Existing work on overcoming coordination failure suggests that an exogenous increase in incentives to coordinate can help groups escape from the productivity trap, but the effect is far from perfect.  Previous work on coordination games (Van Huyck, Battalio, and Beil, 1993), as well as related lab and field studies on incentives (i.e. Lazear, 2000; Dohmen and Falk, 2011), suggest that endogenous changes in incentives with self-selection into contracts will yield a larger effect than exogenously imposed contracts.  Our experimental design allows us to confirm this prediction and decomposes the causes of this effect between the direct effect of changing incentives, selection, and strategic anticipation.  We find that the difference between endogenous and exogenous contract is primarily due to selection.  Fitting a structural model of learning to our data, the key feature needed to track the data is heterogeneity in initial beliefs.  We use this model to predict a “zero sum” effect of incentive contracts.