Nonparametric Analysis of Two Sided Markets

Résumé

This paper considers an empirical semiparametric model for two sided markets. Contrary to existing empirical literature on two sided markets, we do not rely on linear network effects and we specify them nonparametrically. The probability distribution functions of the two sides are not specified neither. The data on German magazine industry is used to estimate both the distribution functions and the network effect functions of readers and advertisers nonparametrically. We show that semi-nonparametric specification is well supported by the data and the network effects are not linear. With a numerical example we demonstrate that the predicted profits of the platforms with the linear network effects are 30 % higher than the case where the network effects are specified nonlinearly. The paper also develops a nonparametric specification test based on bootstrap. We show that linear specifications are rejected whereas quadratic approximations performs as well as the nonparametric alternative.