLICATIONS 1750-1938 AIRPORT MANAGEMENT VOL. 5, NO. 3, 226–238 APRIL–JUNE 2011 Understanding customer preferences for airport access: Implications for forecasting Received (in revised form): 7th January, 2011 DAVID JEVONS leads Oxera’s work in modelling, advising clients on issues surrounding business planning, consumer behaviour and market analysis using advanced economic and analytical tools in commercial, regulatory and litigation contexts. THOMAS HOE has been a member of Oxera’s modelling team for two years and has experience of applying quantitative techniques in a variety of sectors; he specialises in econometric analysis. MATTHEW SHEPHERD has been a member of Oxera’s transport and modelling teams for two years and specialises in providing quantitative analysis for clients in the transport sector. Abstract This paper describes the creation of a model to forecast the demand for, and revenue from, travel on the Heathrow Express services between London Heathrow and central London.The construction of the model, based on a survey of passengers at Heathrow and statistical analysis of these survey data, is described, including the data collection and analytical methodology.The paper concludes with insights from the survey data and the analysis. Keywords market research, statistical analysis, surface access, pricing strategy, demand forecasting INTRODUCTION How passengers travel to airports is important for airports across the globe, with many arriving by car, with the asso- ciated congestion and environmental impacts. However, to provide surface access by other modes (particularly those that require substantial investment, such as rail), it is necessary to produce forecasts of passenger demand and revenue to the airport. This paper discusses the creation of a model which produces demand and revenue forecasts for Heathrow Express, using a survey of passengers at London Heathrow and stated-preference tech- niques to analyse passenger preferences and the implications for demand and revenue. The Heathrow Express and Heathrow Connect services provide fast and frequent rail services between Heathrow Airport and central London’s Paddington Station. Paddington is one of London’s major rail terminals, with connections to both the London Underground network and mainline rail services arriving from the west of England. Heathrow Express and Heathrow Connect have different service propositions and appeal to different mar- kets. Heathrow Express is the name of the operating company and the brand name of one of the services.To avoid confusion, in this paper, ‘Heathrow Express’ is the operating company and the services are referred to as Connect and Express, or the Connect and Express services.

Upload: others

Post on 20-Sep-2019

3 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Understanding customer preferences for airport access ... · rail services between Heathrow Airport and central London’s Paddington Station. Paddington is one of London’s major

David Jevons,Oxera,Park Central40/41 Park End Street,Oxford OX1 1JD,UK

Tel: �44 (0)1865 253097E-mail: [email protected]

226 © HENRY STEWART PUBLICATIONS 1750-1938 AIRPORT MANAGEMENT VOL. 5, NO. 3, 226–238 APRIL–JUNE 2011

Understanding customerpreferences for airport access:Implications for forecastingReceived (in revised form): 7th January, 2011

DAVID JEVONS leads Oxera’s work in modelling, advising clients on issues surrounding business planning, consumer behaviour andmarket analysis using advanced economic and analytical tools in commercial, regulatory and litigation contexts.

THOMAS HOE has been a member of Oxera’s modelling team for two years and has experience of applying quantitative techniquesin a variety of sectors; he specialises in econometric analysis.

MATTHEW SHEPHERD has been a member of Oxera’s transport and modelling teams for two years and specialises in providing quantitativeanalysis for clients in the transport sector.

AbstractThis paper describes the creation of a model to forecast the demand for, and revenue from, travel onthe Heathrow Express services between London Heathrow and central London.The construction of themodel, based on a survey of passengers at Heathrow and statistical analysis of these survey data, isdescribed, including the data collection and analytical methodology.The paper concludes with insightsfrom the survey data and the analysis.

Keywordsmarket research, statistical analysis, surface access, pricing strategy, demand forecasting

INTRODUCTIONHow passengers travel to airports isimportant for airports across the globe,with many arriving by car, with the asso-ciated congestion and environmentalimpacts. However, to provide surfaceaccess by other modes (particularly thosethat require substantial investment, such asrail), it is necessary to produce forecasts ofpassenger demand and revenue to theairport. This paper discusses the creationof a model which produces demand andrevenue forecasts for Heathrow Express,using a survey of passengers at LondonHeathrow and stated-preference tech-niques to analyse passenger preferencesand the implications for demand andrevenue.

The Heathrow Express and HeathrowConnect services provide fast and frequentrail services between Heathrow Airportand central London’s Paddington Station.Paddington is one of London’s major railterminals, with connections to both theLondon Underground network andmainline rail services arriving from thewest of England. Heathrow Express andHeathrow Connect have different servicepropositions and appeal to different mar-kets. Heathrow Express is the name of theoperating company and the brand nameof one of the services.To avoid confusion,in this paper, ‘Heathrow Express’ is theoperating company and the services arereferred to as Connect and Express, or theConnect and Express services.

