Strategic Interactions and Peer Learning in Contests

Many educational systems rely on rank-based rewards to allocate college seats or grades. While rewarding relative performance can boost effort, it may also discourage cooperation, as students strategically limit knowledge sharing to protect their standing, potentially reducing total learning. This paper studies how competition shapes individual effort, peer interactions, and skill production, using a structural model and a tightly linked experiment. The model, in which heterogeneous agents make effort choices, formalizes how competition shapes interactions. The experiment varies competition intensity and peer matching to generate rich behavioral data for structural estimation, collected through a custom-designed learning platform. Results show that moderate competition increases both individual and peer learning, while intense competition shifts effort away from cooperation. Under intense competition, students with similar ability levels interact less than they normally would, crowding out peer learning. LLM-based interaction analysis confirms that cooperative language declines as competition intensifies. Structural model quantifies production and preference components. Peer learning entails lower disutility compared to individual learning. Barriers to participation are higher for lower-ability students, while peer learning helps reduce them. Counterfactual policies suggest that moderate rank-based incentives combined with piece-rate rewards improve learning behavior. When barriers to individual learning participation are reduced, equity improves, while efficiency gains are stronger when collaboration barriers are lowered.