Impersonal Policing vs. Community Policing

RESEARCH ARTICLE

The impact of community policing on crime rates: does the effect of community policing differ in large and small law enforcement agencies?

Mehmet Alper Sozera* and Alida V. Merlob

aSchool of Security Sciences, Turkish National Police Academy, Ankara, Turkey; bDepartment of Criminology, Indiana University of Pennsylvania, Indiana, PA, USA

This study utilized three major data-sets consisting of Law Enforcement Management and Administrative Statistics 2003, two waves of Uniform Crime Report data, and the US Census 2000 data to examine whether the relationship between community polic- ing and crime differs based on the agency size. To operationalize community policing, the authors employed factor analyses. The results yielded three dimensions for the construct of community policing. The multiple-regression analyses indicate that all dimensions of community policing have a significant relationship to crime rates in small agencies; whereas, in large agencies, only problem-solving partnership has a significant positive association with property and violent crime rates. The level of implementation of problem-solving partnership dimension has a positive effect on both small and large agencies, but this effect is more prominent for small agencies. These findings along with possible policy implications are discussed.

Keywords: community policing; crime rates; law enforcement agencies; aggregate- level study

Introduction1

Lab and Das (2003) contend that community policing influenced policing strategies throughout the world. In fact, it has been the most popular trend in policing during the past few decades (Palacios, 2003; Skogan, 2006). Although scholars and professionals have not reached consensus on a single definition for community policing, many pro- grams under the rubric of community policing have surfaced since the 1970s. Initially, community policing was adapted by a few large agencies in major American cities in the form of a pilot study. In general, studies during that time mostly yielded positive results. Those positive results included a reduced fear of crime, improved public–police relation- ships, and the prevention of crime (Police Foundation, 1981; Schneider, 1978; Tien & Cahn, 1986).

In the span of a few years, the positive results revealed in data from sample sites began to attract the attention of politicians. With federal support through the Office of Community Oriented Policing Services (COPS Office), community policing began to be adopted in small and rural departments as well. The percentage of departments employing community policing personnel increased from 34% in 1997 to 64% in 1999 (Hickman &

*Corresponding author. Email: [email protected]

Police Practice and Research, 2013 Vol. 14, No. 6, 506–521, http://dx.doi.org/10.1080/15614263.2012.661151

� 2012 Taylor & Francis

Reaves, 2003). By the end of the fiscal year 2004, COPS Office had funded more than 118,768 community policing officers and deputies, and the total investment of COPS reached $11.3 billion (COPS Office, 2007).

Does allocating a large amount of money to law enforcement agencies to support the implementation of community policing affect the incidence of crime on a national level? According to the Uniform Crime Report (UCR), crime has been consistently declining since 1993 (Federal Bureau of Investigation [FBI], 2006). Scholars have suggested that community policing might have a general role in the last decade’s crime drop (Zhao, Scheider, & Thurman, 2002; Zhao & Thurman, 2004). Contrastingly, some scholars argued that it is difficult to claim that this reduction was attributable to community policing since many other positive social factors such as an improved economy and high employment that might influence crime were evident during the 1990s (Eck & Maguire, 2000).

The previous studies

A single satisfactory answer about the impact of community policing on crime is not readily available as studies have demonstrated inconclusive results (see, e.g. Cordner, 1988, Government Accountability Office [GAO], 2003, 2005; MacDonald, 2002; Mastrofski, 2006; Police Foundation, 1981; Skogan, 2006; Weisburd & Eck, 2004; Zhao, Scheider, & Thurman, 2003). Generally, community policing without a clear focus on specific problems has not been found to be effective in preventing crime. Similarly, it was found that foot patrols, storefront offices, newsletters, and community meetings do not reduce crime (Weisburd & Eck, 2004). Yet, research suggests that these strategies have been somewhat effective in reducing fear of crime and increasing citizen satisfaction with police services (Braga & Weisburd, 2006).

There are a few explanations for the contradictory results in previous studies. First, community policing studies have always suffered from validity issues such as the prob- lem of the definition and the construct (Lurigio & Rosenbaum, 1986). Some programs may have been identified as community policing when they really did not include the core elements of community policing philosophy (Greene & Mastrofski, 1988). Second, for more than a decade, the USA witnessed a decrease in crime, coinciding with the same time period when economic parameters were getting better and harsher penalties had been established. Simultaneously, community policing increasingly was being adopted through- out the nation. Yet, an assessment of which factor has contributed to this crime reduction the most remained inconclusive (Eck & Maguire, 2000).

Early studies about the effectiveness of community policing were limited to large cit- ies (see, e.g. Bowers & Hirsch, 1987; Kessler & Duncan, 1996; Schneider, 1978; Tien & Cahn, 1986; Uchida, Forst, & Annan, 1992; Wycoff & Skogan, 1993). These studies examined the effects of community policing by using a pretest/posttest or a single post- test with or without comparison groups (Lurigio & Rosenbaum, 1986; Yin, 1986). According to Kessler and Duncan (1996), the majority of these studies suffered because of weak research designs. They underused statistical significance tests and measured con- cepts poorly. These studies lacked a valid and reliable measure of program implementa- tion and outcomes, and they consistently failed to address competing explanations for observed effects (Eck & Maguire, 2000; Levitt, 1997; Lurigio & Rosenbaum, 1986).

After a significant federal investment in community policing, national studies began to appear in the literature. Within Oliver’s (2000, p. 373) definition of the ‘evolution of community policing,’ this era represented the institutionalization of community policing. Aggregate-level studies conducted in the institutionalization era revealed inconsistent

Police Practice and Research: An International Journal 507

results. Zhao et al. (2002), Zhao and Thurman (2004) and the GAO (2005) concluded that COPS’ funds reduced crime nationwide. Similarly, Evans and Owens (2007) found that more police added to agencies through the use of COPS grants yielded a reduction in auto thefts, burglaries, robberies, and aggravated assault.

By contrast, when Worrall and Kovandzic (2007) revisited the effect of COPS grants on crime, they found that COPS grants had little or no effect on crime. The focus of these studies was not to assess the extent of implementation of community policing but rather to examine whether COPS’ funding has an impact on crime. Due to the nature of their studies, the operationalization of community policing was not the main concern.

Two studies (Beckman, 2006; MacDonald, 2002) presented some evidence of level of implementation of community policing while analyzing its impact on crime rates, but they overlooked small departments. Excluding small departments is problematic in under- standing the impact of level of implementation of community policing on crime. As Wells and Weisheit (2004, p. 20) noted:

If there are fundamental rural/urban differences in the process by which crime is generated, then focusing almost exclusively on urban areas amounts to little more than convenience sampling in which important sources of variation are omitted. The fact that rural areas can be difficult to study is not a justification for excluding them from research.

Large vs. small departments

Community policing was the impetus for the changes that took place in American polic- ing. It has long been established that police alone cannot solve crime problems. Commu- nity support is necessary to overcome area-specific problems, and it is essential in every department regardless of its size (Benedict, Bower, Brown, & Cuningham, 1999). None- theless, aggregate-level studies, which examine the effect of community policing on crime rates, tend to ignore small departments.

Studies that overlook small agencies may mislead researchers in two ways. First, they will miss the natural laboratory of community policing as these officers generally know the service recipients personally, have face-to-face contact, and are involved in many activities that differ from traditional policing (Weisheit, Wells, & Falcone, 1994). Second, although crime rates are more likely to be lower in small jurisdictions, they are not evenly distributed across all small jurisdictions. Therefore, a conclusion that indicates crime rates and social factors do not vary across small agencies is likely to be inaccurate.

Social disorganization theory is applied to urban communities and largely ignored in rural and suburban settings (Osgood & Chambers, 2000). A study conducted by Wells and Weisheit (2004) found that the key contextual variables varied drastically even in the most rural areas. Furthermore, the magnitude and the direction of impact of contextual variables on crime also differed in urban and rural settings. For example, one of the possible explanations of the reduction in crime during the 1990s was attrib- uted to economic growth (Blumstein & Wallman, 2000). By contrast, it was observed that economic growth was associated with an increase in crime rates in rural settings (Wells & Weisheit, 2004).

In fact, there are also some differences between large and small departments’ police officers. Officers in large departments are less respected and supported by citizens than officers in small departments; and officers in large departments are less responsive to the community needs. According to Weisheit et al. (1994), officers in large departments are more responsive to their departments’ needs and dynamics. In that sense, small depart- ments are deemed to experience the more positive effects of community policing.

508 M.A. Sozer and A.V. Merlo

Differences aside, similarities between large and small departments also support the inclusion of small departments in the analysis. A city-specific study in Choteau, Okla- homa with a population of 1500 residents revealed parallel findings with its large city counterparts. After implementation of community policing in the area, citizens’ percep- tions of the police department, the quality of police services, safety, and crime improved significantly (Brand & Birzer, 2003). These data suggest that community policing does matter even in a very small rural area where researchers do not expect substantial varia- tion in crime rates traditionally.

Another similarity between small and large departments is the focus on public safety. For example, 84% of 207 small departments in one study ranked property crimes at the top of the list just as their larger counterparts did, and ranked violent crime against indi- viduals fifth among 17 types of crime on the list (GAO, 1995). If the goal is to prevent and control crime by establishing quality relationships between the community and police, community policing in small departments is as popular as it is in large depart- ments (Zhao & Thurman, 2003). Between 1998 and 2000, smaller departments signifi- cantly increased their community policing activities (Roth, Roehl, & Johnson, 2004). Furthermore, COPS Office grants entitled ‘Funding Accelerated for Smaller Towns’ (FAST), allocated resources especially for departments serving a population of fewer than 50,000. Additionally, the creation of such funds encouraged small departments to engage in community policing. In particular, jurisdictions with more reported crime were more likely to apply for this grant (GAO, 1995).

The mere existence or implementation of community policing does not guarantee that the officers will assimilate its goals or philosophy. One study found that even though more than half of the agencies in the study sample had a community policing program, few of them experienced internal and external organization change (Hawkins & Weisheit, 2003, p. 26). According to Wang (2006, p. 69), ‘In community policing, CPO sergeants and lieutenants – the first line commanders – are key figures because they play the role of brokers in the process of change.’ When police departments adopt community polic- ing, the supervisors and administrators have the opportunity to employ management styles different than ‘… the universal “by the book” one’ (Wang, 2006, p. 75). Although scholars posit that small departments embody community policing, it must not be assumed that they will engage systematically in community policing activities.

The present study

Different from most of the previous aggregate-level studies on the effect of community policing on crime rates, this study has two main concerns. First, does the level of imple- mentation of community policing differ in its impact on crime rates? Second, does the agency size matter for the effect of level of implementation of community policing on crime rates? Previous studies have failed to examine the impact of community policing by simultaneously taking the level of implementation and various agency sizes into con- sideration. With this limitation as a foundation, this study examines the relationship between the level of implementation of community policing and crime rates by consider- ing variations across agency size.

First, this study conducted analyses on a combined data-set that consists of two national data-sets (UCR and Census) along with data collected at the national level through a sample survey of Law Enforcement Management and Administrative Statistics (LEMAS).2 Agencies serving populations fewer than 50,000 and over 50,000 were sepa- rately analyzed in order to see if there are differences between these two populations.3

Police Practice and Research: An International Journal 509

Data and sample

The LEMAS 2003, the UCR data for the years 2004 and 2005, and Census 2000 data were merged for the analyses. This combined data-set consisted of 2402 agencies. The variables used in this study were classified into three main groups: the dependent variables, the explanatory variables, and the control variables. The control variables were also classified into two groups as departmental and contextual (structural-level) control variables.

The dependent variables

The FBI’s classifications for property and violent crimes were used to generate property and violent crime rates as they are the best measure of reported crime available at the national level (Zawitz et al., 1993). The two-year (2004–2005) average rates were utilized to stabilize the variations in crime rates. To control for the population size, crime per 1000 residents was used.

The independent variables

Explanatory variables

Community policing variables were generated from LEMAS data. Even though commu- nity policing variables are, at best, a weak indicator of community policing, they are the only available national-level data on this issue. Given the limitation of the LEMAS data, it is difficult to claim that the level of implementation of community policing was pre- cisely captured in this study. However, this study measured community policing more thoroughly than most of the previous studies.

Factor analysis was utilized to reduce 35 dichotomous variables to a few components. Results of factor analyses, using varimax rotation, suggested three factors (depicted in Table 1). These factors were used as indicators of the level of implementation of commu- nity policing in agencies. Items having factor loadings above .50 were included in the related measure of community policing.4

It is important to have conceptual evidence to rule out the threat that suggests compo- nents might just be mathematically formed. To name each factor, the authors reviewed the previous literature on the definition and dimensions of community policing.

Three factors, each representing a different dimension, were used to measure level of implementation of community policing in an agency. The first factor relates to collecting systematic information from the public, and the use of that information for various departmental activities. Conversely, the second factor relates to problem-solving activity and community policing training. The last factor represents the commitment of a depart- ment in problem-solving partnership. Factors were named community contribution, train- ing and problem-solving, and problem-solving partnership, respectively.5 Factor scores indicating the level of implementation of a specific dimension of community policing were utilized in the regression analyses.

Control variables

There are two types of control variables used in this study: departmental and contextual. The first departmental control variable is ‘police size’ (the number of police officers per 1000 residents). According to Marvell and Moody (1996), an increase in number of

510 M.A. Sozer and A.V. Merlo

police officers on the street can prevent crime through deterrence and incapacitation. Criminals who observe more officers in the area are less likely to commit a crime due to the high risk of apprehension.

The second departmental control variable is the ‘percentage of community policing officers.’ The implementation of community policing might be based solely on personnel specifically assigned to community policing activities. To create this variable, the number of community policing officers was divided by the number of all actual full-time sworn personnel with general arrest power and multiplied by 100.

The last departmental control variable is the ‘education requirement of officers’ in the department. MacDonald (2002) argued that level of education can be used as a proxy for

Table 1. Factor structure of three-component model.

Components and factor loadings

Items 1 2 3

Surveyed public satisfaction .812 .217 .040 Surveyed public perception .807 .199 .079 Surveyed personal crime experiences .749 .162 .097 Surveyed reporting of crimes .756 .123 .119 Used info for allocating resources .747 .162 .125 Used info for evaluating agency performance .825 .215 .065 Used info for evaluating officer performance .701 .116 .073 Used info for evaluating program effectiveness .774 .171 .120 Used info for prioritizing problems .766 .122 .151 Used info for providing info to patrol officers .811 .151 .101 Used info for redistricting areas .497 .100 .100 Used info for training and development .715 .189 .103 Encouraged SARA-type projects .180 .603 .211 Conducted citizen police academy .204 .511 .168 Creating community policing plan .203 .508 .252 Assigned officers to a specific geographic areas/beats .114 .519 .228 Problem-solving in officer evaluation criteria .171 .544 .194 Training citizens in community-oriented policing (COP) .151 .560 .300 Upgraded technology .129 .526 .217 Partnered with citizen groups .167 .567 .341 Mission statement included COP .127 .534 .149 Training new officers in COP .089 .687 .073 Training (in service) sworn officers in COP .052 .643 .044 Training civilian personnel in COP .118 .544 .055 Partnership w/advocacy groups .098 .130 .649 Partnership w/business group .157 .284 .677 Partnership w/religious groups .149 .193 .644 Partnership w/local government .084 .213 .689 Partnership w/other local law enforcement agencies .087 .127 .602 Partnership w/neighborhood associations .154 .338 .634 Partnership w/senior citizen groups .148 .141 .680 Partnership w/school groups .093 .210 .653 Partnership w/youth service organizations .140 .188 .699 Surveyed other .146 −.007 .114 Used info for other .190 .057 .077 Percent common variance 30.062 11.426 5.385 Eigen values 10.52 4.00 1.58

aBased on varimax rotation.

Police Practice and Research: An International Journal 511

effective policing because officers with a higher level of education are more likely to carry out effective crime control policies. In order to create this variable, the ordinal-level original variable (minimum education requirement of new recruit officers ranging from 1 to 5)6 in the LEMAS 2003 survey was recoded as an indicator variable.

Some structural factors such as poverty, racial diversity, densely populated neighbor- hood, and disrupted family structure are well-known correlates of crime (Lafree, 1998; Messner, 1982; Sampson, 1987; Shaw & McKay, 1942). Cities with higher percentages of divorced males, single-parent households, and unsupervised teenage peer groups have higher rates of crime and delinquency (Sampson, 1987; Sampson & Groves, 1989). Such indicators of social disorganization are most prominent in African-American, inner-city neighborhoods (Sampson & Wilson, 1995). These factors are important in terms of their effect on variation in crime rates across jurisdictions (Kornhauser, 1978). The variables were compiled from the Census 2000 data and each one was used as percentages.

Data analysis

Ordinary least square (OLS) regression was utilized as the primary statistical technique for the study. OLS regression assumptions such as multicollinearity, homoskedasticity, and linearity were tested before performing the actual analyses because significance lev- els, confidence intervals, and other tests are sensitive to a violation of these assumptions7

(Norusis, 2000). Two models were created. In the first model, the influence of predictor variables on

each type of crime rate was examined for only small agencies.8 Then, the same examina- tion was conducted for large agencies.

ŷsmall agencies=large agenciesðproperty=violent crimeper 1000 residentsÞ

¼ a þ b1 police size þ b2 community policing officers þ b3 education requirement for officers þ b4 urban population þ b5 single female headed household w=children under 17 þ b6 renters þ b7 population between the age of 18–24 þ b8 people living below the poverty level þ b9 divorced males; þb10 African Americans þ b11 community contribution þ b12 training problem-solving þ b13 problem-solving partnership þ e:

Finally, the property and violent crime models were further analyzed in terms of the differences in effects. Paternoster, Brame, Mazerolle, and Piquero (1998, p. 862) argued that when the researcher’s aim is to see the difference between two regression coefficients with large samples, the z-test formula illustrated below was to be used.

z ¼ b1 � b2ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ðse1Þ2 þ ðse2Þ2

q :

Results

Table 2 shows the general characteristics of the variables. On average, small agencies have more officers per 1000 residents. Approximately, 1949 (81.1%) of the agencies do not require a college degree or any college credits for new police officer hires. In 2003, both large and small agencies did not place an emphasis on the cadets’ educational degrees. In fact, the percentage of community policing officers is actually higher in small

512 M.A. Sozer and A.V. Merlo

agencies, contrary to the belief that community policing is for big city police departments. Although percentage of community policing officers is higher in small agencies, the level of implementation of community policing on all three dimensions is much lower on average.

Contextual factors of each jurisdiction are likely to influence crime rates. It is argued that structural characteristics in big cities are heterogeneous in nature. However, descrip- tions of these variables illustrate that small cities also have substantial variability. An interesting finding is that the mean of divorced males is the same for both large and small agencies. Not surprisingly, the biggest difference in the mean is for the variable ‘the percentage of urbanized area.’ As expected, large jurisdictions have a higher percentage of urbanized areas than small jurisdictions.

Table 3 depicts the tests of the hypotheses which address whether there is an associa- tion between the level of implementation of community policing and property crime rates, and whether there is a difference between large and small agencies in terms of the effect of community policing variables on property crime rates. The significant positive effects of ‘police size,’ ‘urbanized area,’ ‘population between 18 and 24 years of age,’ and ‘divorced males’ are identified for both large and small agencies as previous studies suggested (Heimer & Coster 1999; Markowitz & Felson, 1998). An increase in any of these variables is likely to result in an increase in property crime rates. When it comes to difference in effect, their effects are more pronounced for large agencies.

The effect of the percentage of the population between 18 and 24 years of age is also moderated by agency size; however, the direction of its association with property crime

Table 2. Descriptive statistics of all variables.

Small Large

Variable Mn SD Mn SD

Departmental control variables Police size 2.50 2.04 1.56 .91 Cop officers 16.27 28.61 6.39 14.19 Education requirement .16 .37 .23 .42

Structural control variables Urbanized area 68.03 39.88 89.11 17.40 Single-female households 6.83 3.39 7.52 2.79 Renters 28.20 12.67 34.08 11.83 Population 18–24 years of age 8.95 5.32 10.52 4.20 Population below poverty 12.86 8.05 12.75 6.18 Divorced males 4.33 1.43 4.33 .98 African-American population 9.54 16.11 13.39 15.46

Explanatory variables Community contribution .080 .860 .138 1.190 Problem-solving partnership .180 .873 .310 1.121 Training and problem-solving .278 .859 .480 1.042

Dependent variables Property crime rate 28.86 26.35 31.47 24.78 Violent crime rate 12.05 13.09 13.43 12.02 N 1521 881

Notes: Mn depicts the item’s mean value. SD depicts the standard deviation.

Police Practice and Research: An International Journal 513

rates is different for large and small agencies. Large agencies which have a greater percentage of the population between 18 and 24 years of age are more likely to have higher property crime rates than small agencies.

With regard to the community policing variables, in small agencies all community policing variables are associated with higher crime rates, whereas, in large agencies, only problem-solving partnership has reached statistical significance. Problem-solving partner- ship is associated with higher property crime rates for both large and small departments; and this impact is more pronounced in small agencies.

Table 4 illustrates the interaction effects that are identified for police size, population between 18 and 24 years of age, and divorced males, which suggest that their effects dif- fer based on an agency’s size. Among these variables, ‘police size’ and ‘divorced males,’ have significant positive effects on violent crime rates for both large and small agencies. The effect of police size is more prominent for large agencies which may be because of the higher incidence of crime in big cities. Unlike the findings of property crime model, the direction of effect of the percentage of the population between 18 and 24 years of age is not different for large and small agencies. Large agencies with a greater percentage of the youth population tend to have lower violent crime rates than small agencies.

In terms of community policing variables, community contribution is not a significant predictor of violent crime rates. Training and problem-solving is only significant for small agencies, and it is positively associated with violent crime. Consistent with the property crime rate model, the problem-solving partnership dimension of community policing has a positive impact on the violent crime rate. However, unlike the property crime model, this effect is equal for both groups.

Study limitations

This study is similar to the published community policing research in its lack of a solid theoretical background and agreed-upon measures of community policing. Second, only

Table 3. Comparison model for property crime rates.

Variable Small agencies Large agencies Z score

Police size 4.559⁄⁄⁄ (.722) 13.952⁄⁄⁄ (1.099) 7.170a Cop officers −.026 (.020) .047 (.040) −.477 Education requirement 1.084 (1.172) 1.845 (1.296) .435 Urbanized area .175⁄⁄⁄ (.021) .335⁄⁄⁄ (.036) 3.902a Single-female households .875⁄ (.370) .303 (.423) 1.010 Renters .237⁄⁄ (.072) −.107⁄⁄ (.069) −1.313 18–24 years of age −.035 (.113) .581⁄⁄⁄ (.159) −2.8a Below poverty .217 (.121) .505⁄⁄ (.173) 1.364 Divorced males 2.523⁄⁄⁄ (.554) 6.513⁄⁄⁄ (.634) 4.744a African-American .032 (.059) −.063 (.064) −.356 Community contribution 2.212⁄⁄ (.759) .716 (.461) 1.684 Training and problem-solving 1.658⁄ (.660) −.068 (.494) −1.924 Problem-solving partnership 4.382⁄⁄⁄ (.734) 2.273⁄⁄⁄ (.558) 2.287a R2 .402 .613 Observations 1521 881

Notes: Unstandardized coefficients were depicted.Robust standard error in parentheses. ⁄p < .05; ⁄⁄p < .01; ⁄⁄⁄p < .001. aIllustrates significant difference.

514 M.A. Sozer and A.V. Merlo

community policing-related activities are included in the analyses, in part, because of the available data. Future studies might control for or examine the impact of other policing strategies which might be more strongly related to crime rates such as hot spots and directed patrols. Additionally, instead of using official crime rates which are subject to criticism, victimization data by itself or in conjunction with official crime data might illustrate a better picture of the relationship between community policing and crime.

Third, even though the presence of community policing implementation is quantified to some extent through three composite measures, it is still uncertain to what extent each of the community policing activities is implemented in an agency. This research, consis- tent with the literature, suggests that testing the effectiveness of community policing is more problematic at the aggregate level where there are many versions of the definition and implementation of community policing. A better strategy might be to examine these activities under the umbrella of community policing at the city level where the definition and implementation can be better identified.

Finally, because of data restrictions, law enforcement agencies are used as the unit of analysis. Besides, demographical variables were gathered at one point in time (Census 2000). Therefore, possible changes over time and their impact cannot be precisely cap- tured.

Discussion

Overall, large agency regression models do a better job than small agency regression models in predicting property and violent crime rates. The number of police per 1000 res- idents makes the strongest unique contribution to predicting crime rates for all of the models. The results of studies on police size and its relationship to crime are ambiguous in terms of both the direction of the relationship and the temporal ordering of the effects (Eck & Maguire, 2000; Levitt, 1997; Marvell & Moody, 1996). Two explanations seem plausible when interpreting why the impact of police size is the strongest, and why this impact is more prominent for large agencies.

Table 4. Comparison model for violent crime rates.

Variable Small agencies Large agencies Z score

Police size 1.336⁄⁄⁄ (.322) 6.717⁄⁄⁄ (.488) 9.214a Cop officers −.003 (.009) .025 (.019) 1.047 Education requirement −.413 (.583) .153 (.630) −.303 Urbanized area .072⁄⁄⁄ (.009) .079⁄⁄⁄ (.016) .388 Single-female households .849⁄⁄⁄ (.195) 1.236⁄⁄⁄ (.195) 1.407 Renters .059 (.031) −.135 (.032) −.627 18–24 years of age −.015 (.051) −.325⁄⁄⁄ (.066) 3.734a Below poverty .189⁄⁄ (.063) .147⁄ (.073) .066 Divorced males 1.403⁄⁄⁄ (.270) 2.531⁄⁄⁄ (.269) 2.960a African American .061 (.032) −.023 (.295) −.128 Community contribution .414 (.367) .174 (.221) .450 Training and problem-solving .940⁄ (.396) −.282 (.224) −1.449 Problem-solving partnership 1.895⁄⁄⁄ (.375) 1.184⁄⁄⁄ (.274) 1.532 R2 .47 .617 Observations 1521 881

Notes: Unstandardized coefficients were depicted.Robust standard error in parentheses. ⁄p < .05; ⁄⁄p < .01; ⁄⁄⁄p < .001. aIllustrates significant difference.

Police Practice and Research: An International Journal 515

First, more officers might increase the size of the beats to be patrolled as well as the frequency of the patrol sequence; hence, more officers might discover more crime. Second, as intensely discussed in previous studies, there is a simultaneity problem. In the police and crime equation, it is very difficult to sort out which one is cause and which one is effect. When it comes to the difference in the effect of police size on small and large agencies, the findings of Marvell and Moody (1996, p. 632) – which suggest that increasing the number of officers in big cities is more effective than additional officers in small cities because the amount of crime is greater in large cities – appear to have merit. In brief, their work suggests that large cities will get more ‘bang for their buck.’

Communities dominated by high residential mobility, population density, economic dis- advantage, and disrupted families are associated with higher crime rates (Sampson & Groves, 1989; Shaw & McKay, 1942). As Wells and Weisheit (2004) argued, this study found that social context is not homogeneous even in small communities. Consistent in all models, the percentage of urbanized area, divorced males, single female-headed households with children under 17, and people living below the poverty level are positively associated with crime rates (though not significant in all models). These findings are supported by pre- vious research. Young men, economically disadvantaged people, and urban residents are more likely to be associated with crime (Heimer & Coster 1999; Markowitz & Felson, 1998). However, the percentage of renters and the percentage of the population between 18 and 24 years of age yielded inconsistent results. The percentage of African-Americans never reached significance in any of the models. Moreover, some contextual factors have different effects across agencies and on the types of crime. The percentage of renters is the most versatile factor that has a different effect in terms of both direction and significance.

In the large agency models, the current study found that jurisdictions having a higher percentage of youth between 18 and 24 years of age are more likely to have higher prop- erty crime rates, but lower violent crime rates. There are two possible explanations why this variable is a significant factor in only large agencies. First, opportunities for crime might be greater while the likelihood of apprehension might be reduced in larger communities. Second, the effects of other variables such as the percentage of renters and African-Americans in the population might suppress the effect of this variable in the small agency models.

The first composite measure, community contribution is positively associated with an increase in property crime rates for small departments only. In terms of violent crime rates, this variable never reached significance in any of the models. Approximately 70% of the agencies in the sample did not implement any community policing activities listed under this measure. For most law enforcement agencies, conducting a survey or sponsoring a sur- vey without any serious political pressure, scientific assistance, and funding is not feasible.

The training and problem-solving dimension consistently shows a significant positive impact on both property and violent crime rates in small agencies contrary to the study’s hypotheses. In large agencies, similar to the effect of community contribution, training and problem-solving does not have a significant impact on property and violent crime rates. In the LEMAS survey questionnaire, the question that asks agencies about person- nel training on community policing provides a cut-off point for training (a minimum of 8 h). For a strategy that requires an understanding of its philosophy and a radical change in the traditional police culture, an 8-h training requirement seems inadequate. If a majority of agencies score on this dimension by providing very brief and superficial training, then it is possible that any crime reduction effect with community policing training will not be detected. For large agencies, even though the direction of relation is negative as hypothe- sized, it never reached statistical significance.

516 M.A. Sozer and A.V. Merlo

The problem-solving partnership dimension of community policing has a positive association with both property and violent crime rates in all models. In comparison with the community contribution dimension, the problem-solving partnership is more prevalent among agencies. Over 70% of the agencies have a written agreement or problem-solving partnership with at least one community entity. It appears that community policing does not produce a crime prevention effect. Overall, in contrast to the hypotheses, an increase in community policing activities results in higher property and violent crime rates.

The authors suggest that there are three main possible explanations for this consistent relationship. First, as evident in number of police and crime nexus, agencies having higher crime rates are likely to implement greater community policing-related activities that make it difficult to truly sort out cause and effect. Second, more community policing activities might also increase two-way interaction between the public and police which, in turn, might yield more crime being brought to the attention of law enforcement agen- cies. Schnebly (2008) found that residents are more likely to report their victimization experiences to the police in jurisdictions where agencies have a higher percentage of offi- cers who have received community policing training. Third, a written problem-solving agreement does not guarantee a real problem-solving partnership between police and the public. Therefore, although agencies reported that they have a written problem-solving partnership on paper, they do not function necessarily through a problem-solving approach.

Conclusion

This study did not find any evidence to justify the crime reduction effect of community policing. For large agencies, community contribution and training and problem-solving activities do not have any association with property and violent crime rates. However, they are associated with higher crime rates in small agencies. The problem-solving part- nership dimension of community policing is associated with higher property and violent crime rates in both large and small agencies. It is important to note that with the present study design, results cannot illustrate a casual relationship between community policing and crime rates. Like in police and crime equation, it is very difficult to sort out whether crime or community policing is the cause. However, to speculate we can refer to Sher- man and Eck (2006, p. 295) who suggested that community policing without a clear focus on crime risk factors generally shows no effect on crime. It is found that the rela- tive importance of some of the structural control variables exceeds the relative importance of explanatory (community policing) variables. If collective awareness and response from a variety of social institutions target the crime problem in a community, then crime reduc- tion is a reasonable expectation (Eck & Maguire, 2000). Therefore, it can be argued that community policing efforts are less likely to provide expected results unless structural- level indicators are improved in the community.

In contrast to the crime reduction expectation, community policing appears to be asso- ciated with higher crime rates. Agencies implementing community policing to a greater extent might fail to address crime risk factors. If so, for the future, agencies utilizing community policing activities should consider crime risk factors and partner with commu- nity agencies and organizations when developing crime prevention strategies. Considering the limitation of data, the study suggests that adaptation of these strategies by police departments might be symbolic and especially some dimensions such as surveying community, evaluating survey results, and problem-solving require expertise that might preclude its thorough implementation.

Police Practice and Research: An International Journal 517

The authors would be remiss if they did not advocate for more research in this area. As Weisburd, Telep, Hinkle, and Eck (2010) have noted, more funding, research, and analysis of the techniques and strategies that work best in policing have to be undertaken. The public, elected officials, and administrators attest to the popularity of the approach, but rigorous research is lacking.

The effect of community policing on crime rate is less than conclusive. Although plausible explanations about why this is the case are presented, it remains a puzzling research question for future research endeavors.

Notes 1. An earlier version of this paper was presented at the Academy of Criminal Justice Sciences

Annual Meeting on 26 February 2010 in San Diego, California. 2. LEMAS 2003 is the latest version of the series available. It has been conducted in three-year

intervals since 1987. 3. Oliver (2000), GAO (1995), Hawkins and Weisheit (2003), Roth et al. (2004), Roth and Ryan

(2000), and Wells and Weisheit (2004) all used a population of 50,000 to categorize agencies regarding their size (small agency = population less than 50,000; large agency = population greater than 50,000). In addition, the COPS office also used the criterion of population less than 50,000 when providing the FAST grants to the small agencies. In lieu of running analy- ses at multiple population points to see whether there is a difference between two different size populations, the authors selected the cut-off point of 50,000 just as the majority of previ- ous literature did.

4. Factor loading value of .497 was accepted for inclusion of the item ‘Used info for redistrict- ing areas’ as it is so close to the cut-off value of .50.

5. Cordner’s dimensions and elements of community policing and the four objectives of COPS and their associated items used by Roth et al. (2004) were cross-matched to provide evidence for the face and content validity of the scales. Cross-matching scales with two other defini- tions’ elements illustrate that scales have merit in providing measures for community policing despite the fact that the name of each scale does not exactly match with the name of Cord- ner’s and Roth et al.’s dimensions

6. 1 = four-year college, 2 = two-year college, 3 = some college but no degree, 4 = high school diploma or equivalent, 5 = no formal education. The variables 4 and 5 are recoded as 0 refer- ring to ‘no college degree’; 1, 2 and 3 are recoded as 1 referring to ‘some college degree’.

7. Several transformations were made and regression analyses were conducted with transformed variables. Since transformed variables did not perform any better in terms of meeting regression assumptions, the original unit of the variables was utilized. White-Pegan tests for heteroskedasticity were run to test for any violation of the assumption of homoskedasticity. Since there was a violation of this assumption, a robust method (Huber’s influence function) was utilized in the regression analyses, and robust standard errors were reported.

8. Instead of using three factors in the OLS regresssion analyses, researchers also utilized a sin- gle community policing varibale (a summated community policing index) that was generated by summing the score of each variable. The mean score for the index is 9.64 with a standard deviation of 7.84, and a skewness of .819. OLS regression analyses were conducted on a full sample (without splitting data as small vs. large departments) with this summated index. A dummy variable agency size (0 = small and 1 = large) included in the analysis as well. Results revealed that for property crime, the level of implementation of community policing is positively associated with property crime rates (b = .515, p < .001); and large agencies are associated with lower property crime rates (b = −2254, p < .010). For violent crime, the level of implementation of community policing is again positively associated with violent crime rates (b = .199, p < .001); and large agencies are associated with lower violent crime rates (b = −1.057, p < .05).

Notes on contributors Mehmet Alper Sozer received his PhD in Criminology at Indiana University of Pennsylvania in 2008. Currently, he is an Assistant Professor at the Turkish National Police Academy in Ankara,

518 M.A. Sozer and A.V. Merlo

Turkey. He teaches criminology, research methods, and crime prevention. In addition to teaching, he is a researcher at the ‘International Center for Terrorism and Transnational Crime’ in Turkey. He has had 10 years field experience in law enforcement. His most recent publications include Community Policing: Society, Crime and Safety (2010) as an editor and Crime Prevention Strategies (2010) co-edited with Daniel Lee in Turkish.

Alida V. Merlo is a Professor of Criminology at Indiana University of Pennsylvania. Previously, she taught at Westfield State College in Westfield, Massachusetts. Her primary research interests are juvenile justice and criminal justice policy. She is the past president of the Academy of Criminal Justice Sciences. Her most recent publications include Controversies in Juvenile Justice and Delinquency (2009) co-edited with Peter J. Benekos; and ‘Policing in Korea: Why Women Choose Law Enforcement Careers’ co-authored with Bitna Kim in the Journal of Ethnicity in Criminal Justice (2010).

References Beckman, K.A. (2006). Community policing and changing crime rates: Does what police do

matter? Masters Abstracts International, 44(05), 133 [UMI No. 1433715]. Benedict, W.R., Bower, D.J., Brown, B., & Cuningham, R. (1999). Small town surveys: Bridging

the gap between police and the community. Journal of Contemporary Criminal Justice, 15(2), 144–154.

Blumstein, A., & Wallman, J. (2000). The crime drop in America. New York, NY: Cambridge University Press.

Bowers, W.J., & Hirsch, J.H. (1987). The impact of foot patrol staffing on crime and disorder in Boston: An unmet promise. American Journal of Police, 6, 17–44.

Braga, A.A., & Weisburd, D.L. (2006). Police innovation and crime prevention: Lessons learned from police research over the past 20 years. Washington, DC: National Institute of Justice.

Brand, M.W., & Birzer, M.L. (2003). The benefits of community policing in rural Oklahoma. In Q.C. Thurman & E.F. McGarrel (Eds.), Community policing in a rural setting (pp. 105–112). Cincinnati, OH: Anderson.

Cordner, G.W. (1988). A problem-oriented approach to community-oriented policing. In J. Greene & S. Mastrofski (Eds.), Community policing: Rhetoric or reality (pp. 135–152). New York, NY: Praeger.

Eck, J.E., & Maguire, E.R. (2000). Have changes in policing reduced violent crime? An assess- ment of the evidence. In A. Blumstein & J. Wallman (Eds.), The crime drop in America (pp. 207–265). New York, NY: Cambridge University Press.

Evans, W.N., & Owens, E.G. (2007). COPS and crime. Journal of Public Economics, 91, 181–201. Federal Bureau of Investigation (FBI). (2006). Crime in the United States. Retrieved October 5,

2007, from http://www2.fbi.gov/ucr/cius2006/offenses/ Government Accountability Office (GAO). (1995). Community policing: Information on the COPS

on the beat grant programs, October, 1995. Washington, DC: Author. Government Accountability Office (GAO). (2003). Technical assessment of Zhao and Thurman’s

2001 evaluation of the effects of cops grants on crime, June, 2003. Washington, DC: Author. Government Accountability Office (GAO). (2005). Interim report on the effects of cops funds on

the decline in crime during the 1990s, June, 2005. Washington, DC: Author. Greene, J.R., & Mastrofski, S.D. (1988). Community policing: Rhetoric or reality. New York, NY:

Prager. Hawkins, C.W., & Weisheit, R.A. (2003). The state of community policing in small towns and

rural areas. In Q.C. Thurman & E.F. McGarrel (Eds.), Community policing in a rural setting (pp. 21–29). Cincinnati, OH: Anderson.

Heimer, K., & Coster, S. (1999). The gendering of violent delinquency. Criminology, 37(2), 277–318. Hickman, M.J., & Reaves, B.A. (2003). Local police departments 2000. Washington, DC:

Department of Justice. Kessler, D.A., & Duncan, S. (1996). The impact of community policing in four Houston neighbor-

hoods. Evaluation Review, 20(6), 627–669. Kornhauser, R. (1978). Social sources of delinquency. Chicago, IL: University of Chicago Press. Lab, S.P., & Das, D.K. (2003). International perspectives on community policing and crime pre-

vention. Upple Saddle River, NJ: Prentice-Hall.

Police Practice and Research: An International Journal 519

LaFree, G. (1998). Social institutions and crime bust of the 1990s. The Journal of Criminal Law in Criminology, 88, 1325–1368.

Levitt, S. (1997). Using electoral cycles in police hiring to estimate the effect of police on crime. American Economic Review, 87, 270–290.

Lurigio, A.J., & Rosenbaum, D.P. (1986). The evaluation research in community crime prevention. A critical look in the field. In D.P. Rosenbaum (Ed.), Community crime prevention: Does it work (pp. 19–44). Beverly Hills, CA: Sage.

MacDonald, J.M. (2002). The effectiveness of community policing in reducing urban violence. Crime & Delinquency, 48(4), 592–618.

Markowitz, F.E., & Felson, B.R. (1998). Social-demographic attitudes and violence. Criminology, 36(1), 117–138.

Marvell, T., & Moody, C. (1996). Specification problems, police levels, and crime rates. Criminol- ogy, 34(2), 609–646.

Mastrofski, S. (2006). Community policing: A skeptical view. In D.L. Weisburd & A.A. Braga (Eds.), Police innovation: Contrasting perspectives (pp. 44–76). New York, NY: Cambridge University Press.

Messner, S. (1982). Poverty, inequality, and the urban homicide rate. Criminology, 20, 103–114. Norusis, M.J. (2000). Guide to data analysis. Upper Saddle River, NJ: Prentice-Hall. Office of Community Oriented Policing Services (COPS Office). (2000). Shared understanding of

community policing. Retrieved September, 12, 2007, from http://www.cops.usdoj.gov Oliver, W.M. (2000). The third generation of community policing: Moving through innovation,

diffusion, and institutionalization. Police Quarterly, 3(4), 367–388. Osgood, D.W., & