Events at CERGE-EI

Libor Dušek, Ph.D.

Department of Economics, Faculty of Law, Charles University
________________________________________
Authors: Libor Dušek and Christian Traxler

Abstract: This paper studies how punishment for past offenses shapes future compliance behavior via learning.
The context of our study is traffic law enforcement through automated speed cameras. We use unique data on speeding tickets and full driving histories of more than one million cars tracked over several years in a suburb of Prague. In our setting, punishment neither implies incapacitation nor do past tickets alter the ‘price’ for future offenses. This allows us to identify specific deterrence effects induced by learning from law enforcement.

We present results from two empirical strategies. Firstly, a regression discontinuity design exploits two speed level cutoffs which provide variation in punishment at the extensive (receiving a speeding ticket) and intensive margin (tickets with low or high fines), respectively. The RDD reveals strong and precisely estimated responses to speeding tickets: the speeding rate drops by a third (10 percentage points) and chances of getting a further ticket fall by 70%. An increase in punishment at the intensive margin -- a more than a doubling of fines -- triggers only a limited additional effect. Secondly, an event study makes use of the high-frequency nature of our data. The average treatment effects on the treated obtained from the event study confirms all LATEs from the RDD. We also document that driving responses are immediate and very persistent over time. Even two years after receiving a ticket there is no evidence on ‘backsliding’ towards speeding. The results reject unlearning and temporary salience effects and support a reinforcement learning model in which agents, after experiencing punishment, update their priors on the expected costs of future offending in a discontinuous, ‘coarse’ manner. Additional results indicate that learning from (local) law enforcement affects drivers' behavior more broadly, including spillovers on non-ticketed drivers.

________________________________________
Full Text: “Learning from Law Enforcement”