Estimating deterrence effects: a public choice perspective on the economics of crime literature.

• Author: Benson, Bruce L.

1. Introduction

   Following Becker's path-breaking theoretical work on the economics of crime, a large econometric literature has developed to test his implications [3]. This literature typically focuses on a supply of offenses in which per capita crimes are related to the probability and severity of punishment for that type of crime, the expected income from the criminal activity, returns from alternative legal activities, and other socio-economic factors. However, the economic theory of crime also implies that the supply of offenses is determined simultaneously with community's demand for enforcement services [24; 25]. Theoretically, the crime rate depends on police deterrence efforts, police deterrence efforts depend upon the level of police resources, and the level of police resources depends on the crime rate. As a result, the "typical" econometric model employees data on reported crimes (Index I crimes) and a simultaneous equation estimating technique with at least two, and generally three (or more) equations in which police deterrence efforts are proxied by the probability of arrest [14].(1) A series of recent empirical studies of the economics of crime have broken with the simultaneous equation estimating technique, however. In particular, Layson [33] and Trumbull [52] both tested for exogeneity using Hausman's [31] test, confirmed it, and used OLS estimators, and Cover and Thistle [22] simply accepted Layson's findings and used OLS.(2) The purpose of the following presentation is to provide a theoretical explanation for why the econometrics of crime literature has not had sufficient data to adequately test for the deterrence hypothesis in a simultaneous equation model: that is, we explain why exogeneity was confirmed by Layson [33] and Trumbull [52].

   Testing for exogeneity could be motivated by the poor performance of the simultaneous estimation techniques in the economics of crime literature which generally finds that the crime rate is responsive to incentives and that higher crime rates are associated with more police resources, but that the probability of arrest often does not appear to be negatively and significantly related to the level of police resources [13; 14; 15; 17; 24; 25; 26; 35; 43; 53].(3) However, we contend that the poor results usually arising with the probability of arrest equation reflects a theoretical issue which suggests that these models are misspecified because of the unavailability of data on the actual allocation of police resources.(4) The economics of crime literature has generally ignored the incentives of the police bureaucrats who have considerable discretion in allocating their resources among a wide array of activities. Consideration of such incentives make it clear that an increase in aggregate police resources does not necessarily lead to an increase in the probability of arrest for any of the reported crime categories. These incentives are examined in section II in order to illustrate that the assumed empirical relationship between the aggregate level of police resources and the probability of arrest is not actually likely to hold. Section III contains concluding remarks.

2. Police Bureaucracies and Their Incentives

   Migue and Belanger [38] and Niskanen [41] proposed that bureaucratic managers seek discretion reflected by a budget with excess revenues over actual costs. These excess revenues are referred to as a "discretionary budget," "discretionary profit," "fiscal residuum," or "organizational slack" in the subsequent literature. The bureau manager cannot pursue discretion without constraint, of course, due to the monitoring and other controls imposed by the bureau's sponsor (e.g., legislature or city council overseeing the agency) [10; 41]. Uncertainty prevents perfect monitoring, however. This uncertainty arises at least in pan because for many bureaucracies output is not measurable and it cannot be objectively evaluated as in private sector markets. Furthermore, many bureaus have a number of different outputs, some of which are easily measurable, and some of which are not. When this is the case, a bureaucrat has incentives to produce the measurable output in quantities that correspond to the monitor's desires, while exploiting the uncertainty associated with unmeasurable outputs in order to gain discretionary budget [34].

   Bureaucrats also have incentives to "educate" the sponsor regarding interest group demands which complement their own goals and to "propagate" their own agenda [4; 18]. In fact, given that they successfully obtain a discretionary budget, they can use part of it to cover lobbying costs [4]. Therefore, they should be relatively influential in the political arena. Furthermore, law enforcement bureaucrats are frequently called upon to provide "evidence" because of their "expertise" on crime matters, so they tend to be a primary source of the information for decision makers and other potential interest groups [28, 22]. With their acknowledged expertise, bureaucrats are obviously a source of information for sponsors, but they might be a source of misinformation as well. Thus, bureaucrats' power depends on the degree of uncertainty, which they themselves can expand by misinforming, or not informing at all on some issues, so the legislative monitor faces the dual problem of determining both what the bureau's output should be and how it should be produced, with the potential for self-interested bureaucrats misleading them on both counts.

   The function of police in the minds of most citizens is to "fight crime." But how can we tell if they are doing a good job? The number of arrests is a natural measure of "effectiveness," as are response times, and these, along with reported levels of crime, tend to be the primary measures that police focus on in their lobbying efforts for expanded budgets [47, 156]. This in turn means that public police have incentives to produce fast response times and large arrest statistics relative to reported crime rates. Thus, incentives to watch or patrol in order to prevent crimes are relatively weak, and incentives to wait until crimes are committed in order to respond and make arrests are relatively strong. It is not surprising, therefore, that after an extensive study of police performance, Sherman [47, 149] concluded: "Instead of watching to prevent crime, motorized police patrol [is] a process of merely waiting to respond to crime." Indeed, Sherman's review of studies of police on duty concluded that about half of an officer's time is spent simply waiting for something to...