Procedural Rationality and Equilibrium Trust*


  • *

     This article was written while both authors were visiting the Santa Fe Institute. We thank the Institute, and especially Sam Bowles, for their hospitality and Carlos Rodriguez for comments on an earlier version. We also thank two referees and the Editor for suggesting a number of improvements.


This article examines the determinants of steady state trust in a population of principals and agents, where the former learn from experience using boundedly rational procedures. For any distribution of agent types, the long-run distribution of principal behaviour is characterised. Heterogeneity in the behaviour of principals persists under both the sampling procedure (Osborne and Rubinstein, 1998) and the maximum average procedure (Rustichini, 2003). Despite its greater sophistication, the maximum average procedure can result in poorer performance than the sampling procedure, both from the perspective of the principal and also with respect to aggregate payoffs.