We consider an observer who makes a nite number of observations of an industry producing a homogeneous good, where each observation consists of the market price and rm-specic production quantities. We develop a revealed preference test (in the form of a linear program) for the hypothesis that the rms are playing a Cournot game, assuming that they have convex cost functions that do not change and the observations are generated by the demand function varying across observations. Extending this basic result, we develop tests for the case where (in addition to changes to demand) rms' cost functions may vary across observations. We also develop tests of Cournot interaction in cases where there are multiple products and where cost functions may be non-convex. Applying these results to the crude oil market, we show that Cournot behavior is strongly rejected. 

VIEW PAPER