Self-Selection Filters Irrationality in One-shot Games

In the field, individual scan choose to self-select into strategic interactions. Incontrast, once seated in the lab, subject shave little choice but to play games that they may have other wise avoided. We here add an explicit self-selection stage, in order to enhance the external validity of laboratory experiments. Using one-shot games, we show that self-selection is mainly driven by two variables: risk-aversion, and a measure of confidence that is new in the context of strategic interactions. Self-selection also greatly reduces the gap between theoretical predictions and actual behavior (for example, the fraction of subjects playing Nash equilibrium doubles).