Do experienced subjects bias experimental results? Evidence from 16 laboratories in six countries

Abstract This paper addresses an area of growing concerns for laboratory researchers: is subjects’ behaviour affected by prior experiences in laboratory experiments? We address the question with a large and highly diverse international dataset and an operationalisation strategy that allows our findings to cohere with previous work while shedding new light on future research. The findings presented in this article are drawn from original data gathered as part of one of the largest tax compliance experiments ever conducted, involving more than 3,000 participants in six countries, across 16 different laboratories. Our results reveal that subjects’ behaviour correlates with their past experimental experiences, in a way that could bias results and compromise a study’s external validity. However, this change in behaviour due to experiences occurs only after subjects have participated in at least two previous laboratory experiments. The findings have implications not just for tax compliance research, but for allocation experiments more generally and for participant recruitment particularly.