JAM189.qxd 4/7/11 5:12 PM Page 226

Page 2: Understanding customer preferences for airport access ... · rail services between Heathrow Airport and central London’s Paddington Station. Paddington is one of London’s major

UNDERSTANDING CUSTOMER PREFERENCES FOR AIRPORT ACCESS

© HENRY STEWART PUBLICATIONS 1750-1938 AIRPORT MANAGEMENT VOL 5, NO. 3, 226–238 APRIL–JUNE 2011 227

The Express service uses dedicated,high-quality rolling stock to provide aservice of four trains per hour between allHeathrow terminals and Paddington. It isa premium service, with a train always atthe platform at Paddington, a non-stopride between the airport and the station,and prices that reflect the features of theservice and the high level of customerservice. Carrying more than 16,000 pas-sengers per day, the service is targeted atbusiness travellers.

The Connect service uses rollingstock similar to that on many suburbancommuter lines, with fewer seats andmore standing space.There are two trainsper hour between the airport and thestation, but the service takes longer thanthe Express as it makes a number of stopsen route; however, the prices are gearedto reflect this different service pattern,and the service is targeted at leisure trav-ellers and airport staff travelling to andfrom the airport.

To assist with its business planning,Heathrow Express uses a model to fore-cast demand and revenue under differentscenarios.The forecasts from the previousmodel had begun to decrease in accuracyfor several reasons, including:

● the addition of a new terminal(Terminal 5) at Heathrow;

● airlines moving terminals due to theextensive work being undertaken atHeathrow, including the closure ofTerminal 2 and the creation ofHeathrow East;

● the existing model not producing fore-casts for the Connect service; and

● the fares structure offered on theExpress service having changed sincethe model was created.

In addition, the previous forecastingmodel was based on information from a

period of strong economic growth. Thesevere recession following the financialcrisis and the changing economic envi-ronment meant that customer tastes andpreferences were likely to have changedsince the original model was developed,which is another possible explanation forthe decline in accuracy of the model.

Consequently, Heathrow Expressneeded to develop a new demand andrevenue forecasting model to assist withits business planning.The new model wasrequired to provide the standard featuresexpected of models of this type (eg sce-nario planning, understanding the likelyimpact of different pricing strategies, theintroduction of new tickets and changesin the characteristics of different accessmodes), but also to forecast the impactof non-standard events such as airlinesmoving terminals.Any new model wouldalso need to be able to forecast the impactof abstraction (passengers switching) fromExpress to Connect, given the differentcharacteristics of the two services. A newmodel was constructed that provided theability to forecast these impacts; however,the model does not forecast the effect ofhigh-impact, low-probability events, suchas the Icelandic volcano eruption in early2010 and its consequences for air services,or the disruption caused at Heathrow bythe severe snowfall in December 2010.The construction of the model isdescribed in the remainder of this paper.

To predict future demand and revenueaccurately, the model has to take intoaccount the fact that customers wanting totravel between London and Heathrow havea choice of multiple modes of transport —ie London Underground, rail (Express andConnect), coach, car and taxi — all ofwhich have different prices, journey timesand inherent features. Different types ofconsumer are likely to have different pref-erences for these characteristics and modes

JAM189.qxd 4/7/11 5:12 PM Page 227

Page 3: Understanding customer preferences for airport access ... · rail services between Heathrow Airport and central London’s Paddington Station. Paddington is one of London’s major

JEVONS, HOE AND SHEPHERD

228 © HENRY STEWART PUBLICATIONS 1750-1938 AIRPORT MANAGEMENT VOL 5, NO. 3, 226–238 APRIL–JUNE 2011

of transport.A model was designed to pro-duce monthly forecasts of revenue andpassengers for 15 monthly periods ahead,plus a five-year, long-term outlook with alow margin of error.The results of statisticalanalysis of the survey of passengers atHeathrow were combined with otherinputs (including prices and GDP fore-casts) to produce the final demand andrevenue forecasting model.These steps arediscussed below, starting with the defini-tion of the relevant market for passengerstravelling to Heathrow using Express andConnect services.

MARKET DEFINITIONTo make the model tractable yet accurate, itwas important to define the market dimen-sions within which Express and Connectoperate. Geographic scope, the potentialtransport options and passenger demo-graphics are the three primary dimensions.

Geographic scopeHeathrow receives departing passengersfrom all over Great Britain and fromother countries for connecting passen-gers. Non-flying visitors to the airport (egstaff) were not considered. The relevantsub-set of these passengers for this exer-cise was considered to be those passengerswho had a realistic option to use Expressor Connect.This exercise divided passen-gers into three categories, those of:

● London origin;● non-London, UK origin having trav-

elled via a London mainline rail station;● non-London, UK origin having trav-

elled directly to Heathrow.

All passengers in the first two categorieswere assumed to have the option of usingExpress or Connect; the second category

picks up passengers who have an inter-mediate stop in London (eg fromCambridge). Passengers in the third cate-gory were assumed never to have consid-ered travelling via London. Certainpassengers not travelling via London maybe relevant (ie they could feasibly haveconsidered travelling via London andHeathrow Express as an alternative route)but were excluded by this assumption.The assumption to exclude such passen-gers was predicated on the grounds thatthey were a relatively small proportion ofpotential Heathrow Express users and thatminor changes to the Heathrow Expressservices were unlikely to affect the major-ity of these passengers’ travel decisions. Inaddition, providing travel time and costsfor all modes of transport across the UKwould have significantly increased thecosts of the stated-preference exercise.

Within the first two categories, themarket was further segmented by specificarea or connecting station. London wassplit into eight areas: the segmentation wasmore granular in east and central Londonto reflect the more diverse journey times inthese areas.West London was split into twobroad areas: ‘near’ (travel zones 1–3) and‘far’ (travel zones 4�). In total, seven main-line connecting stations were included.Together, 16 separate journey origins wereconsidered (nine regions of London plusseven mainline stations). Figure 1 showsthe geographic scope of the market fortransportation to Heathrow.

Field and desk research were used toobtain travel information from each ofthese journey origins. For regional zones(1–8), journeys were assumed to beginfive minutes’ walk from a centrally locatedLondon Underground station. Thisassumption on ‘access time’ was relevant asit adds to the total journey time. For largezones, travel times were collected frommultiple starting points and then averaged.

JAM189.qxd 4/7/11 5:12 PM Page 228

Page 4: Understanding customer preferences for airport access ... · rail services between Heathrow Airport and central London’s Paddington Station. Paddington is one of London’s major

UNDERSTANDING CUSTOMER PREFERENCES FOR AIRPORT ACCESS

© HENRY STEWART PUBLICATIONS 1750-1938 AIRPORT MANAGEMENT VOL 5, NO. 3, 226–238 APRIL–JUNE 2011 229

Potential transport optionsSix alternative options for transport toHeathrow were identified as crediblealternatives to the Express and Connectservices in terms of price, journey timeand frequency. Slower alternatives, such aslocal buses, walking or cycling were con-sidered, but identified as being insuffi-ciently close in terms of substitutability.The complete list of options consideredwas Express, Connect, park and fly, kissand fly, black cab, minicab, coach andLondon Underground. Park and fly isdefined as driving to and parking at theairport; it was assumed that a mid-rangepriced car park was chosen. Kiss and fly isdefined as being driven to the airport bya friend or relative and dropped directly atthe terminal. Black cab refers to a taxihailed on the street, while minicab refersto a pre-booked taxi.

For each mode, six characteristics wereconsidered: fare, access cost, access time,journey time, egress time and waiting time.

Access time is defined as the time from thejourney origin to reaching the main modeof transport (including any initial walkingdistance, assumed to be five minutes to thenearest Tube station in many cases). Egresstime is defined as the time from the mainmode of transport’s destination to the ter-minal building. Waiting time is defined asthe average waiting time between servicesof the main mode of transport. It wasassumed that the London Undergroundwas always chosen as the mode of transportto reach the starting point of the mainmode of transport (ie Paddington Stationfor Express and Connect, or VictoriaStation for coach). Table 1 gives anoverview of the main modes of transportand their broad characteristics.

As Table 1 shows, the offerings acrossmodes of transport are highly differenti-ated. Choices vary from high-cost offer-ings (taxi or park and fly) to low-costofferings (coach or Tube); and from lowjourney-time offerings (Express or

Figure 1 Geographic market definitionNotes:The segments that make up the numbered regions reflect the boundaries of each borough. For example, region 5 is made up ofthe Westminster Borough only. Region 8 is split into two separate journey origins, 8a and 8b, depending on in which London travelzone the journey began.Source: Oxera

JAM189.qxd 4/7/11 5:12 PM Page 229

Page 5: Understanding customer preferences for airport access ... · rail services between Heathrow Airport and central London’s Paddington Station. Paddington is one of London’s major

JEVONS, HOE AND SHEPHERD

230 © HENRY STEWART PUBLICATIONS 1750-1938 AIRPORT MANAGEMENT VOL 5, NO. 3, 226–238 APRIL–JUNE 2011

Connect) to high journey-time offerings(coach or Tube). Other factors beyondthese six attributes also differentiate theofferings, eg brand value, comfort andease of use.

