DidAttitudesInterpretandPredict“Better”Choice...

Research ArticleDid Attitudes Interpret and Predict ldquoBetterrdquo ChoiceBehaviour towards Innovative and Greener AutomotiveTechnologies A Hybrid Choice Modelling Approach

Stefano de Luca and Roberta Di Pace

Department of Civil Engineering University of Salerno Italy Fisciano Italy

Correspondence should be addressed to Stefano de Luca sdelucaunisait

Received 5 December 2019 Revised 29 June 2020 Accepted 3 July 2020 Published 1 August 2020

Academic Editor Ludovic Leclercq

Copyright copy 2020 Stefano de Luca and Roberta Di Pace is is an open access article distributed under the Creative CommonsAttribution License which permits unrestricted use distribution and reproduction in anymedium provided the original work isproperly cited

It is common opinion that traditional approaches used to interpret and model usersrsquo choice behaviour in innovative contexts may lead toneglecting numerous nonquantitative factors that may affect usersrsquo perceptions and behaviours Indeed psychological factors such asattitudes concerns and perceptions may play a significant role which should be explicitly modelled By contrast collecting psychologicalfactors could be a time and cost consuming activity and furthermore real-world applicationsmust rely on theoretical paradigmswhich areable to easily predict choicemarket fractions e present paper aims to investigate the above-mentioned issues with respect to aninnovative automotive technology based on the after-market hybridization of internal combustion engine vehicles In particular threemainresearch questions are addressed (i) whether and how usersrsquo characteristics and attitudes may affect usersrsquo behaviour with respect to newtechnological (automotive) scenarios (eg after-market hybridization kit) (ii) how to better ldquograsprdquo usersrsquo attitudesconcernsperceptionsand in particular which is themost effective surveying approach to observe usersrsquo attitudes (iii) to what extent the probability of choosing anew automotive technology is sensitive to attitudesconcerns changes e choice to installnot install the innovative technology wasmodelled through a hybrid choice model with latent variables (HCMs) starting from a stated preferences survey in which attitudes wereinvestigated using different types of questioning approaches direct questioning indirect questioning or both approaches Finally acomparison with a traditional binomial logit model and a sensitivity analysis was carried out with respect to the instrumental attributes andthe attitudes Obtained results indicate that attitudes are significant in interpreting and predicting usersrsquo behaviour towards the investigatedtechnology and the HCMmakes it possible to easily embed psychological factors into a random utility modelframework Moreover theexplicit simulation of the attitudes allows for a better prediction of usersrsquo choice with respect to the Logit formulation and points out thatusersrsquo behaviour may be significantly affected by acting on usersrsquo attitudes

1 Introduction and Motivations

e diffusion and market penetration of new technologies isbecoming a crucial factor for transport system analysts anddecision-makers Among the several new technologies aiming torender the transportation system more efficient and sustainableElectric Vehicles (EV) or Hybrid Electric vehicles (HEV) rep-resent one of the possible solutions but although electrictechnologies are consolidated and reliable two main issuescontinue to be challenging tasks interpreting and modellingusersrsquo behaviour towards these new technologies and assessingthe potential environmental impacts

Within the cited context it is common opinion thattraditional approaches used to interpret and model usersrsquochoice behaviour may lead to neglecting the numerousnonquantitative factors that may affect usersrsquo perceptionsand behaviours

As a matter of fact psychological factors such as atti-tudes concerns and perceptions may play a significant rolewhich should be explicitly modelled On the other handcollecting psychological factors could be a time and costconsuming activity and furthermore real-world applica-tions must rely on theoretical paradigms able to easilypredict choicemarket fractions

HindawiJournal of Advanced TransportationVolume 2020 Article ID 5135026 22 pageshttpsdoiorg10115520205135026

Over the last two decades usersrsquo propensity to choosealternative fuel vehicles (AFVs) has been widely investigatedand many recent analyses have pointed out the necessity toalso take into account nonobservable variables such as theperceptions and the attitudes of the users (see [1]) along withdirectly observablemeasurable attributes Indeed eventhough models usually adopted in demand modelling aresuitable for the representation of the choice process thesecannot be applied to the representation of perceptions andattitudes [2]

Based on previous considerations several researchershave proposed the integration of latent variables withintraditional econometric frameworks such as the utilitariantheoretical paradigm erefore Latent variables HybridChoice Models (HCMs) have been adopted in several casesand in particular they have been applied to capture theindividual attributes such as attitudes and habit influencingthe mode choice [3ndash10] the residential location choice[11 12] the route choice decision making [13ndash15] and thevehicles choice (see Section 2)

e present paper starting from some preliminary re-sults proposed in de Luca and Di Pace [16] aims to in-vestigate the role of attitudinal factors in the choice of a newautomotive technology the HySolarKit which aims toelectrifyhybridize existing vehicles through an after-marketkit which can be recharged by the grid as well as by solarpower [17ndash21]

e paper first proposes a literature review on the mostsignificant contributions on usersrsquo intention to purchaseelectrichybrid automotive technologies with particularattention to the psychological factors and the possiblemodelling approaches

It then addresses the following three main researchquestions (e main aim is not to investigate the potenti-alities of the HySolarkit but to comprehend the role ofattributes different from typical instrumental attributes)

(1) Whether and how usersrsquo characteristics and attitudesmay affect usersrsquo behaviour with respect to newtechnological (automotive) scenarios (eg after-market hybridization kit) thus if usersrsquo choiceshould be investigated through more advanced andbehaviourally complex models

(2) How to better ldquograsprdquo usersrsquo attitudesconcernsperceptions and in particular which is the mosteffective surveying approach to observe usersrsquo atti-tudes through which type of questions (eg direct orindirect) and through which questions

(3) To what extent the probability of choosing a newautomotive technology may be significantly affectedby attitudesconcerns changes

e above-mentioned issues were addressed through theimplementation of an ldquoad hocrdquo experiment and a modelbased analysis aimed to infer the role of psychological factorsthrough the specification of latent variables within a Ran-dom Utility theory modelling framework

To this aim a Stated Preferences (SP) survey was carriedout on a sample of potential consumers [22] In particular

the survey aimed to collect the propensity to install theHySolarKit but it was specifically designed to ldquograsprdquo fivedifferent attitudesconcerns through a direct questioningapproach (attitudes towards the environment design fuelconsumption technology and reliability of technology)

Different types of questions were tested ldquodirectly relatedto the alternativerdquo and ldquoindirectly related to the alternativerdquoquestions (details in Section 42) Indeed existing literaturemakes use of direct questioning but usually uses alternativerelated questions only [23 24] whilst there are fewer in-stances of resorting to a mix of alternative and non-alternative related questions Experimental results were firstanalysed to investigate the robustness of the adopted hy-pothesesquestions then they were used for calibratinghybrid choice models (HCMs) with latent variables fol-lowed by a traditional binomial Logit model

e estimation results made it possible to understandthe effectiveness of the different surveying methods andthe role of psychological factors in the choice behaviourand also allowed for investigating if and how the in-clusion of attitudinal factors significantly increaseschoice model goodness-of-fit thus leading to much moreeffective results than traditional Logit choice modelsFinally a sensitivity analysis was carried out for assessingthe role of traditional instrumental attributes (egmonetary cost) as well as the impacts that the change ofusersrsquo attitudes may have on the propensity to install thenew technologies

e paper is organised as follows Section 2 introducesthe state of the art on existing approaches and modellingsolutions to the simulation of the propensity to choose newautomotive technologies Section 3 introduces the theoret-ical frameworks whereas Section 4 describes the experi-mental framework introducing the case study the datacollection and the preliminary analyses Finally Section 5discusses the estimation results e main conclusions aresummarised in Section 6

2 Intention to Choose Electric or HybridElectric Automotive Technologies ModellingApproaches and Determinants

e interest in alternative fuel vehicles (AFV) relies on awide range of literature whose main focus is that of Electric(EVs) Hybrid (HEVs) or Plug-In Electric vehicles (PHEVs)

As highlighted in studies of de Luca et al [16] andCartenı et al [25] the choice phenomenon cannot be solelyinterpreted within a unique theoretical framework As amatter of fact eliciting the ldquopreferencerdquo for AFVs is quitecomplicated due to several factors such as the recentadoption of such technologies the unavailability of salesdata the difficulty of estimating the ownership costs and theimpacts on driving habits

Behavioural approaches on EVs choice mainly rely ontwo mainstreams theories (i) psychological and sociologicaltheories (eory of Planned BehaviourmdashTPB Value-Belief-Norm theorymdashVBN habits diffusion of innovation dif-fusion theory) and (ii) Consumer Choice eory (CCT)

2 Journal of Advanced Transportation

Psychological approaches made it possible to compre-hend the complexity of the phenomenon allowing for theunderstanding of the main determinants that may affect EHPHEVs consumer adoption [26ndash34]

e cited psychological approaches have proposedqualitative and quantitative results grounded on descriptivestatistical analyses making it possible to comprehend themain barriers and determinants that may affect EHPHEVsconsumer adoption e main results applied to EVs in-dicated the significance of hedonic and symbolic attributesor emotions and feelings [30] (instrumental attributes referto the functionality or utility that can be derived fromfunctions performed by new technologies hedonic attri-butes refer to the pleasure of using a new technologysymbolic attributes are related to a sense of self or socialidentity that is reflected by the possession of new technol-ogies) factors expressing self-identity andor lifestyle[30 32 35 36] and pleasure or enjoyment gains andminimisations of negative effects when making journeys[37ndash39]

CCT can count on a wide range of models and appli-cations investigating alternative fuel vehicle (AFV) marketpenetration andor consumer behaviour determinants[40ndash46]

e most significant contributions on EVs are mainlygrounded on Random Utility eory and starting from theyear 2000 different modelling solutions have been success-fully proposed the Rank ordered model [47] the Mixed(Hybrid) Multinomial Logit models [48ndash50] theMultinomialLogit models or the Nested Logit models [51ndash55] themultiplediscrete-continuous model [56] and the utility-based and aregret-based model of consumer preferences [57]

Proposed analyses together with typical instrumentaland easily quantifiable attributes (eg range maintenancedriving costs and ownership costs) point out that otherattributes may play a significant role

Indeed the decision to installpurchaseadopt an EPEPHEVs depends on several attributes of different typesinstrumental environmental hedonic and symbolic[35 58ndash62]

EHPHEVsrsquo are usually judged with respect to the in-strumental and functional attributes purchase price run-ning costs reliability performance driving range andrecharging time performance convenience comfort andaesthetics [16 22 35 36 40 53 63ndash68]

Recently an increasing number of contributions havepointed out the role of noninstrumental attributes such ashousehold socioeconomic characteristics [48 50 69] socialinfluence [70] consumer emotions and feelings [28 30 36 39]proenvironmental behaviour [71 72] and [73] consumeradoption of innovations [74 75] governmental incentive [50]and hedonic and symbolic attributes [35 58ndash61]

Among the wide variety of the existing contributionssignificant efforts have been made to analyse the attitudesthat potentially may affect the usersrsquo choice process ofpurchasing an EHPHEVsrsquo

Along with different approaches founded on psycho-logical paradigms but which are not able to foresee theadoption of innovative technologies Hybrid Choice models

(HCMs) with latent variables may allow for the role ofperceptions and attitudes to emerge and at the same timeestimate the probability of choosingadopting the technology

One of the first contributions was by Choo andMokhtarian [76] who analysed the attitudes lifestylespersonalities and mobility which more generally may affectconsumersrsquo choice in identifying the vehicle type In par-ticular seven groups of significant variables were identified(i) the objective mobility such as commute time travelleddistance and frequency of trips (ii) the subjective mobilityfocusing on how that amount of travel is perceived (iii) howrespondents enjoy travelling themselves (iv) the attitudes interms of travel land use and environment (v) the per-sonality and (vi) the lifestyle

A contrasting result was proposed by Kishi and Satoh[77] for the Tokyo and Sapporo residents which thoughwell aware of environment concerns were not inclined topurchase low-pollution cars

Proenvironment propensity was derived by Potoglouand Kanaroglou [52] who observed how the propensity toadopt cleaner vehicles of Canadiansrsquo households was mainlyaffected by the reduction in the monetary costs and purchasetax reliefs along with the low emission rates

Bolduc et al [48] starting from a survey conducted bythe Energy and Materials Research Group (Simon FraserUniversity 2002 and 2003) identified two main latentvariables which have a positive impact on the choiceprobability of the green technology alternatives (i) theenvironmental concern related to transportation and itsenvironmental impact and (ii) the appreciation of new carfeatures related to car purchase decisions

A similar study conducted by the same researchers [78]still observed that an environmentally conscious consumerwould prefer a cleaner automobile technology associatedwith less environmental impact

Jensen [79] carried out a survey based on direct ques-tions and unlike other research studies they included at-tributes regarding charging possibilities and battery lifetimein the experiment and users were also provided with in-formation on recharging types and environmental impactsFurthermore they observed that purchase price (the fullprice paid for the car considering everything) driving range(the maximum distance that could be covered with a fulltank or a fully charged battery) top speed fuel cost batterylife and charging in city centres and train stations have asignificant effect on usersrsquo choices

In [80] Jensen et al investigated the usersrsquo attitudesthrough 27 statements regarding new technologies theenvironment car interest and EVs in general ey carriedout a before and after analysis modelling the choice betweenconventional and electric cars and found that environmentalconcern had a positive effect on the preference for EVs bothbefore and after the test period

In the same year Glerum et al [50] provided a meth-odology based on HCM to forecast the demand for electriccars ey considered an SP survey collected within a projectbetween Renault Suisse S A and EPFLrsquos TransportationCenter (TraCe) In particular respondents were providedwith some statements based on alternative related questions

Journal of Advanced Transportation 3

in order to collect the importance of car design the per-ception of leasing the perception of an electric vehicle as anecological solution the attitude towards new technologiesand the reliability security and use of an electric vehicleeresults indicated that only two attitudes were significant theproleasing attitude and a proconvenience attitude (spa-ciousness comfort and new propulsion)

Soto et al [81] investigated four technological alterna-tives and identified four latent variables the environmentalconcern the support for transport policies and the attitudetowards technology and car Unlike other authors duringthe experiment they provided users with statements basedon direct and nonalternative related questions

More recently Kim et al [23 24] investigated the in-tention to purchase an electric car and observed that theattitudes playing a role regarded the environmental theeconomic the battery the technological aspects and theinnovation value To this aim a questionnaire based onnonalternative related questions was carried out Moreoverthey also considered socioeconomic characteristics in thestructural model and the price of EC the cost electricityrelative to gas the cruising range of the car the time to chargethe battery the maximum speed car and the distance tocharging station as alternative attributes in the logit model

Similar results were obtained by Petsching et al [34] whoinvestigated the interest for hybrid vehicles through theparadigm of the ldquoeory of Innovation Adoptionrdquo andidentified the significant role of the perception of innovationcorrelated to factors such as ecology design profitabilityand ease of use

Finally Tsouros amp Polydoropoulou [82] investigated achoice context including four alternatives (ie a hybrid car anelectric car a diesel car or a standard car) through an SPexperiment and observed environmental awareness as the onlysignificant attitudeey considered the vehiclesrsquo characteristics(eg the engine size the vehicle type and the vehicle edition)and the individualsrsquo socioeconomic characteristics for thestructural equation and depicted the attitude through indicators(indicators of eco-friendliness) based on nonalternative relatedquestions which include perceptions regarding ecological habitssuch as recycling active transportation and how they perceiveeach fuel type As in the studies of Bolduc et al [48] Daly et al[83] and Soto et al [81] a comparison of the obtained resultswith the MNL model is provided confirming that HCM ex-haustively outperforms MNL

Although the comparative analysis among differentcontributions is not straightforward since the surveys(which are not always discussed in depth) and the corre-sponding questionnaires are not comparable and the casestudies are different with regard to EV technologies pene-tration some conclusions may be drawn

Indeed it is interesting to highlight how the attitudetowards the environment is present in the greatest numberof studies In general almost all the existing analyses arefounded on SP experiments and make use of alternativerelated [48 50 79] or nonalternative related questions[23 24 84] to ldquograsprdquo usersrsquo latent characteristics

Latent variables are usually used for describing attitudesbut also concerns andor perceptions and all the

contributions except for the proenvironmental attitudeinvestigate different sets of attitudesperceptionsconcerns

With reference to the identification of modelling ap-proach effectiveness most of the contributions do not pro-pose any comparison whilst Bolduc et al [48] Soto et al [81]and Ioannis et al [84] highlighted that hybrid choice modelsnormally outperform traditional random utility models

In the following table a synoptic framework of the stateof the art is proposed e main contributions discussedabove have been classified with respect to the adopted the-oretical paradigm the investigated case study and the usedattributes (instrumental and noninstrumental) (Table 1)

3 Theoretical Framework

e modelling approach adopted in this paper was based onthe Hybrid Choice Model focusing on the incorporation oflatent variables into discrete choice models

Seminal studies aiming to overstep the boundary ofstandard discrete choice models were conducted by Ortuzarand Hutt [85] and McFadden [86] in which during the 80sthey investigated the possibility of including subjectivevariables in a discrete choice modelling [1 8 9 87ndash97]

Traditionally hybrid choice models are composed of thelatent variable (LV)model and the choicemodel In general therepresentation of parameters related to the psychological modelmay be pursued through different approaches (i) by includingthe indicators directly in the utility function [98ndash100] (ii) byconsidering a sequential method in which at the first stagelatent variables are represented through the Explanatory Fac-torial Analysis (EFA) then these are directly included in theutility function [101] and (iii) by the MIMIC model (MultipleIndicator Multiple Cause [102] in which the relationship be-tween attitudes (latent variables) and sociodemographic char-acteristics is formalised by the structural equation and therelationship between attitudes and perception indicators isformalised by measurement equation then latent variables aredirectly incorporated in discrete choice model

In this paper the MIMIC approach has been adoptedthen the utility in the (hybrid) choice model based on theassumption that each individual n (n 1 N) faced witha set of alternatives i (i 1 I) may be expressed as afunction of a vector of observed attributes Xni (representingthe level of service and the usersrsquo attributes) a vector oflatent variables LVni (Ktimes 1 if K statements are consideredfor each LV) and the error term εin independently andidentically distributed thus for each alternative the per-ceived utility may be expressed as

Uni βx middot Xni + βLV middot LVni + εni (1)

and the choice principle of the alternative considering thechosen yi alternative is based on utility maximisationcriterion

Regarding the LV let be p the generic latent variable tobe measured by psychometric indicators (p 1 P) and kbe the generic psychometric indicator (k 1 K) the latentvariable model consists in two equations for each latentvariable (LV) in particular for each individual n thestructural equation may be expressed as follows


LVni c + β middot Xni + ωni (2)

where c is the intersect Xni is the vector of the usersrsquocharacteristics β is the vector of the coefficients associatedwith the usersrsquo characteristics (to be estimated) and ωn bethe error term which usually is distributed with zero meanand σω standard deviation

Furthermore for each individual n K statements areused thus Ini is a vector of perceptions indicators (Ktimes 1)that are associated to the latent variable with the mea-surement equation given as follows

Ini α + λ middot LVni + vni (3)

where α is the intersect λ is the vector of coefficient asso-ciated with the latent variable (to be estimated) and vni is theerror terms usually assumed normally distributed with zeromean and σv standard deviation

Regarding the psychometric indicators they may be rep-resented in two different ways through continuous and discreteindicators depending on the adopted coding approachey areusually coded through the Likert scale [103] and the structural

equation modelling may be based on the ordered logit modelthe measurement is represented through a discrete variable andthe thresholds are the parameters to be estimated [83] A dia-gram of the modelrsquos specification is shown in Figure 1

In terms of the estimation procedure as previouslyoutlined two approaches may be distinguished the se-quential [3 104 105] and the simultaneous approaches [93]

e sequential approach solves the MIMIC model sepa-rately from the choice model thus two stages must be con-sidered one for estimating latent variables using theperceptions indicators and the other one for estimating theparameters in the choice model related to the latent variablesand to the typical variables However this approach wasdemonstrated as not being efficient (it cannot guaranteeconsistent and unbiased estimators) [43 91 106] e si-multaneous approach is based on a joint estimation procedure

In this paper the applied estimation procedure was basedon the simultaneous approach In general latent variables arenot directly observable indicators are introduced and anyinference must be based on the joint distribution whosedensity can be rewritten as

Table 1 Synoptic framework of the state of the art

Reference

Paradigms Investigated case study

Attribute(s)PST CCT RUT Others

AFVpurchaseintention

Environmentalbehaviour

Adoption ofinnovations

Beggs et al [40] middot middot

Instrumental and functionalattributes purchase pricerunning costs reliability

performance driving rangeand recharging time

performance conveniencecomfort and aesthetics

Bunch et al [63] middot middot

Cheron and Zins [64] middot middot

Ong and Hsselhoff [65] middot middot

Musti and Kockelman [66] middot middot

Graham-Rowe et al [36] middot middot

He et al [67] middot middot

de Luca et al [16] de Luca anddi Pace [22] middot middot

Plotz et al [69] middot middot + not-instrumentalattributes Household

socioeconomiccharacteristics attitudespersonality and lifestyle

Bolduc and Daziano [48] middot middot

Glerum et al [50] middot middot

Choo and Mokhtarian [76] middot middot

Potoglou and Kanaroglou [52] middot middot

Axsen and Kurani [70] middot

+ not-instrumentalattributes norms cognitiveemotions feelings motivessocial factorsinfluence

attitudes anticipated regret

Moons and de Pelsmacker[28] middot middot

Schuitema et al [30] middot middot


Steg [39] middot middot

Bamberg and Moser [71] middot middot

Onwezen et al [72] middot middot

Steg and Vlek [73] middot middot

Shih and Schau [74] middot middot

Watson and Spence [75] middot middot

Petsching et al [34] middot middot

Kishi and Satoh [77] Bolducand Daziano [48] Daziano andBolduc [78] Jensen [79]Glerum et al [50] Soto et al[81] Kim et al [23 24] Ioanniset al [84] Tsouros ampPolydoropoulou [82]

middot middot


g Ini( 1113857 1113946 RLVni

Ini

LVn

λ1113888 1113889hLVni

LVni

Xni

β1113888 1113889dLVni (4)

where RLVniis the range space of the vector of the latent

variablese joint probability of observing the choice yni may be

expressed as

P yniIni

Xni

βx βLV1113888 1113889 1113946RLVniP

yni

Xni

βx βLV1113888 1113889

middot fIni

Ini

LVn

λ1113888 1113889hLVni

LVni

Xni

β1113888 1113889dLVni

(5)

where f(middot) is the probability density function of the per-ception indicators and h(middot) is the probability densityfunction of the latent variables

Parameters estimation is carried out by maximising thejoint likelihood of observed sequence of choices and theobserved answers to the attitudinal questions

L ΠiP yniIni

Xni

βx βLV1113888 1113889

1113946 RLVniΠiP

yni

Xni

βx βLV1113888 1113889fIni

Ini

LVn

λ1113888 1113889hLVni

LVni

Xni

β1113888 1113889dLVni

(6)

the estimation of this class of models requires multidi-mensional integrals with dimensionality given by thenumber of latent variables

Finally particular attention was paid to the implicationsrelated to the evaluation of attitudinal and perceptual var-iables [107 108] Indeed any change in the perceptionrsquosindicators may affect the LVsrsquo meaning to the extent that thewhole model must be reestimated [3 48] As stated in Vij ampWalker [109] twomain approaches may be adopted in orderto observe the choice outcomes the former formulating the

choice probabilities as a function of the observable variablesand the latter formulating the choice probabilities as afunction of both the observable variables and the mea-surement indicators

In this paper the second approach was adopted and thedistribution of the latent variables was assumed given by themeasurement equations

e model parameters were estimated throughPythonBiogeme [110] in which the Maximum SimulatedLikelihood is implemented [92 111]

4 Experimental Framework

41 Observing andMeasuring the Attitudes One of the mainissues related to the specification and estimation of an HCMrelies on how to observe and quantify usersrsquo attitudes[112 113] perceptions [114] or concerns

Attitudes refer to the usersrsquo characteristics and theirapproach in real life and society and may be not related tothe alternatives (nonalternative related attitudes) or relatedto the alternatives (alternative related attitudes)

Perceptions are usually interpreted as ldquoalternative re-latedrdquo and refer to the usersrsquo interpretation and reaction to astimulus [113] However concerns may be related to aspecific problemissue and they may depend on the (choice)context (for example the concern towards the environmentmay depend on the specific problemactivity carried out)

Since attitudesperceptionsconcerns are entities con-structed to represent underlying response behaviour theycannot be measured directly but they could only be inferredstudying behaviour which in turn might be reasonablyassumed to indicate the attitudes themselves

e behaviour may be one that occurs in a natural settingor in a simulated situation In general different approachesto measure attitudes may be pursued

(i) Direct Observations observing the ongoing behav-iour of people in the natural setting or directly asking

Explanatoryvariables

Structural relationshipβ

β

β

Latentvariables

Measurement relationship

Obesrved indicatorX1



Utility

Choiceindicators

Discrete choice model

Latent variable model

Errorterm

Errorterm

Error

Error

Error

λ1

λ2

λ3

Figure 1 Diagram of a hybrid choice model (HCM)


the respondents to state their feelings with regard tothe issue under study Direct observation of be-haviour is not practicable if we want to have data ona large number of individuals Moreover observa-tion of behaviour even when the behaviour is theoutcome of the attitude being studied may tell us thedirection of the underlying attitude (ie whether it ispositive or negative) but it cannot as easily indicatethe magnitude or strength of the attitude

(ii) Direct questioning asking as to what their feelingsare Direct questioning has been applied for studyingattitudes but it mainly serves for limited purpose ofclassifying respondents as favourable unfavourableand indifferent with regard to a psychological objectMoreover individuals may possess certain attitudesand behave accordingly but may not be aware ofthem us direct questioning or any other self-report technique will be of little avail if the re-spondent has no access to his own attitudinal ori-entations buried in the realms of the unconscious

Within direct questioning two further questioning ap-proaches may be pursued

(i) rough direct questions on the investigated attitude(eg how much is the environment important)

(ii) rough indirect questions (eg my home bulbs areenergy efficient)

In general direct questioning is the most pursued so-lution since it makes it possible to control the investigatedcontext (defining the scale of measurement) and it requiressmaller times and costs

In this interpretative context attitudes can be ldquograspedrdquothrough direct or indirect questioning but indirect ques-tioning seems the ldquomost correctrdquo approach whereas per-ceptions can be ldquograspedrdquo through direct questioning onlyand concerns can be ldquograspedrdquo through direct questioningonly

In this paper a direct questioning survey was designedand two different types of questions were submitted to therespondents ldquodirectly relatedrdquo (in the following directquestionsmdashD) and ldquoindirectly relatedrdquo questions (in thefollowing indirect questionsmdashI)

In particular the paper investigates the concerns andattitudesperceptionsconcerns towards the environmentthe vehicle design the fuel consumption the technologyand the reliability of technology

Direct questions aimed to investigate the usersrsquo concernstowards fuel consumption vehicle design environmenttechnology and reliability of technology Indirect questionsaimed to investigate the usersrsquo attitudes towards fuel con-sumption vehicle design and environment

An overview of the questions submitted to the re-spondents will be displayed in the following section

42 Case Study Survey Attributes and Preliminary Analysese analyses and model specifications were carried outwithin a research project supported by the University of

Salerno and regarding the city of Salerno (Salerno is thecapital city of Salerno province (region of Campaniasouthern Italy) situated 55 km southeast of Naples It hasa population of approximately 130000 54500 house-holds an area of about 60 km2 a residential densityof 2240 inhabitants per km2 and an average of 150cars per household Finally four transport modes areusually available car as a driver walking bus andmotorbike)

e questionnaire was built from an early survey [16]and inspired by the existing literature discussed in Section 2

e potential attributes worthy of interest were firstgrouped into five subcategories (i) socioeconomic attri-butes (ii) trip purpose type (iii) owned car characteristicsengine (iv) price compared to userrsquos conventional car and(v) psychological factors

e survey consisted of a sample made up of 700 (the sizewas defined coherently with the indications proposed byLouviere et al (2000)) and involved only respondentsowning at least one car and declaring that heshe had theauthoritypower to make decisions regarding household carownership (mainly householders) e questionnaire con-sisted of three parts

e first part aimed to gather information on familycharacteristics (geographical travel characteristics socio-economic etc) on respondentsrsquo concerns that usually affectthe decision to buy a specific vehicle (eg fuel consumptionenvironmental impact and vehicle design) and about thehousehold cars (fuel supply brand and vehicles age)

Moreover a first set of direct questions (D) were sub-mitted to the respondents (Table 2) As introduced in theprevious section the questions are directly related to en-vironment vehicle design fuel consumption technologyand reliability of technology

In this case each respondent was asked to rate in theLikert scale (1 null 2 mild 3 moderate 4 severe) howmuch each considered statement was important in thechoice of a car

In the second part indirect questions about the usersrsquoattitudes were submitted to respondents (Table 3) eyregarded psychological factors related to fuel consumption(Icons) vehicle design (Idesign) and environment (Ienv) In-direct questions were measured through a five preferencesrating scale (1 totally disagree 2 disagree 3 indifference 4agree 5 totally agree)

e third part investigated the propensity to install theHySolarKit

First of all the respondents were introduced to thetechnology and its main characteristics (see appendix forsome technical details) how it works how it is installed thedifferent performances (eg acceleration speed) and theenvironmental and fuel consumption benefits which can beachieved

To this aim each respondent was presented with a moreaccurate estimate of the benefits obtainable in terms of fuelconsumption (based on the type of trip on the number ofkilometres travelled and on the type of vehicle owned) Inparticular the weekly Δcost was calculated and the userswere asked to state the systematic and nonsystematic trip


characteristics Starting from the stated characteristics eachrespondent was faced with two different scenarios withdifferent installation costs (ranging from 500 to 4000 euros)e cost scenarios submitted to each respondent were in-dependent of one another

All the tested attributes are summarised in Table 4whereas Tables 5 and 6 show the descriptive and statisticalanalyses carried out on the preferences collected throughdirectindirect questions

In Table 5 together with the mean values the standarddeviations and Cronbachrsquos alpha test results are proposed

Means and standard deviations make it possible tounderstand the weight given by the respondents to eachquestion whereas Cronbachrsquos alpha measures how thequestions associated to each attitude are closely related toeach other as a group (internal consistency)

Obtained results made it possible to identify the ques-tions with higher dispersion with respect to the corre-sponding mean values Cronbachrsquos alpha tests confirmed thereliability of the chosen questions but also pointed out theadvisability of an exploratory Principle Factor Analysis(PCA) on all the indicators

e analysis made it possible to identify the correlationbetween the statements allowed for the identification of thelatent variables (factors) and thus the main statementsexplaining them Overall three latent variables were revealedas statistically significant and for each one of them therepresentative statements were identified (Table 5) Inparticular three factors corresponding to three differentlatent variables were clearly identified

(i) Factor 1 representing the attitudes towards fuelconsumption

(ii) Factor 2 representing the attitude towards the ve-hicle design

(iii) Factor 3 representing the attitude towards theenvironment

Observing the loading factors reported in Table 4 it isalso possible to derive the role and significance of the at-titudinal statements for each factor in factor 1 (ldquofuel con-sumptionrdquo) the significant statements were Qcons Icons123(see Tables 2 and 3 for a detailed description of statements)in factor 2 (ldquovehicle designrdquo) were Idesign134 (see Table 3 fora detailed description of statements) in factor 3 (ldquoenvi-ronmentrdquo) were Qenv Ienv2346 (see Tables 2 and 3 for adetailed description of statements)

Such results support an interesting interpretation of thephenomena but also represent an importantfundamentalinput for the specification of the Hybrid choice model whichwill be proposed in the following section

5 Results and Discussion

With regard to the experimental framework previouslyintroduced this section presents the main results obtainedfrom the specification and calibration of different HybridChoice Model (HCM) structures with latent variables Es-timation results are organised into two sections

e first section introduces the estimation results for theHCM and aims to propose a detailed analysis on the

Table 2 Overview of direct questions submitted to the respondents

Direct questions (D)Dcons One of the most important things in a car is the fuel consumption rateDdesign e car design is one of the most important factors in purchasing a carDenv I normally behave or act to reduce the environmental impact of my actionsDrel I prefer driving traditional fuelled vehicles since they guarantee a higher reliabilityDtechnology I am sensitive to all the technological features offered by a car

Table 3 Overview of indirect questions submitted to the respondents

Indirect questions (I)Icons1 e consumption and the energy class significantly influence my choice in purchasing an applianceIcons2 I am usually attentive to the special offers of electric operatorsIcons3 My home bulbs are energy efficientIcons4 I usually evaluate the car efficiency concerning the car cost mileageIcons5 I normally compare the fuel prices among different stations

Icons6When driving I am not willing to behave in a way that reduces the environmental impact (my driving behaviour is normally

aggressive)Idesign1 When parking I am usually careful to avoid having my car damagedIdesign2 I often read journals of designIdesign3 When furnishing I am willing to buy pieces with modern design features and original detailsIdesign4 I am willing to go to the body shop mechanic not only for major damagesIdesign5 I am willing to install not standard equipment (such as antitheft block shaft) in my own carIenv1 I often control the exhaustemission system of my carIenv2 I consciously do separate waste collection (recycling)Ienv3 I really enjoy spending my free time in parks green areas to breathe clean areaIenv4 How much do you agree with the following sentence We must act and make decisions to reduce emissions of greenhouse gasesIenv5 How much do you agree with the following sentence e government should invest in low energy impactIenv6 I am not willing to use the car during the weekend to protect the environment and then reduce air pollution


Table 4 Collected and investigated attributes

Attribute Meaning Type SI Min MaxAge Age of the respondent Continuous Years 24 70Masterrsquos degree Equal to 1 for users achieved this educational attainment Binary 0 1ZonRes Equal to 0 for users living to the historical centre 1 if in the outskirts Binary 0 1Diesel Power supply of the owned car Binary 0 1CarAge Age of the owned car on which the respondent would install the kit Continuous Years 1 10By car-shopping Mode choice car and trip purpose shopping Continuous mdash 0 093By car-personal services Mode choice car and trip purpose personal services Continuous mdash 0 093Interested in electricvehicle purchasing

Equal to 1 for users which declared to be interested in electric vehiclepurchasing Binary mdash 0 1

Conc_ConsumpConc_DesignConc_EnvironConc_ReliabConc_Tech

(i) Design issues concern(ii) Environment concern(iii) Reliability concern(iv) Technology concern

(v) Fuel consumption concern Binary attribute foreach scale mdash 0 1Each respondent was asked to rate how the fuel consumptionvehicle

designenvironmenttechnologyreliability of technology isimportant in the decision of which car to purchase e rating scaleand the value associated to each rate was null importance (1) mild

(2) moderate (3) and severe (4)

Att_Consump

Latent variable representing the attitude towards the fuelconsumption the rating scale and the value associated to each ratewas 1 Totally disagree 2 Disagree 3 Indifference 4 Agree 5 and

Totally agree

Continuous

Att_DesignLatent variable representing the attitude towards the vehicle designthe rating scale and the value associated to each rate was 1 Totallydisagree 2 Disagree 3 Indifference 4 Agree 5 Totally agree

Continuous

Att_EnvironLatent variable representing the attitude towards the environmentthe rating scale and the value associated to each rate was 1 Totallydisagree 2 Disagree 3 Indifference 4 Agree and 5 Totally agree

Continuous

Δcost

Δcost WCwithoutK ndash WcwithK Continuous euro minus44 172e weekly cost considering two scenarios with and without the kit was computed in order to define the usersrsquofinancial gainerefore each respondent was preliminarily informed on the upfront cost and successively heshe was also informed on the weekly cost (combining the fuel consumption the charging cost and the

installation cost)Obviously the cost estimation is based on the weekly kilometres travelled by each respondent

Table 5 Summary of the mean values and the standard deviations of the collected preferences and Cronbachrsquos alpha test

Mean sdFuel consumptionQconsmiddot 376 097Icons1 397 101Icons2 427 162Icons3 328 051Icons4 218 315Icons5 188 094Icons6 235 122

Cronbachrsquos alpha 0658DesignQdesignmiddot 321 092Idesign1 285 065Idesign2 176 079Idesign3 311 096Idesign4 408 094Idesign5 297 05mdash mdash 092

Cronbachrsquos alpha 0527


estimation results on the Latent Variables specification andon the resultant behavioural interpretation e results alsomade it possible to identify the most effective type ofquestions able to ldquograsprdquo usersrsquo attitudes

e second section investigates the differences betweenthe HCM and a traditional Binomial logit model (BLM) inwhich typical sociodemographic characteristics and in-strumental attributes were usede comparison was carriedout in terms of statistically significant attributes ability tointerpret the choice phenomenon goodness-of-fit andsensitivity to the monetary cost and to the attitudes

51 Hybrid Choice Model with Latent Variables EstimationResults eHCMwas specified with utility functions whichare linear in the attributes considering the attributes in-troduced in Section 4 and embedding the LVs within thesystematic utility functions To this aim both the structuraland the measurement equations were specified and jointlycalibrated on the preferences stated by respondents withreference to the questions introduced in Section 42

Overall the estimation results pointed out that thefollowing groups of attributes were statistically significantwith signs of the parameters consistent with the expectations(Table 7)

(a) e respondentsrsquo sociodemographic characteristics(b) e activity-related attributes(c) e level of service attributes

(d) e attitudinal attributes(e) e vehicle characteristics

It may be preliminarily observed that the following usersrsquospecific attributes were statistically significant age educa-tional level and the characteristics of the familyrsquos vehiclefleet Moreover the zone of residence the car age andhaving a diesel vehicle explicated usersrsquo behaviour

In terms of activity-related attributes significant attri-butes were travelling by car if the trip purpose is shoppingandor personal services

Analysing the systematic utility functions it is inter-esting to note how being older increases the not-installchoice thus confirming that younger people are more in-terested in technological innovation regarding the educa-tional level people with a masterrsquos degree show a greaterlikelihood to install the kit

Contrasting results may be observed for users living indifferent zones of residence Indeed people residing in thecity centre show a smaller propensity to install the kit due toreduced trip distance usually travelled and smaller interest incost savings By contrast people living in the outskirts maybenefit from a greater travel cost saving due to the greatertravel distances

As regards vehicle fleet characteristics the car age in-creases the choice to install the kit whereas owning a dieselcar negatively affects the propensity to install the kit Indeedin the former case people may decide due to the actual valueof the vehicle to consider the kit as an opportunity still

Table 5 Continued

Mean sdEnvironmentQenvmiddot 254 091Ienv1 211 102Ienv2 371 059Ienv3 321 091Ienv4 298 076Ienv5 197 073Ienv6 387 087


Table 6 Principal component analysis

Factor Indic Loading

Factor 1 fuel consumption

Qcons 0628Icons1 0357Icons2 0567Icons3 0488

Factor 2 vehicle design

Qdesign 0328Idesign1 0721Idesign3 0548Idesign4 0432

Factor 3 environment

Qenv 0567Ienv2 0355Ienv3 0618Ienv4 0712Ienv6 0667

Significant statements are with values greater than 035


depending on the trade-off between the vehicle market valueand the kit market price for upgrading the owned vehicleand also for revaluating the owned old vehicle (in terms ofemissions reduction and of fuel consumption) Neverthelessa user owning a diesel vehicle is less interested in the kit dueto the smaller benefits in terms of travel costs ese resultsconfirm that the monetary gain achievable installing the kitis the main boost

Furthermore the usual trip purpose also affects theinstallation choice Indeed travelling by car for shoppingand personal services positively increases the utility to in-stall the kit whilst travelling by car for work purposes has anopposite effect In this case travel distances are greater andthe usersrsquo distrust in the technology may play a significantrole

Finally the Δcost which has been defined as the differencebetween the weekly costs with the kit and weekly costwithout the kit [16] as expected was statically significantincreasing the probability of installation choice decreases

Among all the above-mentioned attributes the mostinnovative ones were those aimed at capturing the re-spondentsrsquo nonobservable determinants (attitudinal attri-butes) In particular the three following LVs werestatistically significant (see Table 7 and refer to the pathdiagram in Figure 2)

(i) LVConsumption representing the attitude towards fuelconsumption

(ii) LVDesign representing the attitude towards the ve-hicle design

(iii) LVEnvironment representing the attitude towards theenvironment

e reported results refer to the HCM statistically sig-nificant model out of three different models that werecalibrated using different sets of questions (already intro-duced in Section 42)

(1) With only direct questions (related to the choicecontext-Q)

(2) With only indirect questions (independent from thechoice context-I)

(3) Mixing direct and indirect questions (Q+ I)

Estimation results pointed out that no statistically sig-nificant measurement equations (thus no HCM) were ob-tained using only direct or only indirect questions whilst themix of both types of questions made it possible to obtain aconsistent and robust model (see also Table 8)

Indeed the results highlight that attitudes may be fruit-fully ldquograspedrdquo by a proper mix of direct and indirectquestions In this sense if on one hand the attitudes may beconsidered endogenous to the users and not dependent on thechoice context in which the users are called to make a de-cision on the other hand the attitudes cannot be consideredtotally independent from the alternatives under investigationthus they should be ldquograspedrdquo by direct questions

All the LVs contributed significantly to the systematicutilities but they showed different magnitudes Indeed thelatent variable representing the attitude towards the fuelconsumption had a relative weight greater than the othertwo LVs the latent variable representing the attitude to-wards the environment showed a weight greater than the LVrepresenting the attitude towards the design ese resultsconfirm the preliminary analyses of the psychometric in-dicators discussed in Section 42

Table 7 Estimation results of HCM

AttributeHCM

Install Not-installAge +0160 (+187) mdashMasterrsquos degree +0156 (+216) mdashZonRes +00761 (+198) mdashDiesel mdash +0479 (+352)CarAge +00272 (+155) mdashBy car-shopping +0669 (+1490) mdashBy car-personal services +0192 (+253) mdashΔcost mdash +00638 (+816)LVConsumption +0548 (+255) mdashLVDesign mdash +00682 (+246)LVEnvironment +0104 (+198) mdashδ1 +146 (+ 3890) mdashδ2 +134 (+ 4385) mdashδ3 (the ordinal treatment considers the estimation of threeextra parameters in the measurement model) +152 (+ 4611) mdash

Alternative specific constant (ASV) mdash mdashSTATISTICSobservations 1364Init-log-likelihood (only the log-likelihood associated with the discrete choice component isconsidered) minus944760

Final log-likelihood minus77981Rho-square 0212t-Test values are given in parenthesis δj refers to the thresholds estimation for the ordinal logit model


Analysing the LVs specification the indicators (state-ments) which were statistically significant for each LV havebeen grouped in Table 8

It is thus possible to understand which statement wassignificant in interpreting the observed choice behaviour Inparticular two indicators were significant in LVconsumptionwhilst three indicators were significant in LVdesign andLVenvironment (for each one of two latent variables one moreindicator was considered and normalised thus in all fourindicators are considered for each attitude)

Table 9 reports the relationships between the perceptionindicators and the attitudes and in particular shows the

estimation results for the following parameters the coeffi-cient of the latent variables (λ) the intercept (α) and errorterms (v)

Although the interpretation of the parameters is notimmediate the results indicate the robustness of the esti-mates and of the whole procedure moreover they highlightthat the signs are coherent with the preliminary statisticalanalyses carried out in Section 4 and the coefficient λ plays asignificant role with respect to the intercept (α) and errorterms (v)

Finally the estimation results for the HCM structuralmodel are displayed in Table 10

Consumption

+0725Observed indicator

Qcons

Observed indicatorIcons2

Observed indicatorIdesign1



Observed indicatorQenv

Observed indicatorIenv3



Observed indicatorQdesign

Error+0787

Error1

Error1

Error+143

Error+165

Error+118

Error+0983

Error1

Error+113

Error+138

+0148

+160

+184

+0495

+0729

+0396

1

1

1

Utility Design

Environment

Figure 2 Path diagram of the measurement equations


Overall the trip frequency for specific travel purposesand the following socioeconomic attributes were statisticallysignificant gender age and education level

Regarding the first latent variable the LVConsumption theactivity-related attributes and the level of attainment (masterdegree) are statistically significant and contribute to the LV

In terms of activity-related attributes travelling by carfor shopping purposes or for personal services exhibit dif-ferent signs in particular negative signs in the case ofshopping but positive in the case of personal services As thetrip purpose changes the users perceive the new technologydifferently with respect to the fuel consumption

Regarding the level of attainment having a masterrsquosdegree positively affects the LVConsumption indicating that ahigher cultural level affects such an attitude

e age is significant in both structural equations ofLVDesign and LVEnvironment indeed older people may bemoresensitive to vehicle design due to a higher willingness to payon the other hand the environment is perceived more in-tensely by older people furthermore the gender is signifi-cant and negative LVDesign and LVEnvironment explaining thatfemales are more sensitive to the design and the environ-ment problems

In general it can be observed that the intercepts and theerror terms are significant in all LVs

52 Goodness-of-Fit and Sensitivity Analyses Goodness-of-fitand sensitivity analyses were carried out by comparing theHCM model with a Binomial Logit Model (BNL) displayed

Table 8 Attitudes and indicators

Measurement modelConsumption Design EnvironmentQcons Qdesign QenvOne of the most important thingsin a car is the fuel consumption rate

e car design is one of the mostimportant factors in purchasing a car

I normally behave or act to reduce the environmentalimpact of my actions

Icons2 Idesign1 Ienv3I am usually attentive to the specialoffers of electric operators

When parking I am usually careful toavoid having my car damaged

I really enjoy spent my free time in parks green areas tobreathe clean area

Idesign3 Ienv4When furnishing I am willing to buy

pieces with modern design features andoriginal details

How much do you agree with the following sentence wemust act and make decisions to reduce emissions of

greenhouse gasesIdesign4 Ienv6

I am willing to go to the body shopmechanic not only for major damages

I am not willing to use the car during weekend to protectthe environment and then reduce air pollution

Table 9 Estimation results for the measurement models of HCM

Measurement modelLVconsumption LVDesign LVenvironment

Qcons Qdesign Qenv

α10 +0567 (+346) α20 minus179 (minus277) α30 minus189 (minus2448)λ10 +0725 (+1140) λ20 +0148 (+ 085) λ30 +0495 (+ 1412)v10 +0787 (+1637) ]20 +138 (+3316) v30 +118 (+3106)

Icons2 Idesign1 Ienv3α12 0 α21 0 α33 minus136 (minus1903)λ12 1 λ21 1 λ33 +0729 (+2082)v12 1 v21 1 ]33 +0983 (+2599)

Idesign3 Ienv4α23 +279 (+ 272) α 34 0λ23 +160 (+ 572) λ 34 1v23 +143 (+ 3124) v34 1

Idesign4 Ienv6α24 +279 (+ 269) α36 minus195 (minus2615)λ24 +184 (+ 652) λ36 +0396 (+1218)v24 +165 (+ 2817) v36 +113 (+3180)

e t-test values are given in parenthesis e indicators for each latent variable are highlighted in bold


in Table 11 e comparison was carried out to investigatethe following

(i) If the same socioeconomic and instrumental attributesare statistically significant with or without the explicitsimulation of psychoattitudinal factors through LVs

(ii) If the socioeconomic attributes continue having thesame interpretation and the same role

(iii) If the resulting HCMmodel has the same predictivecapability of a traditional approach andor showssimilar sensitivity to the instrumental attributes

Overall bothmodels presented similar specification of theutility functions except for the LVs ese results highlightthe robustness of hybrid choice modelling based on LVs andconfirm how LVs do not substitute significant attributes intraditional models but make it possible to better interpret thechoice phenomenon enriching the utility functions

Moreover it must be observed that in the HCM thealternative specific constant was not significant in the choicemodel supporting two further considerations In fact HCMembeds the alternative specific constants (intercepts) in thestructural and in the measurement equations thus allowinga different but clearer interpretation of the alternativespecific constantsrsquo role in the systematic utilities

If both models share similar attributes it is interesting toinvestigate if the HCM shows better goodness-of-fit andorhas a different sensitivity to the main instrumental attributerepresented by the Δcost

To this aim together with traditional indicators specificcomparison indicators proposed by de Luca and Cantarella[115] were estimated whereas direct elasticities were cal-culated with respect to the Δcost attribute

To compare the two models the following indicatorswere used

(i) e traditional rho-square indicator(ii) correct that is the percentage of users in the

calibration sample whose observed choices are giventhe maximum probability (whatever the value) bythe model

(iii) FFΣipsimi Nusers isin [01] Fitting Factor (FF) that is

the ratio between the sum over the users in thesample of the simulated choice probability for thechosen alternative psim

user isin [01] and the number ofusers in the sample Nusers

(iv) Clearly Correct which is the Percent-Correctpercentage of users in the sample whose observedchoices are given a probability greater thanthreshold t (considered thresholds are ge60708090) by the model this indicator makes it possibleto obtain a better interpretation than the traditionalcorrect

With respect to the rho-square indicator (see Table 6)the HCM clearly outperforms the BNL model Whilst if correct indicators show similar values FF indicators high-light that the HCM model is able to provide a better sim-ulation of the choices made by users (with reference to eachrespondent) this result is also confirmed by ClearlyCorrectt Indeed with respect to different thresholds tgreater than 05 HCM clearly outperforms the BNL (seeTable 12)

e models were also compared in terms of elasticitywith respect to the Δcost

In Figure 3 the results of the sensitivity analysis ofΔProbability of choice to install against Δcost increasing(from 10 to 50 with intermediate increasing values at20 30 and 40) are shown respectively for BNL andHCM First of all it should be observed that both models aresimilarly sensitive to cost indeed by increasing the Δcost(which means increasing the kit installation cost) theprobability to install the kit is negatively affected and theΔProbability to install the kit shifts from minus1 to minus14 forHCM and from minus1 to ndash13 for BNL

However if the two models show similar sensitivity tothe monetary cost it is interesting to investigate the sen-sitivity of the probability to install the kit with respect to theexplaining attitudes

is nonconventional elasticity analysis was carried outto understand if and how a change in the attitudes may affectthe choice probabilities

If it may be argued the meaning of varying an attitudeandor the actual possibility to modify an attitude on theother hand the aim of such an analysis is twofold first thecomprehension of the weight of the attitudes within themodel specification thus the need for explicitly simulatingthem secondly which effects may be obtained by acting onthe attitudes Indeed the attitudes may be affected by ex-ogenous factors such as different marketing strategies andthe fictitious creation of mediatic phenomena

With regard to our analyses the values of the LV (at-titudes) in the systematic utility functions were increasedfrom 10 until 100 In accordance with previous

Table 10 Estimation results for structural models of HCM

Structural modelAttributes ValueAttitude towards fuel consumption(LVConsumption)βMEAN1 minus232 (minus2980)ω1 +0787 (+1637)By car-shopping minus0448 (minus252)By car-personal services +0550 (+388)Masterrsquos degree +0128 (+222)Attitude towards vehicle design (LVDesign)βMEAN2 minus339 (minus4217)ω2 +0299 (+696)Age minus00955 (minus237)Gender minus0278 (minus665)Attitude towards environment (LVEnvironment)βMEAN3 minus

ω3 +134 (+2718)Age minus106 (minus1560)Gender minus112 (minus1674)


Table 11 Comparison between BNL model and HCM (only significant attributes are listed)

Attribute BNL HCMAge +0352 (+286) +0160 (+187)Masterrsquos degree +0360 (+286) +0156 (+216)ZonRes +0157 (+204) +00761 (+198)Diesel +0356 (+272) +0479 (+352)CarAge +00358 (+211) +00272 (+155)By car-shopping +0867 (+234) +0669 (+1490)By car-personal services +0678 (+180) +0192 (+253)Δcost +00641 (+855) +00638 (+816)LVConsumption mdash +0548 (+255)LVDesign mdash +00682 (+246)LVEnvironment mdash +0104 (+198)ASC +153 (+767) mdashSTATISTICSobservations 1364 1364Init-log-likelihood (only the log-likelihood associated with the discrete choice component isconsidered) minus944760 minus944760

Final log-likelihood minus780962 minus77981Rho-square 0184 0212t-Test values are given in parenthesis δj refers to the thresholds estimation for the ordinal logit model

Table 12 Comparison between BNL and HCM in terms of goodness-of-fit indicators (see [115])

GOF indicators BNL HCMcorrect 7031 7016FF 5964 6547ClearlyCorrect06 1452 2170ClearlyCorrect07 1144 2082ClearlyCorrect08 425 1708ClearlyCorrect09 022 557

ndash1

ndash1 ndash2

ndash4

ndash6

ndash6ndash8

ndash8

ndash4

ndash2

0 10 20 30 40 50 60

∆ Pr

obab

ility

to in

stal

l the

kit

()

∆ Pr

obab

ility

to in

stal

l the

kit

decr

ease

s

0

ndash2

ndash4

ndash6

ndash8

ndash10∆ cost

HSK installation cost increases

HCMBNL

Figure 3 BNLHCM sensitivity analysis of ΔProbability of choice to install against Δcost


considerations sensitivity analyses were referred to the at-titudes towards design and environment e results areproposed in Figure 4

First of all results emphasise that the choice behaviour issignificantly affected by the attitudes towards the design andthe environment

In fact as the LVdesign increases the ΔProbability toinstall the kit decreases from minus3 to minus26 this result in-dicates that car-makers or decision makers could affect thechoice to install the new technology by acting on the atti-tudes towards design but also working on the design of thetechnology itself In our specific case study the after-marketsignificantly changes the overall aesthetic of the owned carWith respect to the LVenvironment the ΔProbability variationto install the kit increases from 3 to 11 As commentedbefore acting on the usersrsquo proenvironment consciousnessand awareness may have significant impact on the choice toinstall the kit Also in this case it could be more effectiveacting on the attitudes than on the instrumental features ofthe automotive technology

6 Conclusions

e market penetration andor the diffusion of innovativetechnologies call for a realistic interpretation of the userrsquoschoice process and require choice models which are effectivebut can be easily implemented in any operational scenario[116]

A choice process can usually be outlined in differentstages (knowledge persuasion decision implementationand confirmation) and existing literature has widely in-vestigated these phenomena through aggregate approaches

founded on the diffusion modelstheory [117] or following(user oriented) disaggregate approaches which are able tointerpretmodel some of the abovementioned stages of thechoice process

Within this framework the paper aims to investigate thechoice behaviour which underpins the decisionimple-mentation stages when using the HySolarkit technologysolely as a case study

In particular the paper aims to respond to three mainresearch questions

(i) If and which attitudinal factors significantly affectusersrsquo choice of new automotive technology (egafter-market hybridization kit) thus if usersrsquo choiceshould be investigated through more advanced andbehaviourally complex models

(ii) Which is the most effective surveying approach toobserve usersrsquo attitudes Indeed one of the mainissues of choice modelling consists in how toldquograsprdquo usersrsquo attitudes and in particular throughthe use of which type of questions (eg direct orindirect) as well as through which specificquestions

(iii) How the propensity of choosing a new automotivetechnology is sensitive to usersrsquo attitudesconcernsand changes

As secondary results the paper also made it possible tocarry out a comparison between a Hybrid choice model(HCM) with Latent Variables (LVs) and a traditional bi-nomial Logit model (BNL) and identified which kind ofattitude plays a role in the propensity to install a new andldquogreenerrdquo automotive technology

3 4 4 5 6 7 8 9 10 11

10ndash3

ndash6

ndash9ndash12

ndash14ndash17

ndash19ndash21

ndash23ndash25

20 30 40 50 60 70 80 90 100∆

Prob

abili

ty to

inst

all t

he k

it (

)

∆ Attitude towards designenvironment

Design attitude increases

Environment attitude increases

∆ Pr

obab

ility

to in

stal

l the

kit

incr

ease

s

141210

86420

ndash2ndash4ndash6ndash8

ndash10ndash12ndash14ndash16ndash18ndash20ndash22ndash24ndash26ndash28

EnvironmentDesign

Figure 4 HCM sensitivity analysis of ΔProbability of choice to install against attitude towards designenvironment


e above research questions were addressed by carryingout a stated preferences survey which investigated the choiceof installing an after-market hybridization kit and collectedattitudinal factors through direct and indirect questions

Estimation results highlighted that attitudinal factorssignificantly affect usersrsquo choice Indeed three (of the fiveinvestigated attitudes) were statistically significant andhighlighted that the HCM with LVs model was able to ef-fectively interpret the usersrsquo choice e statistically signif-icant attitudes were

(i) Attitude towards the ldquoenvironmentrdquo(ii) Attitude towards the ldquofuel consumptionrdquo(iii) Attitude towards the ldquovehicle designrdquo

If the first two attitudes are in line with the study ex-pectations the attitude towards the design reveals the role ofaesthetic factors By contrast the attitudes regarding thetechnology and the reliability of technology did not turn outto be statistically significant highlighting that the usersrsquobehaviour is not influenced by being a ldquotechnologicallyorientedcaptiverdquo user

In terms of magnitude the latent variable representingthe attitude towards ldquofuel consumptionrdquo showed a relativeweight greater than the other two LVs whereas the latentvariable representing the attitude towards the ldquoenviron-mentrdquo showed a weight greater than the LV representing theattitude towards the ldquodesignrdquo

e analysis of the Logit model formulation was carriedout to compare the systematic utility specifications andsecondarily to compare the models in terms of the good-ness-of-fit

e primary consideration is that the two models sharealmost the same specification of the utility functions exceptfor the LVs ese results highlight how attributes that aresignificant in traditional models continue to be significant inthe HCM and how the LVs do not substitute these attributesbut enrich the utility functions allowing for a better inter-pretation of the choice phenomenon

Furthermore the socioeconomic attributes which arestatistically significant in the BNL become statistically sig-nificant in the specification of the LVs only Such a resultconfirms that the HCM specification also makes it possible tobetter classify the socioeconomic attributes highlighting theirrole in the interpretation of the usersrsquo attitudes within the LVs

Regarding the goodness-of-fit it has been shown that theHCM outperforms the BNL also allowing for a bettermarket share prediction

With regards to the approach that aims to ldquograsprdquo theattitudes no statistically significant results were obtainedusing only direct or only indirect questions whilst the mixof both types of questions made it possible to obtain aconsistent and robust model erefore the first attitudesmay be considered endogenous to the users thus theyshould be ldquograspedrdquo by indirect questions regarding therespondentrsquos generic preferences secondly the attitudescannot be considered as being totally independent fromthe choice context under investigation thus they shouldbe ldquograspedrdquo by direct questions regarding the

respondentrsquos preferences specifically related to the choicecontext

Finally although both models showed a similar sensi-tivity to the instrumental attributes (installation cost) thesensitivity analysis on the HCM model showed how thechoice probabilities are extremely sensitive to the attitudevariation In particular the choice behaviour is more sen-sitive to the attitudes towards the ldquodesignrdquo and theldquoenvironmentrdquo

Such results indicate that decision makers andorindustries may pursue sustainability goals acting not onlyon the instrumental features of a technology but alsotrying to induce a behavioural change through themodification of the userrsquos attitudes concerns andorperceptions ey could work on specific strategies suchas the promotion of educational programmes the de-velopment of ldquoad hocrdquo marketing strategies andor thecreation of social phenomena through the current socialplatforms

In conclusion the adoption of new technologies requireseffective modelling solutions which are able to simulate thechoice process andor allow for a wide interpretation of thephenomenon From a political point of view the marketpenetration of sustainable technologies depends on theinstrumental features of the technology but it can be sig-nificantly affected by acting on andor cultivating ldquousersrsquobackground feelings and emotionsrdquo

Indeed this field is complex and worthy of further re-search and it is the authorsrsquo opinion that future perspectivesshould focus on three main streams

Firstly a significant amount of effort should be placedon the developmentimplementation of alternativetheoretical paradigms which are able to embed thepsychological factors within behavioural paradigmsunlike the utilitarian framework Moreover such para-digms should be integrated in a unique behaviouralframework coherent with the five stages of the diffusionparadigms knowledge persuasion decision imple-mentation and confirmation e integration ofbehavioural choice models within the ldquodiffusion para-digmrdquo might give interesting insights for a better un-derstanding of the diffusion process and would be ofsupport in interpreting and modelling the ldquoknowledgerdquoand ldquopersuasionrdquo stages which are rather overlooked inthe literature

Secondly it could also be relevant to investigate if andhow the diffusion of a new technology could be affected byits environmental impact Indeed even though a tech-nology might be attractive this could determine negativefeelings from the potential users is aspect which doesnot hold for the HySolarKit should be explicitly observedand modelled

Finally another crucial research field concerns the in-vestigation of different and reliable but easy to deployapproaches for capturing and measuring the userrsquos psy-chological factors which in turn affect usersrsquo behaviourIndeed the use of the Likert scale andor the use of absolutejudgments may not effectively represent the userrsquosperceptions


Appendix

Technological Framework

In this paper the HySolarKit (HSK see Figure 5) developedby the E-Prob Lab of the University of Salerno (Italy) hasbeen considered e technology is an aftermarket mildhybridization kit based on the idea that conventional ve-hicles may be upgraded to hybrid vehicles by means of theelectrification of the rear wheels in front-wheel-drive ve-hicles adopting in-wheel motors plug-in HEVs[17 18 20 21] Concerning the battery it may be rechargedin three alternative ways by the rear wheels when operatingin generation mode by photovoltaic panels or by a regularelectric power outlet when the vehicle is connected to thepower grid in plug-in mode [19] In terms of reliability it isestimated that the battery duration in fully charged con-ditions is around 15 Km in hybrid mode and in an urbancontext

A more detailed description of the vehicle managementunit and on-board diagnostics protocol gate is provided in[22]

Data Availability

e data are available under request to the University ofSalerno

Conflicts of Interest

e authors declare that there are no conflicts of interestregarding the publication of this paper

Acknowledgments

is research was partially supported by the University ofSalerno Italy EU under local grant no ORSA171328 fi-nancial year 2017 e authors wish also to thank profGianfranco Rizzo and the LIFE-SAVE project (Solar AidedVehicle Electrification LIFE16 ENVIT000442) that in-spired the line of research

References

[1] F J Bahamonde-Birke U Kunert H Link andJ d D Ortuzar ldquoAbout attitudes and perceptions finding

Figure 5 HySolar kit (HSK)-vehicle integration


the proper way to consider latent variables in discrete choicemodelsrdquo Transportation vol 44 no 3 pp 475ndash493 2017

[2] R A Daziano and D Bolduc ldquoIncorporating pro-envi-ronmental preferences towards green automobile technol-ogies through a Bayesian hybrid choice modelrdquoTransportmetrica A Transport Science vol 9 no 1 pp 74ndash106 2013

[3] M Vredin Johansson T Heldt and P Johansson ldquoeeffects of attitudes and personality traits on mode choicerdquoTransportation Research Part A Policy and Practice vol 40no 6 pp 507ndash525 2006

[4] C Domarchi A Tudela and A Gonzalez ldquoEffect of atti-tudes habit and affective appraisal on mode choice anapplication to university workersrdquo Transportation vol 35no 5 pp 585ndash599 2008

[5] K M N Habib L Kattan and M T Islam ldquoWhy do thepeople use transit A model for explanation of personalattitude towards transit service qualityrdquo in Proceedings of the88th Annual Meeting of the Transportation Research BoardWashington DC USA January 2009

[6] B Atasoy A Glerum R Hurtubia and M Bierlaire ldquoDe-mand for public transport services integrating qualitativeand quantitative methodsrdquo in Proceedings of the 10th SwissTransport Research Conference Ascona Switzerland Sep-tember 2010

[7] M F Yantildeez S Raveau and J D D Ortuzar ldquoInclusion oflatent variables in mixed logit models modelling andforecastingrdquo Transportation Research Part A Policy andPractice vol 44 no 9 pp 744ndash753 2010

[8] D Bolduc and R Alvarez-Daziano ldquoOn estimation of hybridchoice modelsrdquo in Choice Modelling Ae State-Of-Ae-Artand the State-Of-Practice Proceedings from the InauguralInternational Choice Modelling Conference pp 259ndash287Emerald Group Publishing Limited Bingley UK 2010

[9] G Correia and J M Viegas ldquoCarpooling and carpool clubsclarifying concepts and assessing value enhancement pos-sibilities through a Stated Preference web survey in LisbonPortugalrdquo Transportation Research Part A Policy andPractice vol 45 no 2 pp 81ndash90 2011

[10] G H de Almeida Correia J de Abreu e Silva andJ M Viegas ldquoUsing latent attitudinal variables estimatedthrough a structural equations model for understandingcarpooling propensityrdquo Transportation Planning and Tech-nology vol 36 no 6 pp 499ndash519 2013

[11] K Choocharukul H T Van and S Fujii ldquoPsychologicaleffects of travel behavior on preference of residential locationchoicerdquo Transportation Research Part A Policy and Practicevol 42 no 1 pp 116ndash124 2008

[12] Y O Susilo and R Kitamura ldquoStructural changes in com-mutersrsquo daily travel the case of auto and transit commutersin the Osaka metropolitan area of Japan 1980-2000rdquoTransportation Research Part A Policy and Practice vol 42no 1 pp 95ndash115 2008

[13] E Parkany R Gallagher and P Viveiros ldquoAre attitudesimportant in travel choicerdquo Transportation Research RecordJournal of the Transportation Research Board vol 1894 no 1pp 127ndash139 2004

[14] C G Prato S Bekhor and C Pronello ldquoLatent variables androute choice behaviorrdquo Transportation vol 39 no 2pp 299ndash319 2012

[15] S Bekhor and G Albert ldquoAccounting for sensation seekingin route choice behavior with travel time informationrdquoTransportation Research Part F Traffic Psychology and Be-haviour vol 22 pp 39ndash49 2014

[16] S de Luca R Di Pace and V Marano ldquoModelling theadoption intention and installation choice of an automotiveafter-market mild-solar-hybridization kitrdquo TransportationResearch Part C Emerging Technologies vol 56 pp 426ndash4452015

[17] I Arsie G Rizzo and M Sorrentino ldquoOptimal design anddynamic simulation of a hybrid solar vehicle (no 2006-01-2997)rdquo SAE TransactionsmdashJournal of Engines vol 115 no 3pp 805ndash811 2007

[18] G Rizzo I Arsie andM Sorrentino ldquoHybrid solar vehiclesrdquoSolar Collectors and Panels Aeory and Applications InTechWest Palm Beach FL USA 2010

[19] G Rizzo I Arsie and M Sorrentino ldquoSolar energy for carsperspectives opportunities and problemsrdquo in Proceedings ofthe GTAA Meeting Universite de Mulhouse MulhouseFrance May 2010

[20] I Arsie M DrsquoAgostino M Naddeo G Rizzo andM Sorrentino ldquoToward the development of a through-the-road solar hybridized vehiclerdquo IFAC Proceedings Volumesvol 46 no 21 pp 806ndash811 2013

[21] G Rizzo V Marano C Pisanti et al ldquoA prototype mild-solar-hybridization kit design and challengesrdquo EnergyProcedia vol 45 pp 1017ndash1026 2014

[22] S de Luca and R Di Pace ldquoAftermarket vehicle hybrid-ization potential market penetration and environmentalbenefits of a hybrid-solar kitrdquo International Journal ofSustainable Transportation vol 12 no 5 pp 353ndash366 2018

[23] J Kim S Rasouli and H Timmermans ldquoExpanding scope ofhybrid choice models allowing for mixture of social influ-ences and latent attitudes application to intended purchaseof electric carsrdquo Transportation Research Part A Policy andPractice vol 69 pp 71ndash85 2014

[24] J Kim S Rasouli and H J P Timmermans ldquoe effects ofactivity-travel context and individual attitudes on car-sharing decisions under travel time uncertainty a hybridchoice modeling approachrdquo Transportation Research Part DTransport and Environment vol 56 pp 189ndash202 2017

[25] A Cartenı E Cascetta and S de Luca ldquoA random utilitymodel for park amp carsharing services and the pure preferencefor electric vehiclesrdquo Transport Policy vol 48 pp 49ndash592016

[26] I Ajzen ldquoe theory of planned behaviorrdquo OrganizationalBehavior and Human Decision Processes vol 50 no 2pp 179ndash211 1991

[27] B Lane and S Potter ldquoe adoption of cleaner vehicles in theUK exploring the consumer attitudendashaction gaprdquo Journal ofCleaner Production vol 15 no 11-12 pp 1085ndash1092 2007

[28] I Moons and P De Pelsmacker ldquoEmotions as determinantsof electric car usage intentionrdquo Journal of MarketingManagement vol 28 no 3-4 pp 195ndash237 2012

[29] P C Stern ldquoNew environmental theories toward a coherenttheory of environmentally significant behaviorrdquo Journal ofSocial Issues vol 56 no 3 pp 407ndash424 2000

[30] G Schuitema J Anable S Skippon and N Kinnear ldquoerole of instrumental hedonic and symbolic attributes in theintention to adopt electric vehiclesrdquo Transportation ResearchPart A Policy and Practice vol 48 pp 39ndash49 2013

[31] E H Noppers K Keizer J W Bolderdijk and L Steg ldquoeadoption of sustainable innovations driven by symbolic andenvironmental motivesrdquo Global Environmental Changevol 25 pp 52ndash62 2014

[32] J Axsen and K S Kurani ldquoHybrid plug-in hybrid orelectric-What do car buyers wantrdquo Energy Policy vol 61pp 532ndash543 2013


[33] A Peters and E Dutschke ldquoHow do consumers perceiveelectric vehicles A comparison of German consumergroupsrdquo Journal of Environmental Policy amp Planning vol 16no 3 pp 359ndash377 2014

[34] M Petschnig S Heidenreich and P Spieth ldquoInnovativealternatives take actionmdashinvestigating determinants of al-ternative fuel vehicle adoptionrdquo Transportation ResearchPart A Policy and Practice vol 61 pp 68ndash83 2014

[35] S Skippon and M Garwood ldquoResponses to battery electricvehicles UK consumer attitudes and attributions of symbolicmeaning following direct experience to reduce psychologicaldistancerdquo Transportation Research Part D Transport andEnvironment vol 16 no 7 pp 525ndash531 2011

[36] E Graham-Rowe B Gardner C Abraham et al ldquoMain-stream consumers driving plug-in battery-electric and plug-in hybrid electric cars a qualitative analysis of responses andevaluationsrdquo Transportation Research Part A Policy andPractice vol 46 no 1 pp 140ndash153 2012

[37] B Gardner and C Abraham ldquoWhat drives car use Agrounded theory analysis of commutersrsquo reasons for driv-ingrdquo Transportation Research Part F Traffic Psychology andBehaviour vol 10 no 3 pp 187ndash200 2007

[38] E Mann and C Abraham ldquoe role of affect in UK com-mutersrsquo travel mode choices an interpretative phenome-nological analysisrdquo British Journal of Psychology vol 97no 2 pp 155ndash176 2006

[39] L Steg ldquoCar use lust and must Instrumental symbolic andaffective motives for car userdquo Transportation Research PartA Policy and Practice vol 39 no 2-3 pp 147ndash162 2005

[40] S Beggs S Cardell and J Hausman ldquoAssessing the potentialdemand for electric carsrdquo Journal of Econometrics vol 17no 1 pp 1ndash19 1981

[41] K Train ldquoe potential market for non-gasoline-poweredautomobilesrdquo Transportation Research Part A Generalvol 14 no 5-6 pp 405ndash414 1980

[42] D A Hensher ldquoFunctional measurement individual pref-erence and discrete-choice modelling theory and applica-tionrdquo Journal of Economic Psychology vol 2 no 4pp 323ndash335 1982

[43] K Train Qualitative Choice Analysis Econometrics and anApplication to 604 Automobile Demand MIT Press Cam-bridge MA USA 1986

[44] T E Golob and D A Hensher ldquoDriving behaviour of longdistance truck drivers the effects of schedule compliance ondrug use and speeding citationsrdquo International Journal ofTransport EconomicsRivista internazionale di economia deitrasporti vol 23 pp 267ndash301 1996

[45] D Brownstone D S Bunch T F Golob and W Ren ldquoAtransactions choice model for forecasting demand for al-ternative-fuel vehiclesrdquo Research in Transportation Eco-nomics vol 4 pp 87ndash129 1996

[46] D Brownstone D S Bunch and K Train ldquoJoint mixed logitmodels of stated and revealed preferences for alternative-fuelvehiclesrdquo Transportation Research Part B Methodologicalvol 34 no 5 pp 315ndash338 2000

[47] J K Dagsvik T Wennemo D G Wetterwald andR Aaberge ldquoPotential demand for alternative fuel vehiclesrdquoTransportation Research Part B Methodological vol 36no 4 pp 361ndash384 2002

[48] D Bolduc N Boucher and R Alvarez-Daziano ldquoHybridchoice modeling of new technologies for car choice inCanadardquo Transportation Research Record Journal of theTransportation Research Board vol 2082 no 1 pp 63ndash712008

[49] A Hackbarth and R Madlener ldquoConsumer preferences foralternative fuel vehicles a discrete choice analysisrdquo Trans-portation Research Part D Transport and Environmentvol 25 pp 5ndash17 2013

[50] A Glerum L Stankovikj M emans and M BierlaireldquoForecasting the demand for electric vehicles accounting forattitudes and perceptionsrdquo Transportation Science vol 48no 4 pp 483ndash499 2014

[51] D J Santini and A D Vyas Suggestions for a New VehicleChoice Model Simulating Advanced Vehicles IntroductionDecisions (AVID) Structure and Coefficients Argonne Na-tional Laboratory Argonne IL USA 2005

[52] D Potoglou and P S Kanaroglou ldquoHousehold demand andwillingness to pay for clean vehiclesrdquo Transportation Re-search Part D Transport and Environment vol 12 no 4pp 264ndash274 2007

[53] K Sikes T Gross Z Lin J Sullivan T Cleary and J WardldquoPlug-in hybrid electric vehicle market introduction studyrdquoFinal report ORNLTM-2009019 US Department of En-ergy Washington DC USA 2010

[54] J Struben and J D Sterman ldquoTransition challenges foralternative fuel vehicle and transportation systemsrdquo Envi-ronment and Planning B Planning and Design vol 35 no 6pp 1070ndash1097 2008

[55] L Qian and D Soopramanien ldquoHeterogeneous consumerpreferences for alternative fuel cars in Chinardquo TransportationResearch Part D Transport and Environment vol 16 no 8pp 607ndash613 2011

[56] J Ahn G Jeong and Y Kim ldquoA forecast of householdownership and use of alternative fuel vehicles a multiplediscrete-continuous choice approachrdquo Energy Economicsvol 30 no 5 pp 2091ndash2104 2008

[57] C G Chorus J M Rose and D A Hensher ldquoRegretminimization or utility maximization it depends on theattributerdquo Environment and Planning B Planning and De-sign vol 40 no 1 pp 154ndash169 2013

[58] H DittmarAe Social Psychology of Material Possessions ToHave Is to Be Harvester Wheatsheaf Hemel HempsteadUK 1992

[59] K E Voss E R Spangenberg and B Grohmann ldquoMea-suring the hedonic and utilitarian dimensions of consumerattituderdquo Journal of Marketing Research vol 40 no 3pp 310ndash320 2003

[60] G Roehrich ldquoConsumer innovativenessrdquo Journal of BusinessResearch vol 57 no 6 pp 671ndash677 2004

[61] S Shepherd P Bonsall and G Harrison ldquoFactors affectingfuture demand for electric vehicles a model based studyrdquoTransport Policy vol 20 pp 62ndash74 2012

[62] E Cascetta and A Cartenı ldquoe hedonic value of railwaysterminals A quantitative analysis of the impact of stationsquality on travellers behaviourrdquo Transportation ResearchPart A Policy and Practice vol 61 pp 41ndash52 2014

[63] D S Bunch M Bradley T F Golob R Kitamura andG P Occhiuzzo ldquoDemand for clean-fuel vehicles in Cal-ifornia a discrete-choice stated preference pilot projectrdquoTransportation Research Part A Policy and Practice vol 27no 3 pp 237ndash253 1993

[64] E Cheron and M Zins ldquoElectric vehicle purchasing in-tentions the concern over battery charge durationrdquoTransportation Research Part A Policy and Practice vol 31no 3 pp 235ndash243 1997

[65] P Ong and K Hasselhoff ldquoHigh interest in hybrid carsrdquo SCSFact Sheet vol 1 no 5 2005


[66] S Musti and K M Kockelman ldquoEvolution of the householdvehicle fleet anticipating fleet composition PHEV adoptionand GHG emissions in Austin Texasrdquo Transportation Re-search Part A Policy and Practice vol 45 no 8 pp 707ndash7202011

[67] L He W Chen and G Conzelmann ldquoImpact of vehicleusage on consumer choice of hybrid electric vehiclesrdquoTransportation Research Part D Transport and Environmentvol 17 no 3 pp 208ndash214 2012

[68] S de Luca and R Di Pace ldquoModelling the propensity inadhering to a carsharing system a behavioral approachrdquoTransportation Research Procedia vol 3 pp 866ndash875 2014

[69] P Plotz U Schneider J Globisch and E Dutschke ldquoWhowill buy electric vehicles Identifying early adopters inGermanyrdquo Transportation Research Part A Policy andPractice vol 67 pp 96ndash109 2014

[70] J Axsen and K S Kurani ldquoAnticipating plug-in hybridvehicle energy impacts in California constructing consumer-informed recharge profilesrdquo Transportation Research Part DTransport and Environment vol 15 no 4 pp 212ndash219 2010

[71] S Bamberg and G Moser ldquoTwenty years after HinesHungerford and Tomera a new meta-analysis of psycho-social determinants of pro-environmental behaviourrdquoJournal of Environmental Psychology vol 27 no 1 pp 14ndash25 2007

[72] M C Onwezen G Antonides and J Bartels ldquoe normactivation model an exploration of the functions of antic-ipated pride and guilt in pro-environmental behaviourrdquoJournal of Economic Psychology vol 39 pp 141ndash153 2013

[73] L Steg and C Vlek ldquoEncouraging pro-environmental be-haviour an integrative review and research agendardquo Journalof Environmental Psychology vol 29 no 3 pp 309ndash3172009

[74] E Shih andH J Schau ldquoTo justify or not to justify the role ofanticipated regret on consumersrsquo decisions to upgradetechnological innovationsrdquo Journal of Retailing vol 87no 2 pp 242ndash251 2011

[75] L Watson and M T Spence ldquoCauses and consequences ofemotions on consumer behaviourrdquo European Journal ofMarketing vol 41 no 56 pp 487ndash511 2007

[76] S Choo and P L Mokhtarian ldquoWhat type of vehicle dopeople drive e role of attitude and lifestyle in influencingvehicle type choicerdquo Transportation Research Part A Policyand Practice vol 38 no 3 pp 201ndash222 2004

[77] K Kishi and K Satoh ldquoEvaluation of willingness to buy alow-pollution car in Japanrdquo Journal of the Eastern AsiaSociety for Transportation Studies vol 6 pp 3121ndash3134 2005

[78] R Alvarez-Daziano and D Bolduc ldquoCanadian consumersrsquoperceptual and attitudinal responses toward green auto-mobile technologies an application of Hybrid ChoiceModelsrdquo European Summer School in Resources Environ-mental Economics Economics Transport and EnvironmentVenice International University Venice Italy 2009

[79] A F Jensen Development of a Stated Preference Experimentfor Electric Vehicle Demand Masterrsquos esis TechnicalUniversity of Denmark Lyngby Denmark 2010

[80] A F Jensen E Cherchi and S L Mabit ldquoOn the stability ofpreferences and attitudes before and after experiencing anelectric vehiclerdquo Transportation Research Part D Transportand Environment vol 25 pp 24ndash32 2013

[81] J J Soto V Cantillo and J Arellana ldquoHybrid choicemodelling of alternative fueled vehicles in colombian citiesincluding second order structural equationsrdquo in Proceedings

of the Transportation Research Board 93rd Annual Meeting(No 14-4545) Washington DC USA January 2014

[82] I Tsouros and A Polydoropoulou ldquoWho will buy alternativefueled or automated vehicles a modular behavioral mod-eling approachrdquo Transportation Research Part A Policy andPractice vol 132 pp 214ndash225 2020

[83] A Daly S Hess B Patruni D Potoglou and C RohrldquoUsing ordered attitudinal indicators in a latent variablechoice model a study of the impact of security on rail travelbehaviourrdquo Transportation vol 39 no 2 pp 267ndash297 2012

[84] T Ioannis P Amalia and T Athena ldquoA hybrid choicemodel for alternative fuel car purchaserdquo in Proceedings of theInternational Choice Modelling Conference Cape TownSouth Africa 2015

[85] J D D Ortuzar and G A Hutt ldquoLa influencia de elementossubjetivos en funciones desagregadas de eleccion discretardquoIngenierıa de Sistemas vol 4 no 2 pp 37ndash54 1984

[86] D McFadden ldquoe choice theory approach to market re-searchrdquo Marketing Science vol 5 no 4 pp 275ndash297 1986

[87] K G Joreskog ldquoSimultaneous factor Analysis in severalpopulationsrdquo ETS Research Bulletin Series vol 1970 no 2pp indash31 1970

[88] M Ben-Akiva D McFadden T Garling et al ldquoFrameworkfor modeling choice behaviorrdquo Marketing Letters vol 10no 3 pp 187ndash203

[89] J L Larichev ldquoExtended discrete choice models integratedframework flexible error structures and latent variablesrdquopp 80ndash116 Massachusetts Institute of Technology Cam-bridge MA USA 2001 Doctoral dissertation

[90] D McFadden ldquoEconomic choicesrdquo American EconomicReview vol 91 no 3 pp 351ndash378 2001

[91] M Ben-Akiva J Walker A T Bernardino D A GopinathT Morikawa and A Polydoropoulou ldquoIntegration of choiceand latent variable modelsrdquo Perpetual Motion Travel Be-haviour Research Opportunities and Application Challengespp 431ndash470 Emerald Group Publishing Bingley UK 2002

[92] J Walker and M Ben-Akiva ldquoGeneralized random utilitymodelrdquo Mathematical Social Sciences vol 43 no 3pp 303ndash343 2002

[93] S Raveau R Alvarez-Daziano M Yantildeez D Bolduc andJ de Dios Ortuzar ldquoSequential and simultaneous estimationof hybrid discrete choice models some new findingsrdquoTransportation Research Record Journal of the Trans-portation Research Board vol 2156 no 1 pp 131ndash139 2010

[94] M Kroesen and C Chorus ldquoe role of general and specificattitudes in predicting travel behaviormdasha fatal dilemmardquoTravel Behaviour and Society vol 10 pp 33ndash41 2018

[95] R Krueger A Vij and T H Rashidi ldquoNormative beliefs andmodality styles a latent class and latent variable model oftravel behaviourrdquo Transportation vol 45 no 3 pp 789ndash8252018

[96] Morhauge E Cherchi J L Walker and J Rich ldquoe roleof intention as mediator between latent effects and behaviorapplication of a hybrid choice model to study departure timechoicesrdquo Transportation vol 46 no 4 pp 1421ndash1445 2019

[97] W Li and M Kamargianni ldquoAn integrated choice and latentvariable model to explore the influence of attitudinal andperceptual factors on shared mobility choices and their valueof time estimationrdquo Transportation Science vol 54 no 12019

[98] F S Koppelman and J R Hauser ldquoDestination choice be-haviour for non-grocery-shopping tripsrdquo TransportationResearch Record vol 673 pp 157ndash165 1978


[99] P E Green ldquoHybrid models for conjoint analysis an ex-pository reviewrdquo Journal of Marketing Research vol 21no 2 pp 155ndash169 1984

[100] K M Harris and M P Keane ldquoA model of health planchoice inferring preferences and perceptions from a com-bination of revealed preference and attitudinal datardquo Journalof Econometrics vol 89 no 1-2 pp 131ndash157 1998

[101] J N Prashker ldquoScaling perceptions of reliability of urbantravel modes using indscal and factor analysis methodsrdquoTransportation Research Part A General vol 13 no 3pp 203ndash212 1979

[102] K A Bollen Structural Equations with Latent VariablesWiley New York NY USA 1989

[103] R Likert ldquoA technique for the measurement of attitudesrdquoArchives of psychology vol 22 no 140 1932

[104] K Ashok W R Dillon and S Yuan ldquoExtending discretechoice models to incorporate attitudinal and other latentvariablesrdquo Journal of Marketing Research vol 39 no 1pp 31ndash46 2002

[105] M L Tam W H K Lam and H P Lo ldquoModeling airpassenger travel behavior on airport ground access modechoicesrdquo Transportmetrica vol 4 no 2 pp 135ndash153 2008

[106] F J Bahamonde-Birke and J D D Ortuzar ldquoOn the vari-ability of hybrid discrete choice modelsrdquo TransportmetricaA Transport Science vol 10 no 1 pp 74ndash88 2014

[107] X Cao P L Mokhtarian and S L Handy ldquoe relationshipbetween the built environment and nonwork travel a casestudy of Northern Californiardquo Transportation Research PartA Policy and Practice vol 43 no 5 pp 548ndash559 2009

[108] S Fujii and T Garling ldquoApplication of attitude theory forimproved predictive accuracy of stated preference methodsin travel demand analysisrdquo Transportation Research Part APolicy and Practice vol 37 no 4 pp 389ndash402 2003

[109] A Vij and J L Walker ldquoHow when and why integratedchoice and latent variable models are latently usefulrdquoTransportation Research Part B Methodological vol 90pp 192ndash217 2016

[110] M Bierlaire andM Fetiarison ldquoEstimation of discrete choicemodels extending BIOGEMErdquo in Proceedings of the SwissTransport Research Conference (STRC) Leiden NetherlandsSeptember 2009

[111] D Bolduc and A Giroux Ae Integrated Choice and LatentVariable (ICLV) Model Handout to Accompany the Esti-mation Software Dacuteepartement drsquoacuteeconomique UniversitacuteeLaval Quebec Canada 2005

[112] G W Allport Handbook of Social Psychology Addison-Wesley Boston MA USA 1935

[113] J Pickens ldquoAttitudes and perceptionsrdquo Organizational Be-havior in Health Care pp 43ndash75 Jones and Bartlett Pub-lishers Sudbury MA USA 2005

[114] P H Lindsay and D A Norman Human InformationProcessing An Introduction to Psychology Academic PressCambridge MA USA 2013

[115] S de Luca and G E Cantarella ldquoValidation and comparisonof choice modelsrdquo in Success and Failure of Travel DemandManagement Measures W Saleh Ed Vol 37ndash58 AshgatePublications Farnham UK 2011

[116] S de Luca and A Papola ldquoEvaluation of travel demandmanagement policies in the urban area of Naplesrdquo Advancesin Transport vol 8 pp 185ndash194 2001

[117] F M Bass K Gordon T L Ferguson and M L GithensldquoForecasting diffusion of a new technology prior to productlaunchrdquo Interfaces vol 31 no 3 pp S82ndashS93 2001

[118] K A Bollen Structural Equations with Latent VariablesJohn Wiley amp Sons Hoboken NJ USA 2014

[119] G Correia J Abreu e Silva and J Viegas ldquoUsing latentattitudinal variables for measuring carpooling propensityrdquoin Proceeding of 12th World Conference on TransportResearch Lisbon Portugal July 2010

[120] T F Golob ldquoStructural equation modeling for travel be-havior researchrdquo Transportation Research Part B Method-ological vol 37 no 1 pp 1ndash25 2003

[121] T F Golob D S Bunch and D Brownstone ldquoA vehicle useforecasting model based on revealed and stated vehicle typechoice and utilisation datardquo Journal of Transport Economicsand Policy vol 31 pp 69ndash92 1997

[122] S Hess N Sanko J Dumont and A Daly ldquoA latent variableapproach to dealing with missing or inaccurately measuredvariables the case of incomerdquo in Proceedings of the AirdInternational Choice Modelling Conference pp 3ndash5 SydneyAustralia July 2013

[123] T Ohsaki T Kishi T Kuboki et al ldquoOvercharge reaction oflithium-ion batteriesrdquo Journal of Power Sources vol 146no 1-2 pp 97ndash100 2005


Over the last two decades usersrsquo propensity to choosealternative fuel vehicles (AFVs) has been widely investigatedand many recent analyses have pointed out the necessity toalso take into account nonobservable variables such as theperceptions and the attitudes of the users (see [1]) along withdirectly observablemeasurable attributes Indeed eventhough models usually adopted in demand modelling aresuitable for the representation of the choice process thesecannot be applied to the representation of perceptions andattitudes [2]

Based on previous considerations several researchershave proposed the integration of latent variables withintraditional econometric frameworks such as the utilitariantheoretical paradigm erefore Latent variables HybridChoice Models (HCMs) have been adopted in several casesand in particular they have been applied to capture theindividual attributes such as attitudes and habit influencingthe mode choice [3ndash10] the residential location choice[11 12] the route choice decision making [13ndash15] and thevehicles choice (see Section 2)

e present paper starting from some preliminary re-sults proposed in de Luca and Di Pace [16] aims to in-vestigate the role of attitudinal factors in the choice of a newautomotive technology the HySolarKit which aims toelectrifyhybridize existing vehicles through an after-marketkit which can be recharged by the grid as well as by solarpower [17ndash21]

e paper first proposes a literature review on the mostsignificant contributions on usersrsquo intention to purchaseelectrichybrid automotive technologies with particularattention to the psychological factors and the possiblemodelling approaches

It then addresses the following three main researchquestions (e main aim is not to investigate the potenti-alities of the HySolarkit but to comprehend the role ofattributes different from typical instrumental attributes)

(1) Whether and how usersrsquo characteristics and attitudesmay affect usersrsquo behaviour with respect to newtechnological (automotive) scenarios (eg after-market hybridization kit) thus if usersrsquo choiceshould be investigated through more advanced andbehaviourally complex models

(2) How to better ldquograsprdquo usersrsquo attitudesconcernsperceptions and in particular which is the mosteffective surveying approach to observe usersrsquo atti-tudes through which type of questions (eg direct orindirect) and through which questions

(3) To what extent the probability of choosing a newautomotive technology may be significantly affectedby attitudesconcerns changes

e above-mentioned issues were addressed through theimplementation of an ldquoad hocrdquo experiment and a modelbased analysis aimed to infer the role of psychological factorsthrough the specification of latent variables within a Ran-dom Utility theory modelling framework

To this aim a Stated Preferences (SP) survey was carriedout on a sample of potential consumers [22] In particular

the survey aimed to collect the propensity to install theHySolarKit but it was specifically designed to ldquograsprdquo fivedifferent attitudesconcerns through a direct questioningapproach (attitudes towards the environment design fuelconsumption technology and reliability of technology)

Different types of questions were tested ldquodirectly relatedto the alternativerdquo and ldquoindirectly related to the alternativerdquoquestions (details in Section 42) Indeed existing literaturemakes use of direct questioning but usually uses alternativerelated questions only [23 24] whilst there are fewer in-stances of resorting to a mix of alternative and non-alternative related questions Experimental results were firstanalysed to investigate the robustness of the adopted hy-pothesesquestions then they were used for calibratinghybrid choice models (HCMs) with latent variables fol-lowed by a traditional binomial Logit model

e estimation results made it possible to understandthe effectiveness of the different surveying methods andthe role of psychological factors in the choice behaviourand also allowed for investigating if and how the in-clusion of attitudinal factors significantly increaseschoice model goodness-of-fit thus leading to much moreeffective results than traditional Logit choice modelsFinally a sensitivity analysis was carried out for assessingthe role of traditional instrumental attributes (egmonetary cost) as well as the impacts that the change ofusersrsquo attitudes may have on the propensity to install thenew technologies

e paper is organised as follows Section 2 introducesthe state of the art on existing approaches and modellingsolutions to the simulation of the propensity to choose newautomotive technologies Section 3 introduces the theoret-ical frameworks whereas Section 4 describes the experi-mental framework introducing the case study the datacollection and the preliminary analyses Finally Section 5discusses the estimation results e main conclusions aresummarised in Section 6

2 Intention to Choose Electric or HybridElectric Automotive Technologies ModellingApproaches and Determinants

e interest in alternative fuel vehicles (AFV) relies on awide range of literature whose main focus is that of Electric(EVs) Hybrid (HEVs) or Plug-In Electric vehicles (PHEVs)

As highlighted in studies of de Luca et al [16] andCartenı et al [25] the choice phenomenon cannot be solelyinterpreted within a unique theoretical framework As amatter of fact eliciting the ldquopreferencerdquo for AFVs is quitecomplicated due to several factors such as the recentadoption of such technologies the unavailability of salesdata the difficulty of estimating the ownership costs and theimpacts on driving habits

Behavioural approaches on EVs choice mainly rely ontwo mainstreams theories (i) psychological and sociologicaltheories (eory of Planned BehaviourmdashTPB Value-Belief-Norm theorymdashVBN habits diffusion of innovation dif-fusion theory) and (ii) Consumer Choice eory (CCT)





















































Reference





































middot middot


g Ini( 1113857 1113946 RLVni

Ini

LVn

λ1113888 1113889hLVni

LVni

Xni

β1113888 1113889dLVni (4)



expressed as

P yniIni

Xni

βx βLV1113888 1113889 1113946RLVniP

yni

Xni

βx βLV1113888 1113889

middot fIni

Ini

LVn

λ1113888 1113889hLVni

LVni

Xni

β1113888 1113889dLVni

(5)



L ΠiP yniIni

Xni

βx βLV1113888 1113889

1113946 RLVniΠiP

yni

Xni

βx βLV1113888 1113889fIni

Ini

LVn

λ1113888 1113889hLVni

LVni

Xni

β1113888 1113889dLVni

(6)















β

β

Latentvariables





Utility

Choiceindicators



Errorterm

Errorterm

Error

Error

Error

λ1

λ2

λ3
























































Att_Consump


Totally agree

Continuous


Continuous


Continuous

Δcost



















Table 5 Continued




























AttributeHCM











Consumption


Qcons










Error+0787

Error1

Error1

Error+143

Error+165

Error+118

Error+0983

Error1

Error+113

Error+138

+0148

+160

+184

+0495

+0729

+0396

1

1

1

Utility Design

Environment

























Qcons Qdesign Qenv






































ndash1

ndash1 ndash2

ndash4

ndash6

ndash6ndash8

ndash8

ndash4

ndash2

0 10 20 30 40 50 60

∆ Pr

obab

ility

to in

stal

l the

kit

()

∆ Pr

obab

ility

to in

stal

l the

kit

decr

ease

s

0

ndash2

ndash4

ndash6

ndash8

ndash10∆ cost


HCMBNL






6 Conclusions










3 4 4 5 6 7 8 9 10 11

10ndash3

ndash6

ndash9ndash12

ndash14ndash17

ndash19ndash21

ndash23ndash25

20 30 40 50 60 70 80 90 100∆

Prob

abili

ty to

inst

all t

he k

it (

)




∆ Pr

obab

ility

to in

stal

l the

kit

incr

ease

s

141210

86420



EnvironmentDesign






















Appendix




Data Availability




Acknowledgments


References























































































































































































Reference





































middot middot


g Ini( 1113857 1113946 RLVni

Ini

LVn

λ1113888 1113889hLVni

LVni

Xni

β1113888 1113889dLVni (4)



expressed as

P yniIni

Xni

βx βLV1113888 1113889 1113946RLVniP

yni

Xni

βx βLV1113888 1113889

middot fIni

Ini

LVn

λ1113888 1113889hLVni

LVni

Xni

β1113888 1113889dLVni

(5)



L ΠiP yniIni

Xni

βx βLV1113888 1113889

1113946 RLVniΠiP

yni

Xni

βx βLV1113888 1113889fIni

Ini

LVn

λ1113888 1113889hLVni

LVni

Xni

β1113888 1113889dLVni

(6)















β

β

Latentvariables





Utility

Choiceindicators



Errorterm

Errorterm

Error

Error

Error

λ1

λ2

λ3
























































Att_Consump


Totally agree

Continuous


Continuous


Continuous

Δcost



















Table 5 Continued




























AttributeHCM











Consumption


Qcons










Error+0787

Error1

Error1

Error+143

Error+165

Error+118

Error+0983

Error1

Error+113

Error+138

+0148

+160

+184

+0495

+0729

+0396

1

1

1

Utility Design

Environment

























Qcons Qdesign Qenv






































ndash1

ndash1 ndash2

ndash4

ndash6

ndash6ndash8

ndash8

ndash4

ndash2

0 10 20 30 40 50 60

∆ Pr

obab

ility

to in

stal

l the

kit

()

∆ Pr

obab

ility

to in

stal

l the

kit

decr

ease

s

0

ndash2

ndash4

ndash6

ndash8

ndash10∆ cost


HCMBNL






6 Conclusions










3 4 4 5 6 7 8 9 10 11

10ndash3

ndash6

ndash9ndash12

ndash14ndash17

ndash19ndash21

ndash23ndash25

20 30 40 50 60 70 80 90 100∆

Prob

abili

ty to

inst

all t

he k

it (

)




∆ Pr

obab

ility

to in

stal

l the

kit

incr

ease

s

141210

86420



EnvironmentDesign






















Appendix




Data Availability




Acknowledgments


References




































































































































g Ini( 1113857 1113946 RLVni

Ini

LVn

λ1113888 1113889hLVni

LVni

Xni

β1113888 1113889dLVni (4)



expressed as

P yniIni

Xni

βx βLV1113888 1113889 1113946RLVniP

yni

Xni

βx βLV1113888 1113889

middot fIni

Ini

LVn

λ1113888 1113889hLVni

LVni

Xni

β1113888 1113889dLVni

(5)



L ΠiP yniIni

Xni

βx βLV1113888 1113889

1113946 RLVniΠiP

yni

Xni

βx βLV1113888 1113889fIni

Ini

LVn

λ1113888 1113889hLVni

LVni

Xni

β1113888 1113889dLVni

(6)















β

β

Latentvariables





Utility

Choiceindicators



Errorterm

Errorterm

Error

Error

Error

λ1

λ2

λ3
























































Att_Consump


Totally agree

Continuous


Continuous


Continuous

Δcost



















Table 5 Continued




























AttributeHCM











Consumption


Qcons










Error+0787

Error1

Error1

Error+143

Error+165

Error+118

Error+0983

Error1

Error+113

Error+138

+0148

+160

+184

+0495

+0729

+0396

1

1

1

Utility Design

Environment

























Qcons Qdesign Qenv






































ndash1

ndash1 ndash2

ndash4

ndash6

ndash6ndash8

ndash8

ndash4

ndash2

0 10 20 30 40 50 60

∆ Pr

obab

ility

to in

stal

l the

kit

()

∆ Pr

obab

ility

to in

stal

l the

kit

decr

ease

s

0

ndash2

ndash4

ndash6

ndash8

ndash10∆ cost


HCMBNL






6 Conclusions










3 4 4 5 6 7 8 9 10 11

10ndash3

ndash6

ndash9ndash12

ndash14ndash17

ndash19ndash21

ndash23ndash25

20 30 40 50 60 70 80 90 100∆

Prob

abili

ty to

inst

all t

he k

it (

)




∆ Pr

obab

ility

to in

stal

l the

kit

incr

ease

s

141210

86420



EnvironmentDesign






















Appendix




Data Availability




Acknowledgments


References

























































































































































































Att_Consump


Totally agree

Continuous


Continuous


Continuous

Δcost



















Table 5 Continued




























AttributeHCM











Consumption


Qcons










Error+0787

Error1

Error1

Error+143

Error+165

Error+118

Error+0983

Error1

Error+113

Error+138

+0148

+160

+184

+0495

+0729

+0396

1

1

1

Utility Design

Environment

























Qcons Qdesign Qenv






































ndash1

ndash1 ndash2

ndash4

ndash6

ndash6ndash8

ndash8

ndash4

ndash2

0 10 20 30 40 50 60

∆ Pr

obab

ility

to in

stal

l the

kit

()

∆ Pr

obab

ility

to in

stal

l the

kit

decr

ease

s

0

ndash2

ndash4

ndash6

ndash8

ndash10∆ cost


HCMBNL






6 Conclusions










3 4 4 5 6 7 8 9 10 11

10ndash3

ndash6

ndash9ndash12

ndash14ndash17

ndash19ndash21

ndash23ndash25

20 30 40 50 60 70 80 90 100∆

Prob

abili

ty to

inst

all t

he k

it (

)




∆ Pr

obab

ility

to in

stal

l the

kit

incr

ease

s

141210

86420



EnvironmentDesign






















Appendix




Data Availability




Acknowledgments


References















































































































































Table 5 Continued




























AttributeHCM











Consumption


Qcons










Error+0787

Error1

Error1

Error+143

Error+165

Error+118

Error+0983

Error1

Error+113

Error+138

+0148

+160

+184

+0495

+0729

+0396

1

1

1

Utility Design

Environment

























Qcons Qdesign Qenv






































ndash1

ndash1 ndash2

ndash4

ndash6

ndash6ndash8

ndash8

ndash4

ndash2

0 10 20 30 40 50 60

∆ Pr

obab

ility

to in

stal

l the

kit

()

∆ Pr

obab

ility

to in

stal

l the

kit

decr

ease

s

0

ndash2

ndash4

ndash6

ndash8

ndash10∆ cost


HCMBNL






6 Conclusions










3 4 4 5 6 7 8 9 10 11

10ndash3

ndash6

ndash9ndash12

ndash14ndash17

ndash19ndash21

ndash23ndash25

20 30 40 50 60 70 80 90 100∆

Prob

abili

ty to

inst

all t

he k

it (

)




∆ Pr

obab

ility

to in

stal

l the

kit

incr

ease

s

141210

86420



EnvironmentDesign






















Appendix




Data Availability




Acknowledgments


References



















































































































































AttributeHCM











Consumption


Qcons










Error+0787

Error1

Error1

Error+143

Error+165

Error+118

Error+0983

Error1

Error+113

Error+138

+0148

+160

+184

+0495

+0729

+0396

1

1

1

Utility Design

Environment

























Qcons Qdesign Qenv






































ndash1

ndash1 ndash2

ndash4

ndash6

ndash6ndash8

ndash8

ndash4

ndash2

0 10 20 30 40 50 60

∆ Pr

obab

ility

to in

stal

l the

kit

()

∆ Pr

obab

ility

to in

stal

l the

kit

decr

ease

s

0

ndash2

ndash4

ndash6

ndash8

ndash10∆ cost


HCMBNL






6 Conclusions










3 4 4 5 6 7 8 9 10 11

10ndash3

ndash6

ndash9ndash12

ndash14ndash17

ndash19ndash21

ndash23ndash25

20 30 40 50 60 70 80 90 100∆

Prob

abili

ty to

inst

all t

he k

it (

)




∆ Pr

obab

ility

to in

stal

l the

kit

incr

ease

s

141210

86420



EnvironmentDesign






















Appendix




Data Availability




Acknowledgments


References










































































































































Consumption


Qcons










Error+0787

Error1

Error1

Error+143

Error+165

Error+118

Error+0983

Error1

Error+113

Error+138

+0148

+160

+184

+0495

+0729

+0396

1

1

1

Utility Design

Environment

























Qcons Qdesign Qenv






































ndash1

ndash1 ndash2

ndash4

ndash6

ndash6ndash8

ndash8

ndash4

ndash2

0 10 20 30 40 50 60

∆ Pr

obab

ility

to in

stal

l the

kit

()

∆ Pr

obab

ility

to in

stal

l the

kit

decr

ease

s

0

ndash2

ndash4

ndash6

ndash8

ndash10∆ cost


HCMBNL






6 Conclusions










3 4 4 5 6 7 8 9 10 11

10ndash3

ndash6

ndash9ndash12

ndash14ndash17

ndash19ndash21

ndash23ndash25

20 30 40 50 60 70 80 90 100∆

Prob

abili

ty to

inst

all t

he k

it (

)




∆ Pr

obab

ility

to in

stal

l the

kit

incr

ease

s

141210

86420



EnvironmentDesign






















Appendix




Data Availability




Acknowledgments


References


























































































































































Qcons Qdesign Qenv






































ndash1

ndash1 ndash2

ndash4

ndash6

ndash6ndash8

ndash8

ndash4

ndash2

0 10 20 30 40 50 60

∆ Pr

obab

ility

to in

stal

l the

kit

()

∆ Pr

obab

ility

to in

stal

l the

kit

decr

ease

s

0

ndash2

ndash4

ndash6

ndash8

ndash10∆ cost


HCMBNL






6 Conclusions










3 4 4 5 6 7 8 9 10 11

10ndash3

ndash6

ndash9ndash12

ndash14ndash17

ndash19ndash21

ndash23ndash25

20 30 40 50 60 70 80 90 100∆

Prob

abili

ty to

inst

all t

he k

it (

)




∆ Pr

obab

ility

to in

stal

l the

kit

incr

ease

s

141210

86420



EnvironmentDesign






















Appendix




Data Availability




Acknowledgments


References



































































































































































ndash1

ndash1 ndash2

ndash4

ndash6

ndash6ndash8

ndash8

ndash4

ndash2

0 10 20 30 40 50 60

∆ Pr

obab

ility

to in

stal

l the

kit

()

∆ Pr

obab

ility

to in

stal

l the

kit

decr

ease

s

0

ndash2

ndash4

ndash6

ndash8

ndash10∆ cost


HCMBNL






6 Conclusions










3 4 4 5 6 7 8 9 10 11

10ndash3

ndash6

ndash9ndash12

ndash14ndash17

ndash19ndash21

ndash23ndash25

20 30 40 50 60 70 80 90 100∆

Prob

abili

ty to

inst

all t

he k

it (

)




∆ Pr

obab

ility

to in

stal

l the

kit

incr

ease

s

141210

86420



EnvironmentDesign






















Appendix




Data Availability




Acknowledgments


References







































































































































6 Conclusions










3 4 4 5 6 7 8 9 10 11

10ndash3

ndash6

ndash9ndash12

ndash14ndash17

ndash19ndash21

ndash23ndash25

20 30 40 50 60 70 80 90 100∆

Prob

abili

ty to

inst

all t

he k

it (

)




∆ Pr

obab

ility

to in

stal

l the

kit

incr

ease

s

141210

86420



EnvironmentDesign






















Appendix




Data Availability




Acknowledgments


References























































































































































Appendix




Data Availability




Acknowledgments


References




































































































































DidAttitudesInterpretandPredict“Better”Choice...

Documents

Transcript of DidAttitudesInterpretandPredict“Better”Choice...