Anchoring Trust. Order of moves and agent behavior in two Trust Games

Abstract The paper reports and discusses the results of a randomized experiment testing whether the order of moves in two subsequent Trust Games, with reversed roles, affects the level of trust, measured by the share of their endowment that Trustors send to Trustees. We find that on average, participants playing Trustor in Game 2 (TSG) send a larger share to their partner compared to those who played Trustor in Game 1. We provide micro-level evidence that this aggregate effect is driven by anchoring. TSG subjects systematically anchor their decisions to the amount they observed being sent by their counterpart in the previous game. The absolute distance between their own choice and the observed choice is significantly smaller, and 20.9% of subjects exactly replicate the observed amount, i.e. more than twice what would be expected under random choice. Our findings suggest that anchoring, rather than learning or belief updating, is a key mechanism shaping trust, with clear behavioral and payoff implications, since TSG subjects earn significantly lower total payoffs. This is consistent with anchoring effect, since subjects increase trust by imitating an observed reference point, even when doing so reduces their own payoff.