Passenger demographicsDifferent types of passenger are expectedto have different preferences for modes oftransport. To reflect this in the forecasts,consideration was given to the demo-graphics and characteristics that individ-ual passengers may have. Passengerswere categorised across the followingdemographics:

● travel purpose (business versus leisure);● group size (1, 2, 3� passengers);● UK resident status (resident versus non-

resident);● income (high versus low).

This gave rise to 24 sub-groups, each ofwhich was analysed separately using themethodology outlined below.

METHODOLOGYCreating the forecasting frameworkrequired three key building blocks:

● a survey;● statistical analysis of the survey results;

and● construction of an interactive modelling

tool.

Surveying passengers at Heathrow AirportA survey of 1,000 passengers travellingto or from Heathrow was undertakenduring the summer of 2009. The periodincluded two weeks before and twoweeks during the school summer break.Passengers were drawn from the variousdeparture lounges across the five terminalswith the help of BAA’s market researchteam. Fixed quotas for the proportion ofrespondents per terminal were enforcedto ensure that the sample was represent-ative of the airport as a whole.Respondents were initially filtered via a

Table 1 Modes of transport and their characteristics

Mode Relative fare Access cost Access time Journey time Egress time Waiting time

Express Medium Tube price Tube plus Low Dependent Mediumfive-minute on terminalwalk

Connect Medium Tube price Tube plus Low–medium Dependent Mediumfive-minute on terminalwalk

Park and fly Very high None None Medium Dependent Noneon terminal

Kiss and fly Low None None Medium Low NoneBlack cab High None None Medium Low NoneMinicab High None None Medium Low NoneCoach Low Tube price Tube plus High Dependent High

five-minute on terminalwalk

London Very low None Five-minute High Dependent LowUnderground walk on terminal

Note: The park-and-fly fare includes the average cost of parking for a two-week holiday.Source: Oxera

JAM189.qxd 4/7/11 5:12 PM Page 230

Page 6: Understanding customer preferences for airport access ... · rail services between Heathrow Airport and central London’s Paddington Station. Paddington is one of London’s major

UNDERSTANDING CUSTOMER PREFERENCES FOR AIRPORT ACCESS

© HENRY STEWART PUBLICATIONS 1750-1938 AIRPORT MANAGEMENT VOL 5, NO. 3, 226–238 APRIL–JUNE 2011 231

series of questions so that only those whosatisfied the market definition criteria (asdiscussed above) were asked the mainbody of questions.

The main body of the survey was splitinto two sections: one series of revealed-preference questions and another ofstated-preference questions. The formerasked about actual, historical behaviour ofrespondents, eg ‘How did you travel to theairport today?’ The latter asked questionsabout future or hypothetical intentions,eg ‘How might you travel to the airporttomorrow?’

The factual information ascertainedfrom the revealed-preference questionshad two purposes. First and foremost, itwas used to inform the stated-preferencequestions. Using factual information oneach respondent (eg their journeyorigin) allowed the hypothetical ques-tions to be personalised so that thesequestions were as realistic as possible. Forexample, the simple example given above(‘How might you travel to the airporttomorrow?’) is likely to elicit a morerealistic answer if the available choicesare realistic (eg with a travel time by carthat accurately reflects the respondent’sdistance from the airport). Field and deskresearch were used to create a database ofrealistic journey times.

A second use for the revealed-preferencequestions was to gain information forHeathrow Express’s wider market researchpurposes.

Stated-preference survey techniquesStated-preference techniques enableresearchers to understand what passen-gers (or equivalently, customers) woulddo in hypothetical situations, ie theyenable researchers to ask, in a sophisti-cated way (avoiding response bias as far as possible), what people would do in a

given situation.This can provide powerfulinsights into customer willingness to payfor changes in service quality, what theywould do if a new piece of infrastructurewere available, and so on. In this case, itwas important to know how passengerswould react to changes in price and otheroperational characteristics, such as fre-quency, in both the Express and Connectservices, and the competing modes, suchas taxi.

The process of obtaining the informa-tion from a stated-preference survey con-sists of two parts: conducting the surveyand obtaining the data, and econometric/statistical analysis (see the followingsection).

