Item

Abstract

We study a dynamic multi-stage contest that resolves in each stage only with a given probability. Assuming that there is unobservable heterogeneity in intrinsic motivations we derive properties of the equilibrium efforts across the different stages. Whereas in the corresponding complete information benchmark equilibrium efforts are stable across the stages, uncertainty about the type distribution of possible opponents generates learning. We identify reasons for dynamic adjustments of efforts caused by belief formation and updating and by selection of certain types into continuing conflict. A corresponding experimental setup provides evidence for escalation of efforts in later stages, for type heterogeneity, for belief formation and belief updating, and for self-selection. Overall, our results suggest the importance of an appropriate benchmark model when testing predictions on behavior in conflict or related strategic interactions.