Factors Explaining the Hypothetical Bias: How to Improve Models for Meta-analyses

Abstract

Meta-analyses are getting more common in economics. However, little information is available regarding the choice of econometric models and its impact on results. Moreover, outlier data are common in meta-analyses and some authors have simply chosen to remove arbitrarily such data. We use the rich literature of meta-analysis on hypothetical bias (HB) related to contingent valuation methods to illustrate our point. More specifically, we review and update the meta-analyses of HB using a Meta-Regression Hierarchical Mixed Effect (MRHME) model and we apply a Bayesian Gibbs Sampling as classical results robustness check. A set of 462 observations from 87 economic valuation studies is used to this effect. The findings indicate that MRHME model is more efficient to explain HB. While respondents overstate their stated willingness-to-pay for a good by a factor of two, cheap talk, certainty correction, Ex Ante and Ex Post mitigation techniques significantly reduce the HB. Notwithstanding, mitigation techniques are more effective in private goods economic valuation.