International Choice Modelling Conference, International Choice Modelling Conference 2017

Font Size: 
Rewarding truthful-telling in stated preference studies
Pierre Mahieu, Romain Crastes, Jordan Louviere, Ewa Zawojska

Last modified: 28 March 2017


Stated preferences (SP), although widely used since late 80’s, is still under doubts on whether it provides true value estimates. The critique mainly points out that in SP surveys, respondents usually have no incentive to answer truthfully. Struggling to make the SP data as reliable as possible, researchers have proposed various ex-ante techniques, as well as formulated the conditions under which SP surveys are incentive compatible, that is, create such conditions under which a respondent’s optimal strategy is to reveal preferences truthfully. Because of important shortcomings of the existing approaches, in particular as the former are not based on economic theory, and because of important limitations placed on SP studies by incentive-compatibility requirements, we propose an alternative ex-ante method, which has economic theoretical properties and can improve SP data reliability.

On a basis of mechanism design theory, researchers have recently defined a set of conditions which assures incentive compatibility of a SP survey. Carson and Groves (2007) identify the following conditions: (a) the authority can enforce the payment by voters upon the program implementation (namely a coercive payment mechanism), (b) the valuation question is viewed as a take-it-or-leave-it offer, (c) the valuation question has a single, binary format (involves a yes-or-no answer), (d) participants care about the survey outcome(s), and (e) the probability that the proposed program is implemented increases monotonically with the proportion of votes in favor of it. The last two conditions are jointly termed as survey’s consequentiality by Carson and Groves (2007) – respondents who care about the survey’s outcome view their responses as potentially influencing the finally undertaken action. Vossler, Doyon, and Rondeau (2012) further develop the conditions for the case when a survey consists of a sequence of binary valuation questions, and set two more requirements to maintain an incentive compatible character of a survey. As seen, the conditions for incentive compatibility of SP surveys impose strict limitations on the design of the surveys.

The ex-ante methods are thought (and often empirically observed) to enhance incentives of SP respondents to truthfully reveal their preferences. The (i) “oath” approach asks respondents at the beginning of the survey to sign a form that they swear to tell the truth, (ii) a “ten commandments” approach asks respondents to adhere, to the best of their ability, to ethical guidelines about lying and (iii) an “honesty priming” approach invites honest behavior of respondents by exposing them prior to the valuation question to a task involving honesty concepts. Although the methods (i) and (iii) might appear effective in bringing the respondents’ hypothetical answers closer to their true valuations, the incentives they provide to encourage truthful preference revelation are not based on economic theory. In other words, according to economic theory, the behaviors of respondents who answer surveys with and without any of the proposed approaches should not differ; the approaches do not change the incentive structure.

We propose a novel method to reward participants for providing truthful answers in SP surveys. The idea behind our method is to employ lie detection device and reward participants who answer truthfully. The main advantages of our approach over methods (i) to (iii) are that economic theory provides predictions how the technique revises the strategies of respondents how to optimally answer in a SP survey.

Inconsequential surveys are commonly conducted. A possible strategy for rational individuals in this type of surveys is to provide random responses. Obviously, it takes some effort and time to respond truthfully to complex questions and it is unclear why rational people would invest such a level of effort in absence of incentive. First, we theoretically discuss how the lie detection method contributes to enhance reliability of stated preference data in inconsequential surveys. Second, we empirically test how the method works using data from a lab study conducted in France February 2015 where an oximeter is employed to record cardiac pulse. More precisely, we test if people take more time responding the valuation question and provide less random responses when their cardiac pulse is recorded. We find that people spend more time completing the valuation task in the lie detection treatment. Also, the variance of the error term is lower in an hybrid choice model that accounts for possible heterogeneity. Overall, lie detection approach seems to increase the validity of the SP data.

Conference registration is required in order to view papers.