game1 crime

J Quant Criminol (2007) 23:243-258 DOI 10.1 007/s 1 0940-007 -9028-0

Community Variation in Crime Clearance: A Multilevel Analysis with Comments on Assessing Police Performance

Paul-Philippe Paré • Richard B. Felson • Marc Ouimet

Published online: 7 June 2007

© Springer Science+Business Media, LLC 2007

Abstract We use data from the Province of Quebec to examine how the characteristics of a crime and the community context in which it occurs affect the likelihood that it will be cleared by the police. Based on a sample of 362,295 crime incidents clustered in 93 communities, multilevel analyses reveal that the police are more likely to clear crimes in small communities than in large urban areas and in communities with greater levels of poverty. Workload is not very important, having only a slight effect on the clearance of misdemeanors. The fact that offenders are much more likely to evade the law in some communities than others may have important implications for deterrence. Two methods to improve the evaluation of police departments at crime clearance are also proposed.

Keywords Community size • Crime clearance • Police • Poverty • Workload

Introduction

A core function of the police is criminal investigation and the identification and appre- hension of offenders (American Bar Association 1980; Burton et al. 1993; Greenwood et al. 1977; S.P.C.U.M. 1996). Given the importance of the certainty of punishment for deterrence (see for example Blumstein et al. 1978; Cusson 1998; Nagin 1998), a police

P.-P. Paré (Kl) Department of Sociology, University of Western Ontario, London, ON, Canada N6A 5C2 e-mail: [email protected]

R. B. Felson

Crime, Law, and Justice Program, Department of Sociology, Pennsylvania State University, University Park, PA, USA

M. Ouimet

School of Criminology, Université de Montréal, Montreal, QC, Canada

Ô Springer

ORIGINAL PAPER

This content downloaded from 35.8.11.3 on Mon, 15 Apr 2019 12:56:26 UTC All use subject to https://about.jstor.org/terms

244 J Quant Criminol (2007) 23:243-258

department's success at these tasks is important. If offenders can avoid prosecution be- cause of ineffective law enforcement they may be more likely to commit crime. A standard measure of success at law enforcement is the crime clearance rate (Alpert

and Moore 1993). The clearance rate typically refers to the number of criminal incidents in which at least one suspect is charged and/or arrested for the offense, divided by the total number of incidents known to the police. In this research we use data from the Province of Quebec to examine how the types of crimes the police handle and the community context in which these crimes occur affect their success at clearing crimes. In our multilevel analyses, we isolate the effects of types of crimes from those of community characteristics. Our community variables include the level of poverty, the size of the community, and the crime workload, i.e., the number of crimes in the community relative to the number of police. Before presenting our hypotheses and reviewing the empirical literature, we discuss measurement issues related to crime clearance.

Measuring crime clearance

Some scholars have questioned the use of crime clearance rates as an indicator of police performance (e.g. Bay ley 1993; Brodeur 1998; Hoover 1996; Marx 1976; Murphy 1985).,They point to variation in recording practices across police departments and dif- ferences in the definition of what is considered a cleared incident (Brodeur 1998; Burrows

1986; Murphy 1985; Petersilia et al. 1990; Reiner 1998). For example, an incident may be defined as cleared if a suspect is charged, if a suspect is arrested even if there is no charge, or if a likely suspect is identified, even if no arrest is made. In addition, prior research (Reiner 1998; Walker 1992) shows that some police departments manipulate their records to increase their clearance rate. Tactics include failure to record difficult cases, the use of

an overly broad definition of clearance, and the encouragement of suspects to confess a series of offenses in exchange for more lenient charges.

The use of clearance rates to evaluate the performance of police departments raises additional issues. The judgment of performance in any endeavor requires an understanding of variation in task difficulty. The task environments of police departments are likely to vary widely. For example, a police department could have a low clearance rate because it handles a greater proportion of difficult cases (see below). Thus, if one wishes to use clearance rates to judge a department's performance it is important to control for the difficulty of their task. On the other hand, if wants to know what factors affect clearance rates, these measures of task difficulty become the variables of interest.

Despite its limitations, crime clearance is still a commonly used measure of police performance (Ouimet and Pare 2003; Reiner 1998). Its use, in part, reflects the lack of alternative measures. Measures of citizens' satisfaction with the police or the crime rate also have their limitations. In this research, we address some of the limitations of crime

clearance by using a standardized measure and a modeling strategy that takes differences between types of crimes and contextual factors into account. The literature suggests a number of factors that affect crime clearance. We discuss each of them in turn.

