Probability vs Severity of Punishment : the case of the non-point source pollution

Abstract

In a world of scarce resources it is a main concern for a regulator how to deal with crime in an optimal way. While we know that increases in the expected penalty facing violators increase the level of compliance, with resource constraints a regulator has to decide in order to create an enforcement strategy whether to enhance the probability of punishment (by investing in monitoring and security) or in the penalty facing violators (years in jail or fines)1 . To separate the effect of both and to get the optimal combination for the most effective deterrence is a main issue in economics of crime. A special case is when the regulator deals with non-point pollution problems. These problems arise when it is impossible for the regulator to know the origin of the pollution or it is extremely expensive to monitor for every individual so only the ambient pollution level in known. This could be the case of excess fertilizers, herbicides, toxic chemicals, lakes pollution, atmospheric deposition, hydromodification, etc. In this context, the regulator can choose the best mechanism to avoid pollution and to get the highest possible level of abatement of the firms. The utilization of different mechanisms sets up a good basis for researchers to study the behaviour of firms (choosing their level of abatement) with different combinations of probability of being caught and the severity of punishment when emissions exceed the ambient pollution level set by the regulator. However, there is not so much empirical evidence or at least it is not conclusive. The main issue about the empirical analyses is the difficulty in obtaining good data. In particular, in most situations, individual data is not available and it can only be collected at the aggregate through various calculations and estimations. Also, as pointed out by Anderson et al (2003), “for many offenses the probability of punishment is computed as the ratio of the number of convictions to the number of reported offenses. Because reporting is costly to the victim, offenses are often underreported and thus there may be systematic measurement error in this variable that can bias estimation” (this bias could be because offenses are generally underreported or because some types of offenses are more reported than others). Data obtained from experiments can overcome most of these issues. In particular, this paper uses experimental results obtained from Camacho and Requate (2012). Experimental data gives us the opportunity to accurately measure and control the conditions subjects deal with during the experiment. On one side, the experiment controlled for both probability of being caught and the amount fined using different treatments so the expected penalty was the same. On the other side, the experimental data obtained allows us to take into account the risk aversion distribution for all the individuals. It is crucial to take into account the risk preferences of the individuals (firms) because their decision on abatement will be directly linked to it. The classic model used in crime literature is the one of Becker (1968), who considers criminals as any other individual so they behave as utility maximizers. Using his economic model of crime, he derives the different roles individuals have. This is: riskneutral individuals will be not affected by changes in the composition of the expected penalty (neither probability nor size of the fine), risk-averse individuals will comply more with an increase in the size of the fine than with an increase in the probability of being caught, and risk-seeking individuals are deterred more by an increase in the probability of punishment than an increase in the severity of punishment This paper tries to test the classic Becker´s model of crime for the individuals’ behaviour facing different distributions of the expected fine (the probability of punishment and the severity of punishment) depending on their aversion to risk. In this context, it would be easy to see how individuals react to changes in the probability and the size of the fine. To test this, data from the experiment mentioned above concerning non-point pollution problems is used. The main results obtained from this paper do not seem to fit the Becker´s model perfectly. Most of the predictions from the model regarding risk attitudes and changes in the composition of the expected fine are poorly linked with the result from data and in some cases the effects seem to be completely the opposite.