Opting-in: Participation bias in economic experiments

• Examine characteristics of laboratory participations to population recruited from.
• Subjects have less income, work less, volunteer more, greater interest in lab tasks.
• Magnitude of the participation bias is large; often 2 to 1 over-representation.
• Discuss recruitment and econometric methods to reduce biases.

Assuming individuals rationally decide whether to participate or not to participate in lab experiments, we hypothesize several non-representative biases in the characteristics of lab participants. We test the hypotheses by first collecting survey and experimental data from a typical recruitment population and then inviting them to participate in a lab experiment. The results indicate that lab participants are not representative of the target population on almost all the hypothesized characteristics, including having lower income, working fewer hours, volunteering more often, and exhibiting behaviors correlated with interest in experiments and economics. The results reinforce the commonly understood limits of laboratory research to make quantitative inferences. We also discuss several methods for addressing non-representative biases to advance laboratory methods for improving quantitative inferences and consequently increasing confidence in qualitative conclusions.