The relative performance of ex-ante and ex-post measures to mitigate hypothetical and strategic bias in a stated preference study

Abstract

Bias related to the hypothetical setting remains controversial regarding the reliability and validity of value estimates from discrete choice experiments (DCEs). This has motivated a large body of literature to investigate approaches for mitigating hypothetical and strategic bias. Our study provides further evidence to inform this debate by testing whether a combination of ex-ante or ex-post mitigation strategies might be effective in reducing bias in DCEs. Specifically, we employ individual and multiple ex-ante reminders alongside an ex-post data treatment and analyse how their individual or joint use affects willingness to pay (WTP) estimates. The econometric analysis makes use of innovative semi-parametric logit-mixed logit in addition to the state-of-the-art mixed logit model. The empirical case study focuses on preferences for the environmental and social impacts of organic olive production. By comparing the three experimental treatments with a control treatment, we test whether ex-ante cheap talk, a reminder of the project's relative spatial extent, or a combination of both affect stated WTP. In addition, we use an ex-post data treatment to correct WTP estimates. WTP estimates of treatments related to ex-ante mitigation strategies did not differ significantly from those obtained from a control treatment with standard budget constraint reminders. However, the ex-post approach results in a significant reduction in mean WTP estimates and is used to investigate whether the observed choice inconsistencies are due to unintentional errors or strategic behaviour. We argue that ex-post mechanisms deserve greater attention and highlight the need to distinguish strategic behaviour from other sources of hypothetical bias.