A stated-preference survey is oftenconducted face to face with the respon-dent, with the aid of a computer, or overthe internet (such surveys are known ascomputer-aided personal interviewingsurveys). Respondents were asked tochoose repeatedly between severaloptions (see Figure 2 for an example),with the characteristics of the optionschanging slightly each time. The optionsthat the respondent was given were cali-brated to be as realistic as possiblethrough the use of the revealed-prefer-ence data, as discussed above.This calibra-tion should result in the respondentengaging with the survey and providing

Figure 2 Choice screenNote:This is an example of a simplified choice screen, it is notthe screen used in the survey reported in this paper.Source: Oxera

JAM189.qxd 4/7/11 5:12 PM Page 231

Page 7: Understanding customer preferences for airport access ... · rail services between Heathrow Airport and central London’s Paddington Station. Paddington is one of London’s major

JEVONS, HOE AND SHEPHERD

232 © HENRY STEWART PUBLICATIONS 1750-1938 AIRPORT MANAGEMENT VOL 5, NO. 3, 226–238 APRIL–JUNE 2011

robust results. Of course, the design of thesurvey was one of the key aspects of theanalysis, which should be carefully con-sidered in any analysis of this type toensure that the survey data are as robust aspossible.

The use of this technique enables largevolumes of data to be compiled fromrelatively few respondents, which pro-vides a cost-effective way of generatingthe data required for the analysis. In thiscase, the survey data were used to helpunderstand how passengers might changetheir access decisions should differentfactors vary. For example, how many pas-sengers would change their mode oftravel to the Connect service if the fre-quency were to be increased. Economistsuse the concept of elasticities to describethese changes, where the elasticity ofdemand with respect to A (price, fre-quency etc) is defined as the percentagechange in demand divided by the per-centage change in A. Stated-preferencesurveys enable researchers to estimatethese elasticities by fitting a statisticalmodel to the data.

Analysing survey dataA number of different statistical modelscan be used to analyse the results ofstated-preference surveys. In this case, themultinomial logit model was used.1 Thismodel works by assigning a probabilityto a respondent picking a particularmode of transport to access Heathrow, egExpress or taxi.The probability of a cus-tomer using a particular mode can beinfluenced by the characteristics of therespondent, eg age and journey purpose.Using this model, an equation was esti-mated to model the probability that aconsumer travels by a given mode as afunction of access time, access cost, egresstime, journey time, access journey fare,

main journey fare and income. Equation1 is as follows:

Prob(Yi = j )= f (access time, egress time,journey time, access journey (1)fare, main journey fare, income, ...”)

Equation 2 presents how the probabilityof mode choice is calculated using themultinomial logit approach:

(2)

where Yi is the mode chosen by passengeri to access the airport of all the optionswhich were available to them; j, �j are themodel parameters for option j; and xi ischaracteristic x for passenger i, eg journeytime. Note that the stated-preferencesurvey changes the values of x slightly foreach consecutive question (where appro-priate). Elasticities can be derived usingthe estimates of the model parameters(�s), which were used in the forecastingmodel as described below. The analysiswas conducted for 24 separate demo-graphic groups, ie a separate equation wasestimated for each demographic group.This enabled the identification of groupsthat were particularly sensitive to changesin different variables, eg high-incomeleisure travellers are less sensitive tochanges in travel cost than low-incomeleisure travellers. When the statisticalanalysis was completed, the elasticity esti-mates were used as one input into theinteractive spreadsheet model, which pro-duced demand and revenue forecasts.

Interactive modellingOutputs of the statistical analysis andadditional third-party data, eg fromHeathrow, were the essential inputs to theforecasting model. The model itself was

Prob Y j expexp

ix

x

j i

j ij 1

N( )=

=

=′

′∑

β

β

JAM189.qxd 4/7/11 5:12 PM Page 232

Page 8: Understanding customer preferences for airport access ... · rail services between Heathrow Airport and central London’s Paddington Station. Paddington is one of London’s major

UNDERSTANDING CUSTOMER PREFERENCES FOR AIRPORT ACCESS

© HENRY STEWART PUBLICATIONS 1750-1938 AIRPORT MANAGEMENT VOL 5, NO. 3, 226–238 APRIL–JUNE 2011 233

designed as an interactive spreadsheet.Figure 3 illustrates the model structure.

