This paper is a survey of applied work with auction data First, we sum
marize the pre-game theoretic competitive bidding literature based on
decision theory and the associated empirical work, centered essentiall
y around the winner's curse debate. Then, we review the main distribut
ion free predictions of the three major game theoretic models, the ide
ntification of their stochastic structure, and their confrontation wit
h data. We study successively, the symmetric common value model, the i
ndependent private values model and the asymmetric common value model.
Conditional upon being in one of these models, game theoretic restric
tions can be tested. A major conclusion is that predictions are highly
dependent on the unobservable restrictions on the distributions of ch
aracteristics. The paper concludes with a suggestion to develop a stru
ctural econometrics of auction data with only a few restrictions on th
ese distributions.