1 Scholars have also been critical of an overemphasis on crime clearance in evaluation of the effectiveness of police function. They point out that the police have other important functions including order mainte- nance, problem solving, and social services (Bay ley 1993; Brodeur 1998; Hoover 1996; Marx 1976; Murphy 1985; Walker and Katz 2002).

â Springer

This content downloaded from 35.8.11.3 on Mon, 15 Apr 2019 12:56:26 UTC All use subject to https://about.jstor.org/terms

J Quant Criminol (2007) 23:243-258 245

Type of crime

Crime clearance rates are much higher for some types of crimes than others (see e.g., Ahlberg and Knutsson 1987; Cordner 1989; Ouimet and Pare 2003; Riedel and Jarvis 1998). For example, according to FBI (2000) data, the clearance rate is 69% for murder, 59% for aggravated assaults, 14% for burglaries, and 15% for auto thefts. Violent crimes are more likely to be cleared than property crimes, presumably because they involve face- to-face confrontation between the offender and the victim, and the victim serves as a

witness. For some crimes the offender is typically identified when the crime is detected, so almost every crime encountered is cleared. Examples include drug possession, driving while intoxicated, and being an obstacle to a peace officer on duty.

The evidence is clear in showing that the types of crime the police handle are a major predictor of a police department's clearance rate. For example, Cordner (1989) found that clearance rates are lower for departments that handle higher proportions of property crimes, since these crimes are more difficult to clear. Ouimet and Pare (2003) also found that clearance rates of police departments are strongly related to the types of crime they handle. Finally, Riedel and Jarvis (1998) found that clearance rates for homicides are strongly influenced by the relative proportion of easy cases (e.g., spousal homicides) and difficult cases (e.g., homicides related to the organized crime).

Workload

The more offenses the police handle, the less time and resources are available for a particular case. Thus, many studies show that heavy crime workloads are associated with lower clearance rates (Bay ley 1994; Borg and Desbiens 1997; Greenwood et al. 1977; McClintock and Avison 1968; Ouimet and Pare 2003; Parker 2001; Sullivan 1985). Litwin (2004), however, found no association between homicide clearance and the homicide rate (see also Lattimore et al. 1997). Workload may also have different effects on clearance for different types of crime. For example, it may be that when the workload is heavy, the police concentrate their efforts on more serious crimes and ignore minor crimes. It is possible that efforts to combat violent crimes by hiring more police only affect the clearance of minor or nonviolent crimes.

Workload effects have an important policy implication. They suggest that more crimes will be cleared if the police force is expanded. The hiring of more police is a common crime-fighting policy but results are mixed regarding its effects on crime rates, and the evidence leans toward supporting the null hypothesis (Cameron 1988; Eck and Maguire 2000; Levitt 1997, 2002; Loftin and McDowall 1982; Marvell and Moody 1996; McCrary 2002).

Community size

Crime clearance may be more difficult in large urban areas than in areas that are less heavily populated. Urban areas provide more anonymity to offenders (Willmer 1970). Witnesses are less likely to be able to identify offenders and the police may be less likely to know offenders and criminal networks (Felson 1998). The evidence on the relationship between community size and crime clearance is mixed, however. Some studies have found that size has a positive effect (e.g., Greenwood et al. 1977), some have found a negative

Ô Springer

This content downloaded from 35.8.11.3 on Mon, 15 Apr 2019 12:56:26 UTC All use subject to https://about.jstor.org/terms

246 J Quant Criminol (2007) 23:243-258

effect (Ahlberg and Knutsson 1987; Cordner 1989; FBI 2004; Ouimet and Pare 2003), and some have found no effect (Borg and Parker 2001; Litwin 2004). Thus, it is not clear whether offenders can more easily evade the law in big cities or small towns.

Poverty

According to a conflict perspective (Black 1976; Chamblis and Seidman 1982; Turk 1969), the criminal justice system discriminates against the poor and minorities because these groups have lower power and status. The argument takes two forms, each implying opposite predictions. According to one argument the criminal justice system discriminates against low status offenders by paying more attention to their offenses and treating them more severely. For example, the police may patrol poor communities more closely than wealthier communities, enforce the law more vehemently, and show less respect for the civil liberties of suspects (e.g. Sampson 1986; Smith 1986). From this perspective, one would expect clearance rates to be higher in impoverished areas.