As Figure 3 shows, the two main inputsare the survey and the Heathrow passen-ger forecasts. Analysis of the former leadsto data on current market shares and esti-mates of consumers’ sensitivity to modalattributes. The latter provides the modelwith a baseline for the total number ofpotential passengers that could ever takeHeathrow Express services in a givenmonth.As this input is itself a forecast, it issubject to its own uncertainty and marginof error. Necessarily, this uncertainty alsoapplies to the eventual forecasts forHeathrow Express services that are basedon the Heathrow passenger numbers. Byanalysing the historical passenger numberdata, the trends and seasonal patterns of airdemand have been incorporated into theforecasts.

Estimates of how sensitive customersare to changes in modal attributes and thecurrent market shares of each transportmode allow the model to be calibrated,which is the final process that combinesall the input data. Calibration ensures thatthe results from the statistical analysis arealigned with the factual market shares. Inpractice, an optimisation algorithm is iter-atively applied until the model minimisesits forecast error in the current period.

Baseline forecasts — ie estimates ofhow many passengers will travel on eachtransport mode, assuming that no factorsother than airport traveller numberschange each month — are calculated byapplying the market shares to the rele-vant proportion of the Heathrow pas-senger forecasts.This gives a benchmarkof future passengers and revenues foreach mode.

Figure 3 Stylised forecasting model — Inputs, analysis and outputsSource: Oxera

JAM189.qxd 4/7/11 5:12 PM Page 233

Page 9: Understanding customer preferences for airport access ... · rail services between Heathrow Airport and central London’s Paddington Station. Paddington is one of London’s major

JEVONS, HOE AND SHEPHERD

234 © HENRY STEWART PUBLICATIONS 1750-1938 AIRPORT MANAGEMENT VOL 5, NO. 3, 226–238 APRIL–JUNE 2011

The scenario forecast is calculated bysimulating changes in the market andevaluating the implications for the marketshares, and thus for passenger numbersand revenues. For business planning, vari-ous scenario forecasts can be comparedwith the baseline (‘no change’) forecast.This allows an assessment of the futurerevenue streams. Examples of parametersin the model that can be adjusted to sim-ulate future market changes include:

● changes in ticket prices for each modeof transport;

● changes in travel time and frequency ofservice for each mode of transport;

● introduction of additional ticket typesfor Express and Connect;

● airline migration between terminals;● changes in income levels between

countries.

Some of these are parameters over whichHeathrow Express has control (eg pric-ing), while others are parameters that mayhave changed since the time of the survey(eg coach frequency). Each simulationcan include one or more of these aspectsin the forecast passenger numbers andrevenues.

FINDINGSThis section is split into findings from thesurvey and findings from the forecastingmodel.

Survey evidenceEvidence from each section of the surveyis considered separately below.

Revealed-preference findingsThe revealed-preference questions provideinformation on the factual behaviour of

customers in the market at the point intime of the survey. This was a usefulmarket research exercise in itself, and wasalso used to personalise the stated-prefer-ence questions to each respondent. Themost common mode of transport wasfound to be the London Undergroundnetwork, with 42 per cent of surveyrespondents having used that mode on theday of the survey. This was followed byminicab (20 per cent) and Express stan-dard class (17 per cent). The majority ofpassengers originated their Heathrowjourneys in central London (88 per cent);the remainder of passengers originatedtheir journey outside London, but camevia a mainline London rail station.WithinLondon, the Westminster region wasstated as the most prevalent journeyorigin (22 per cent), compared with thewest central (15 per cent) and the north-east central and City (14 per cent) areas.Figure 4 illustrates the full modal andorigin splits at the time of the survey.

Another important dimension revealedby the survey was the split between busi-ness and leisure passengers. Differences inthe time and money trade-offs that indi-viduals make were expected between thetwo groups. These innate preferences arelikely to affect modal choices and the levelof response to changes in the market (egprice rises). The survey revealed that 77per cent of customers were travelling forleisure purposes.Variation in this propor-tion by current travel mode providesindicative evidence of the differences inthe time and money trade-off betweenthe two groups (see Figure 5).

The variation in business and leisuremix by current choice of transport washigh, ranging from almost 50 per centbusiness on Express standard class to 0 percent for coach. Figure 5 shows a roughpattern between the business passengerproportion and average journey time,

JAM189.qxd 4/7/11 5:12 PM Page 234

Page 10: Understanding customer preferences for airport access ... · rail services between Heathrow Airport and central London’s Paddington Station. Paddington is one of London’s major

UNDERSTANDING CUSTOMER PREFERENCES FOR AIRPORT ACCESS

© HENRY STEWART PUBLICATIONS 1750-1938 AIRPORT MANAGEMENT VOL 5, NO. 3, 226–238 APRIL–JUNE 2011 235

with the faster modes of transport gener-ally attracting more business passengers.For instance, the average business propor-tion for modes taking under or aroundone hour on average was 29 per cent,compared with 17 per cent for the othermodes; however, there were exceptionsto this such as the car options, being fastbut having low proportions of businesspassengers. The revealed-preference dataalone were not sufficient to untangle the

factors that determine individuals’ modalchoices.

