Pressure in Crowdsourcing Contests

This project investigates the relationship between pressure and performance. We analyze 1,677 unique coders competing in 38 software algorithm competitions with random assignment. Part of these contestants compete among the top-performers of a low-ability group, the others compete among the bottom-performers of a high-ability group. We compare the performance of these contestants using a Regression Discontinuity (RD) design. We find that bottom-performers of a high-ability group are characterized by a performance that is 22% lower than that of contestants who have the same ability-level but compete as top-performers of a low-ability group. A decrease in effort cannot explain the performance differentials we observe. Instead, we find that a higher willingness to take risks and psychological factors like choking under pressure affect the problem-solving behavior of the contestants.