International Choice Modelling Conference, International Choice Modelling Conference 2019

Font Size: 
Controlling for endogeneity of perceived consequentiality in preference modelling
Ewa Zawojska, Wiktor Budziński, Mikołaj Czajkowski

Last modified: 12 July 2019

Abstract


Stated preference, discrete choice experiment (DCE) surveys find application in many fields, such as environmental economics and transportation, to estimate the value of public goods. The value estimates are used to evaluate benefits in cost-benefit analyses of public policy projects and to estimate losses in litigation processes over natural damages, among others. In order for respondents to reveal preferences truthfully in such surveys, it is recommended to make the surveys be (viewed by respondents as) consequential; that is, the survey result should be seen as potentially influencing actual decisions regarding the public good provision and the related payment collection. Although consequentiality is recognised as a necessary condition for truthful preference revelation, research and guidance on how to elicit respondents’ perceptions about survey consequentiality are scarce. What seems even scarcer is the investigation on how to correctly include the consequentiality perceptions as reported by respondents in the econometric models of preferences revealed through choices in DCE surveys. This study aims at undertaking these research questions.

 

Information on consequentiality perceptions is typically collected in DCE surveys via a follow-up question(s) after the preference elicitation, such as how likely do you think it is that the result of this survey will affect final decisions regarding the project presented? A question such as this is commonly answered on a several-level Likert scale. This method of eliciting the perceptions is widely accepted and applied, but hardly any research has investigated alternative ways of the perception elicitation. To the best of our knowledge, the only stated preference studies that used more than one consequentiality perception indicator questions in preference modelling are Vossler and Holladay (2018) and Zawojska et al. (2019), while the only study that played with the location of the consequentiality perception question (that is, before or after preference elicitation) is Lloyd-Smith et al. (forthcoming).

 

One of the main challenges with including stated consequentiality perceptions in preference modelling is potential endogeneity. Endogeneity can arise because responses to the consequentiality question(s) are likely driven by similar factors, unobserved from the perspective of a researcher, as those which affect responses to preference elicitation questions. The outcome that the majority of empirical studies find a positive relationship between consequentiality and the value of a good considered in a survey can possibly result from endogeneity, An early study that pointed to this problem was Herriges et al. (2010), however, since then, despite the use of consequentiality perceptions in preference modelling getting broader and broader, the problem has not been solved, neither comprehensively addressed. A recent study by Lloyd-Smith et al. (forthcoming) revives this issue by proposing a special regressor approach to control for the potential endogeneity when modelling preferences revealed through choices in a single dichotomous choice question and including at the same time a single indicator of perceived consequentiality.

 

Our study further builds upon the existing literature and contributes in three important directions. First, in modelling preferences revealed in a DCE survey, we control for consequentiality perceptions, which are measured by several statements (in contrast to the commonly used, single question). Second, we do not simplify respondents’ answers to the consequentiality questions (for instance, by transforming Likert scale responses into a zero-one-coded variable), but we use them as stated by respondents. Third, we model the preferences using an augmented hybrid choice framework that allows for controlling potential endogeneity of the consequentiality indicator questions (Budzinski and Czajkowski, 2018). The hybrid choice model helps decompose several consequentiality statements into a (smaller) number of latent factors and investigate how these different dimensions of consequentiality affect preferences. The model makes it simultaneously possible to examine whether the location of the consequentiality perception questions in the survey influences stated consequentiality perceptions, and whether it changes the effect of the perceived consequentiality on stated preferences.

 

We use data from a DCE survey that inquires about preferences towards extending the theatre offer in Poland. The considered project involves increasing the number of shows and premiers within entertainment, drama repertory, children and experimental performances. The study includes two treatments differing with respect to the location of the consequentiality perception questions in the survey—they are either located before choice tasks used for preference elicitation or after them. The survey was conducted in Poland from September to December 2018, and the data was collected through Computer Assisted Web Interviews (CAWI). The survey involved a sample of 2,864 respondents representative for the population of Poland.

 

We believe that results of this study can contribute twofold to the methodology of DCE research. First, as this is the first implementation of a set of consequentiality perception questions in modelling of preferences, this may provide insights on advantages and drawbacks of this alternative way of collecting data on consequentiality perceptions. This could help applied researchers design such questions in future DCE studies, informing at the same time about the preferred location of such questions in the survey (before versus after preference elicitation), among others. Second, to the best of our knowledge, this is the first empirical application of the augmented hybrid choice framework that might address the potential problem of endogeneity. This can provide a substantial methodological advancement, given the importance of controlling for (possibly endogenous) perceptions and attitudes in choice modelling.

 

Budziński, W., & Czajkowski, M. (2018). Hybrid choice models vs. endogeneity of indicator variables: a Monte Carlo investigation. WNE UW Working Paper no. 21/2018.

 

Herriges, J., Kling, C., Liu, C. C., & Tobias, J. (2010). What are the consequences of consequentiality? Journal of Environmental Economics and Management, 59(1):67-81.

 

Lloyd-Smith, P., Adamowicz, W., & Dupont, D., (forthcoming). Incorporating stated consequentiality questions in stated preference research. Land Economics.

 

Vossler, C. A., & Holladay, J. S. (2018). Alternative value elicitation formats in contingent valuation: Mechanism design and convergent validity. Journal of Public Economics, 165:133-145.

 

Zawojska, E., Bartczak, A., & Czajkowski, M. (2019). Disentangling the effects of policy and payment consequentiality and risk attitudes on stated preferences. Journal of Environmental Economics and Management, 93:63-84.


Conference registration is required in order to view papers.