Stated-preference findingsThe stated-preference exercise allowedcustomer preferences and their sensitivityto different modal characteristics to bequantified. This approach goes beyondobserving statistics such as those presentedin Figure 4. Instead, it attempts to isolate

Figure 4 Modal split and origin split estimated at the time of the survey (August 2009)Source: Oxera analysis

Figure 5 Business and leisure splits by transport modeNote: HEx is the Express service; HC is the Connect service; LU tube is London Underground tube. Park and fly includes the averagecost of parking for a two-week period.Source: Oxera analysis

JAM189.qxd 4/7/11 5:12 PM Page 235

Page 11: Understanding customer preferences for airport access ... · rail services between Heathrow Airport and central London’s Paddington Station. Paddington is one of London’s major

JEVONS, HOE AND SHEPHERD

236 © HENRY STEWART PUBLICATIONS 1750-1938 AIRPORT MANAGEMENT VOL 5, NO. 3, 226–238 APRIL–JUNE 2011

the individual impact of specific modalattributes (eg price) on decisions. At ahigh level, this part of the survey high-lighted customers’ strong preferences fortheir current mode of transport.This wasreflected in the finding that the majorityof passengers were not swayed away fromrepeatedly selecting their current choice.

Brand value and loyalty schemes areone potential explanation for this modalallegiance. The Express service hasinvested significant sums in marketingover the past five years, and is now listedas one of the top 500 superbrands in theUK.2 Such branding is expected toenhance the loyalty of customers. LondonUnderground, on the other hand, offersits users a discount for purchasing ticketsthat cover periods of time (travelcards).Passengers who already have such a ticket(eg for inter-London commuting) cantravel to Heathrow without incurringany additional cost and thus effectivelytravel without incurring further cashexpense. The fact that many Londonerswill typically have this effective discountwill increase their allegiance to theLondon Underground. (Oyster cardswere also accounted for in the analysis byreflecting this lower (but not zero) cost inthe offerings.)

A second explanation for the strongpreferences may be the fact that manygroups of customers prize certain modalattributes. Given the strong differences inmodal attributes — see Table 1 — thisleads to many customers always choosingbased on one factor. For example, a jour-ney from the most common origin,Westminster, is expected to take around60 minutes on average via LondonUnderground and 45 minutes via Express(including an initial trip to Paddington).If a passenger’s priority is time, a largeincrease in journey time would be neces-sary before the modal choice would be

affected. Alternatively, if a passenger’s pri-ority is ease of use (eg a non-residentbusiness user) then a minicab, at around45 minutes, may offer strong advantagesthat price and time changes will notaffect. Furthermore, even though the taxijourney is comparable with the Expressservice in terms of time, the journey timevia taxi is considerably less certain. Thefact that taxis can command up to doublethe Express fare reflects these otherfactors (ease of use, comfort, no accesstime or cost).

The statistical analysis of the stated-preference exercise leads to elasticity esti-mates. Estimates are calculated fordifferent demographic sub-groups of therespondents. The sensitivity of customersto time and fare, sometimes referred to as‘elasticities’, are shown in Figure 6. Thenumbers in Figure 6 show the likelyreduction in demand for a mode of trans-port given a particular change in journeytime or fare. For example, in the case ofhigh-income business travellers, a £10increase in travel cost is estimated to havean impact of –7 per cent on demand.Using the full range of estimates, it ispossible to estimate to which mode thereduced demand for one mode of trans-port would switch.There are several char-acteristics to note from the estimates:

● All passengers are more sensitive tochanges in price than changes in time.

● Leisure passengers are more sensitive tocost than business passengers.

● Business passengers are more sensitiveto time than leisure passengers.

● Low-income passengers are moresensitive to cost than high-incomepassengers.

● The relative sensitivity to cost and timeis greater for leisure passengers thanbusiness passengers. For example, forhigh-income leisure passengers, it

JAM189.qxd 4/7/11 5:12 PM Page 236

Page 12: Understanding customer preferences for airport access ... · rail services between Heathrow Airport and central London’s Paddington Station. Paddington is one of London’s major

UNDERSTANDING CUSTOMER PREFERENCES FOR AIRPORT ACCESS

© HENRY STEWART PUBLICATIONS 1750-1938 AIRPORT MANAGEMENT VOL 5, NO. 3, 226–238 APRIL–JUNE 2011 237

would require a change in travel timeof around ten times larger than a changein travel cost to elicit a similar demandresponse; for high-income business pas-sengers, it would only require a change intravel time of around double that of travelcost to elicit a similar demand response.

Elasticities were calculated for a widerrange of demographic groups than pre-sented in Figure 6. Group size (1, 2, 3�passengers) and UK resident status werealso used to distinguish between respon-siveness levels. These elasticities drive theforecasting model’s ability to predict howchanges in the time and fare values willaffect future demand for each mode.Therelative size of the demographic groupshas been determined by the revealed-preference part of the survey. Implicitly,the forecasting model assumes that theseproportions remain constant from thetime of the survey and do not vary overthe course of the forecast period.Repeating the survey exercise at different

points during the year and at regularintervals (to detect any shifts in the demo-graphics over time) would allow thisassumption to be relaxed.

Insights from forecastingThe complete forecasting model allows ascenario analysis of any future or potentialchanges that may affect the Heathrowaccess market. This model is being usedby Heathrow Express in its business plan-ning. Insights from its usage to date sug-gest three key messages that may hold forother airports:

● The competitive landscape is one of the mostimportant drivers of passenger access deci-sions. Heathrow’s highly differentiatedairport access options mean that cus-tomers display strong allegiance to par-ticular modes. This has been enhancedby brand value and loyalty schemes.Changes to fare or journey time withina reasonable range (up to 30 per cent)

Figure 6 Responsiveness of different customer groups to changes in journey time and travel costNotes: Elasticities are approximate averages across all modes of transport. Income groups were specified relative to the median in the sample.Source: Oxera analysis

JAM189.qxd 4/7/11 5:12 PM Page 237

Page 13: Understanding customer preferences for airport access ... · rail services between Heathrow Airport and central London’s Paddington Station. Paddington is one of London’s major

JEVONS, HOE AND SHEPHERD

238 © HENRY STEWART PUBLICATIONS 1750-1938 AIRPORT MANAGEMENT VOL 5, NO. 3, 226–238 APRIL–JUNE 2011

may not have a major impact on themodal choices. This assumes that cus-tomers respond in a rational fashiononly and perceptions of unfair and/orrepeated fare rises may damage the brandvalue; however, new modal options maycause major shifts in modal choices.

● Airline migration between terminals mayhave a noticeable effect on customer choices.The geographic features of HeathrowAirport mean that travel times fromcentral London vary substantially withthe destination terminal. For example,the egress time from the coach station isaround five minutes for Terminals 1–3compared with more than 15 minutesfor Terminals 4 and 5. Other modes alsohave egress times that vary by terminaland, for the faster modes of transport(eg Express service), this can materiallyaffect the total journey time. Terminalclosures or airline migration can resultin large groups of customers beingsubjected to changes in journey times,which has implications for both airportplanning and the design of airportaccess transport.

● Changes in the wider economy may onlyhave a limited impact on modal choices.Customers’ modal choices are lessresponsive to changes in aggregateincome measures (eg GDP) relative toother factors (eg fares). For most trans-port modes, income was estimated ashaving no role in the decision-makingprocess.This suggests that, although thedemand for air travel as a whole may

well be closely linked to the economicclimate, choices between airport accesstravel modes may not be. For example,if air travel falls during a recession, thedemand for airport access as a whole islikely to reduce, but for those passengerswho do travel, their modal choice maybe unaffected.

CLOSING REMARKSThis paper has illustrated how consumersurveys can be used to understand con-sumer decision making. The techniquesthemselves are more widely applicableand can be used across a range of differentcontexts. Combining survey techniqueswith statistical analysis can lead to sophis-ticated business planning tools. Such toolscan help airport access operators (and,indeed, airport operators and users moregenerally) to plan for the future.

Acknowledgments

The authors would like to thank DrStephane Hess, Institute of TransportStudies, Accent, and Elizabete Ernstsone,Oxera, for their help with this paper. Ofcourse, any errors or omissions are thesole responsibility of the authors.

References(1) Hensher,A., Rose, J. and Greene,W. (2005)

‘Applied Choice Analysis:A Primer’,Cambridge University Press, Cambridge.

(2) Business Superbrands 2011: Heathrow Express,see: http://www.superbrands.uk.com/heathrowexpress (accessed 28th January, 2011).

JAM189.qxd 4/7/11 5:12 PM Page 238