Dynamic Contests with Feedback and Heterogeneity: Empirical Evidence from a Gaming App

This paper empirically analyzes how performance feedback and information on heterogeneity affect behavior in dynamic contests, using data on two-player-contests from a smartphone/tablet application called ``Wordblitz for Friends''. We find that players increase output as underdogs and decrease output as favorites, where being an underdog (favorite) refers to being matched against an opponent who was more (less) successful in previous contests. Performance feedback, i.e. information about past performance within one particular contest, induces players to reduce output when being ahead and increase output when being behind. The strength of these effects increases in output differences. An analysis of the interaction of information on heterogeneity and performance feedback reveals that the output reducing effect of being ahead is entirely driven by favorites, as players generally increase their output when being an underdog, independent of being ahead or behind.