Please click here if you are unable to view this page.
TOPIC:
PRICE DISCRIMINATION, SWITCHING COSTS AND WELFARE: EVIDENCE FROM THE DUTCH MORTGAGE MARKET
ABSTRACT
In many markets with switching costs firms charge a lower price to new customers than to existing customers, a practice called history-based price discrimination. By exploiting a ban on history-based price discrimination in the Dutch mortgage market, this paper estimates the effects of history-based price discrimination on consumer surplus, profits and welfare. These effects are theoretically ambiguous because history-based price discrimination can make markets more or less competitive. I estimate a structural model, of which the supply side consists of a dynamic game. To deal with the curse of dimensionality, I employ techniques from machine learning to reduce the dimension of this game’s state space. I implement these techniques in a new estimation method for dynamic games, which I justify with a new solution concept: Sparse Markov Perfect Equilibrium (SMPE). In an SMPE, firms optimally pay attention to a subset of state variables instead of the full state. Therefore, the state space is considerably smaller than under the standard assumption of Markov Perfect Equilibrium. I show that the Lasso identifies which variables firms pay attention to. For an average mortgage, banning history-based price discrimination increases welfare by €125 per year and consumer surplus by €415 per year, while bank profits drop by €290 per year.