This paper develops a novel approach to estimate the incentive effects of contracts when the matching between agents and principals, and the contract features, derive from an auction process. The post-auction outcome of each agent depends on the principal-agent observable characteristics, the contract features and also on unobserved signals which drive how principals bid for the agents. We propose a two-step procedure which consists of, first, estimating the primitives of an interdependent values auction model which is shown to be non-parametrically identified and, second, constructing two control terms that reflect some expected signals of the auction winner and of the principal previously matched with the given agent. A Monte Carlo study shows that the extent of the bias can be large and that our estimator, which accounts for the endogeneity of contracts, can perform well, even in small samples. We apply our methodology to a labor market application: we estimate the effect of sports players' auction-determined wages on their individual performance.

joint work with Manasa Patnam (IMF) and Michael Visser (CREST and CRED Paris 2)