An alternative argument is that the criminal justice system discriminates against low status victims by taking their grievances less seriously. Thus, Donald Black (1976) and others (e.g. Gross and Mauro 1989; Hawkins 1987; LaFree 1989) have argued that the law is less available to victims from low status groups. The police may ignore crime in poor communities and devote less effort to apprehending offenders. Identifying suspects may also be more difficult in impoverished communities if citizens are less cooperative with the police (e.g. Riedel and Jarvis 1998). The literature on collective efficacy and social dis- organization also suggests that impoverished communities are less successful in fighting crime (e.g. Bursik 1988; Sampson et al. 1997). Citizens in these communities may be less successful than citizens in wealthier communities at pressuring the police to address a crime problem. From this perspective, crime clearance should be lower in impoverished communities.

The evidence on the relationship between poverty and crime clearance is mixed. Some studies found that clearance rates are higher in impoverished areas (Ouimet and Pare 2003) or areas with a large number of minorities (Borg and Parker 2001). Other studies have found lower crime clearance in poor communities (e.g., Sullivan 1985). Litwin (2004) found mixed evidence in his study of homicide clearance rates. Clearance rates were positively related to home ownership, but unrelated to unemployment and the median level of income in an area.

In sum, prior studies support the ideas that crime clearance is higher in some com- munities than others because of the types of crime the police handle and because of their workload. On the other hand, the evidence is mixed regarding the effects of community size and the prevalence of poverty.

The current study

In this study we examine the effects of workload, community size, poverty, and types of crimes on the likelihood that crimes will cleared by the police. We use a standardized measure of crime clearance that addresses, at least to some degree, the problem of recording variation across departments.

Unlike previous research, which is typically based on aggregate level data, we use a multilevel design. Our equations include a set of dummy variables representing each type

Springer

This content downloaded from 35.8.11.3 on Mon, 15 Apr 2019 12:56:26 UTC All use subject to https://about.jstor.org/terms

J Quant Criminol (2007) 23:243-258 247

of crime, thereby providing a better control for types of crimes than previous research. Most studies of crime clearance either control for whether an offense involves violent or

property crime, or only examine one type of crime (e.g. homicide). We also estimate separate equations for violent crimes, property crimes, and misdemeanors. Based on the literature reviewed above, we predict that police departments with a

heavier workload should be less likely to clear crimes. We predict lower crime clearance in large communities because they provide offenders greater anonymity. Finally, it is not clear whether to predict higher or lower crime clearance in impoverished areas. Dis- crimination against offenders of low status would lead to higher clearance in impoverished areas while discrimination against victims of lower status would have the opposite effect. It is also possible that these forms of discrimination cancel each other out.2 Another goal of our study is to propose two practical methods for assessing the per-

formance of police departments at clearing crimes. The first method utilizes the residuals from a multilevel analysis to determine which police departments are particularly effective or ineffective at clearing crimes. While this first method works well, it is computationally demanding and requires knowledge of multilevel modeling. A second method based on standardization for crime types followed by OLS regression including community vari- ables is also proposed. This alternative method is not as statistically elegant as the first one, but some analysts will find it simpler to use, especially with large datasets. Both methods follow a very similar logic. We believe that our methods are an improvement over raw comparisons of clearance rates because they control for differences in the difficulty of the task that departments face. Specifically, they control for differences in types of crimes, workload, community size, and level of poverty.

Methods

Our data are based on the Canadian Uniform Crime Report (UCR2) from the Province of Quebec in 1998. The UCR2 provides information about every incident. Thomassin (2000) reports that 95% of criminal incidents registered by the police were included in the UCR2 in 1997. Our second data source is an official report by Statistics Canada (1997) that provides data related to police organizations, such as the number of sworn police officers and the number of crimes per police officer. The third data source is the 1996 Census for the Province of Quebec, which provides data on population size and poverty in Quebec's municipalities.

Our data set includes a total of 362,295 incidents clustered in 93 municipal police departments. We omitted departments in rural territories and municipal departments with fewer than 10 police officers.3 Many of the police departments studied in our analyses no longer exist because of changes in Quebec's municipalities and organizational changes during the last few years. Some departments have been merged into regional police forces while others have been integrated into the provincial police.

2 We did not include a measure of racial composition of communities in our study. In the Province of Quebec, many communities have a very low proportion of minorities, and minorities tend to be socially well integrated.

3 We omitted smaller and rural communities from analysis because policing duties are often handled by the provincial police rather than the local force. Also, we were concerned that the measurement of community characteristics may be less valid in these areas because of small samples.

Ê Springer

This content downloaded from 35.8.11.3 on Mon, 15 Apr 2019 12:56:26 UTC All use subject to https://about.jstor.org/terms