International Choice Modelling Conference, International Choice Modelling Conference 2015

Font Size: 
Incorporating a budget constraint into discrete choice experiments
John Rose

Last modified: 11 May 2015


Ever since the very first stated preference experiment (Thurston 1931), economists have derided the approach claiming that the hypothetical nature of such experiments will result in respondents systematising their answers in such a way as to produce plausible but spurious results (Wallis and Friedman 1942). The same arguments against the use of stated preference surveys remain today (e.g., Camerer and Hogarth 1999; Diamond and Hausman 1994; List 2001). Indeed, encapsulated by the concept of incentive compatibility, there exists growing evidence supporting the early criticisms of Wallis and Friedman (1942) that respondents acting in a utility maximising manner may act in a strategic manner rather than reveal their true preferences when answering stated preference surveys, including stated choice (SC) type response formats (Bateman et al. 2008; Carson and Groves 2007). Rather than abandon SC experiments however, a number of researchers have explored alternative strategies to either mitigate the ability of respondents to act in a way how they would if faced with similar choices in real markets, or minimise any biases that may arise if they do not make choices that reflect their true preferences.


Recent attempts to make SC choice tasks more realistic and less prone to hypothetical bias have taken many forms. Making SC choice tasks more incentive-compatible (e.g., Lusk et al. 2008; Ding 2007; Alfnes et al. 2006) and individual customisation of SC choice tasks to respondent specific experiences (e.g., Rose et al. 2008; Train and Wilson 2008) represent just two approaches researchers employ. Other strategies that have been explored include innovations such as the introduction of methods such as information acceleration (e.g., Urban et al. 1996, 1997) where researchers create a choice environment that mimics better the context in which future consumption will be made, the use of cheap talk (Champ et al. 2004; Carlsson et al. 2005; Ladenburg et al. 2010) to emphasise the importance of the answers respondent give and how these should reflect their true preferences, the use of certainty scales (e.g., Norwood 2005; Garcia et al. 2008; Ready et al. 2010) designed to force respondents to reflect on the choices they make, or giving each respondent only a solitary binary choice task involving the choice between a status quo and a single hypothetical alternative (Farquharson 1969) or informing respondents that only one choice task out of the sequence will be randomly selected and used for modelling purposes (Carson and Groves 2007; McNair et al. 2011) so that they believe that their specific choices matter. More recently, although not a SC experiment, Jacquemet et al. (2013) applied social psychology theories on the oath taking as a truth-telling-commitment device to assist in mitigating hypothetical bias in an auction based stated preference study. Other methods have focused on increasing the consequentiality of the choices respondents make when undertaking SC experiments (Vossler and Evans 2009; Herriges et al. 2010) by ensuring that respondents face some real outcome or consequence from the choices they make, referred to as incentive alignment (Harrison 2007). Typically such studies involve respondents being informed that at least one of the choices they make whilst undertaking a SC experiment will be selected at random and be binding (see e.g., Lusk and Schroeder 2004; List et al. 2006; Alfnes et al. 2006; Ding 2007; Carlsson et al. 2010; Moser et al. 2010).


The requirement that respondents experience one of their choices may not always be feasible, particularly when the experiment involves some new hypothetical alternative or technology (e.g., a new metro mode that is yet to be built), some non-market good that may not be readily accessible to them (e.g., a study designed to value the protection of a river system in Africa), or an scenario in which the alternatives are particularly costly (e.g., new car choice). As such, most studies tend to ignore the issue completely or rely on the soft approach of reminding respondents about the impact their choices will have on their current expenditure patterns, as with cheap talk (e.g., Arrow et al. 1993, Lomis et al. 1994). Even with such reminders however, it is possible for respondents undertaking SC tasks to ignore or underestimate their true budget constraints, which will significantly impact on the study outcome. Wardman (2001) and Brownstone and Small (2005) both argue that respondents tend to ignore budget constraints and select higher-cost alternatives more frequently in real life than in hypothetical surveys.


Added to this is the possibility that many SC experiments fail to fully account for the full set of viable alternatives that individuals face in real markets. For example, most mode choice experiments involve respondents selecting from amongst a set of competing modes, however a decision maker operating in a real market might also have the option to select within their current mode, but at an earlier or later time. At the very least, failure to capture the full set of choices available to respondents may invalidate the task, even if one carefully handles the issue of incentive compatibility.


In this work, we present the results of a SC experiment involving automobile choice whereby respondents are first asked about their household expenditure and current vehicle fleet, after which they complete the SC task. A split sample is employed where half the respondents are asked after making their choice, how they would change their vehicle fleet (if at all) by selling one or more existing vehicles, and given that, how their expenditure would change. Respondents are further allowed to change their vehicle choice, after considering the implications upon the household budget. In this way, the task is designed to better reflect the decisions that households face when purchasing a new vehicle, in terms of possible vehicle replacement, as well as forces respondents to equilibrate their expenditure in light of the selected vehicle.

Conference registration is required in order to view papers.