Designing Choice Sets for Stated Preference Methods: The Effects of Complexity on Choice Consistency

We vary the complexity of choice sets used in stated preference methods to evaluate the impact on choice consistency. We define five measures of complexity that capture either the amount of information or the correlational structure of information in a choice set. Using these measures as randomly assigned treatments in a large experiment, we analyze the consistency of respondentsʹ choices using a heteroskedastic multinomial logit model. We find that: (i) all of our measures of choice set complexity affect choice consistency, (ii) changes in the correlational structure of information have the largest impact on choice consistency, and (iii) choice complexity significantly distorts welfare estimates. Our results confirm economistsʹ initial worry that moving from contingent valuation to conjoint methods will increase complexity-induced choice inconsistency. However, our analysis also shows that these negative impacts may be mitigated if precautions are taken at the design and estimation stages of stated preference methods.