Strategic Teaching and Equilibrium Models of Repeated Trust and Entry Games

This paper tests a model of strategic teaching in repeated trust and entry games with incomplete information. The model assumes 'short-run' players follow a one-parameter learning model (functional experience-weighted attraction). 'Long-run' players either realize how others are learning and 'teach' by maximizing their long-run payoff, or always behave honestly or aggressively. For precision, the fraction of honest/aggressive types was first measured in an experiment with one-shot games. Using data from 28 experimental sessions of eight-period trust and entry supergames (25,000 observations), the model fits modestly better than a quantal-response equilibrium benchmark, and both models predict much more accurately than chance. Estimates show most players are sophisticated, and become more sophisticated with experience. Direct measures of subjects' beliefs are weakly correlated with implicit model beliefs, but are extremely accurate and do not show the overconfidence found in many psychological studies.