Abstract
This paper studies the relationship between the intensity of competition and innovative outcomes, using unique, detailed data on 1050 elite software programming contests. In these contests, competitors were randomly assigned to compete in groups of different size and in solving different problems. These characteristics of the contests provide us with sources of exogenous variation in the level of competition, as measured by number of contestants, and in the type of innovation or problem the coders were called to solve. We find that, for simpler problems, increasing the number of contestants reduces the overall innovative performance, consistent with a standard view of the negative impact of competition on innovation incentives.
For more complex problems, however, the addition of competitors increases the performance of at least the best coder in a given competition. This finding is consistent with a view of innovation that focuses less on incentives and "racing", and defines innovative effort as a search for solutions. Especially for complex problems, increasing the number of competitors also increases the diversity of approaches to the problems, thus making the search more likely to succeed. We discuss the implications of these novel findings about the competition-innovation link being contingent on the type of innovation, in relation to a number of contexts, including competition and innovation in markets, technology licensing problems, open innovation contexts, distributed innovation communities and platform strategies.