Addressing Attribute Value Substitution in Discrete Choice Experiments to Avoid Unintended Consequences

• Published in: Environmental & resource economics Vol. 77; no. 4; pp. 813 - 838
• Main Authors: Howard, Gregory, Roe, Brian E., Interis, Matthew G., Martin, Jay
• Format: Journal Article
• Language: English
• Published: Dordrecht Springer Netherlands 01.12.2020; Springer Nature B.V
• Subjects: Attributes; Beliefs; Bias; Design; Discrete choice; Econometrics; Economic Policy; Economics; Economics and Finance; Ecosystem services; Ecosystems; Efficiency; Environmental Economics; Environmental Law/Policy/Ecojustice; Environmental Management; Estimates; Experiments; Farmers; Marketing; Mathematical models; Parameter estimation; Preferences; researchers; Substitutes; transportation; Value; Welfare; Subjective beliefs; Attribute value substitution; Discrete choice; Scenario adjustment
• ISSN: 0924-6460; 1573-1502
• DOI: 10.1007/s10640-020-00519-9

Choice experiments are a popular method of generating stated preference data for a variety of fields from marketing to health, transportation and environmental economics. They allow researchers to systematically vary choice attributes in a manner that can both increase estimation efficiency and allay endogeneity concerns. An increasing number of studies have included elicited subjective beliefs in their stated preference models. We discuss why this strategy may be warranted in some cases, specifically when the researcher suspects subjects will engage in attribute value substitution or scenario adjustment. While there are multiple ways one could integrate subjective beliefs, in all cases a proper understanding of the econometric ramifications of their inclusion is necessary. We show that excluding subjective beliefs yields biased parameter estimates yet policy-relevant welfare measures, whereas including subjective beliefs yields unbiased parameter estimates but can generate less policy-relevant welfare estimates. We demonstrate how policy-relevant welfare measures should be calculated from models that include subjective beliefs and illustrate our theory with an application to payment for ecosystem services to farmers.