Abstract
This article shows that including inconsistent subjects in a Holt-and-Laury analysis will bias the mean, as well as the variance of the risk attitudes of the subject group of interest to an extent that cannot be determined a priori and that must not be neglected. One might be tempted to simply drop inconsistent subjects from the analysis to avoid such biases in a population-level analysis. Unfortunately, however, this is not a solution: first, the sample size may fall to an unacceptably low level. Second – and even more important – simply dropping inconsistent subjects from the analysis may introduce another unknown bias since systematic differences may exist in the risk preferences of those who answer consistently and those who do not. One must thus conclude that, if the group of interest contains a large proportion of inconsistent subjects, the whole set-up of the Holt-and-Laury lottery (HLL) experiment must be critically reconsidered and the experiment eventually repeated.
Acknowledgements
We thank the editor and two anonymous referees. We gratefully acknowledge the financial support received from the German Research Foundation (DFG) and the ScienceCampus Halle (WCH).