Subject pool bias in economics experiments

Abstract

In this paper we consider data from a large number of economic experiments, and look for demographic effects that may be a source of subject pool bias if not carefully accounted for in the subsequent statistical analysis. Our dataset contains information on 2,408 subjects and 597 experimental sessions from 74 experiments recorded over more than 2 years at an experimental laboratory. Using different estimation methods and model specifications, we identify the significant demographic determinants of personal earnings, and find that they account for less than 4% of the observed variation. Thus we deliver empirical evidence supporting the experimental method as monetary incentives, and therefore some kind of strategic behavior, seem to be more important than demographics in the laboratory. Exploiting the timeseries nature of the data we also study some dynamic issues of the subject pool: we analyze the factors that influence subjects’ decisions on returning to the laboratory.