Classical and Belief-Based Gift Exchange Models: Theory and Evidence

We derive, compare, and test the predictions of three models of gift exchange: Classical (CGE); Augmented (AGE) based on unexpected wage surprises and first order beliefs; and Belief-based (BGE) that uses second order beliefs to formally model guilt-aversion. Motivated by Akerlof (1982), we also introduce signals of exogenous industry wage norms, θw, and effort norms, θe. We study the effects on the worker’s optimal effort of exogenous variation in the wage, w, the signals θw, θe, and a signal of firm’s expectations of effort from the worker, s. All three models successfully predict gift exchange (higher effort in response to higher w). The AGE and the CGE models fail to explain the data for the worker’s effort responses to the signals θw, θe, and s. The BGE model successfully explains the data in all these respects. Gift exchange is underpinned by guilt-aversion and a formal modelling of second order conditional beliefs.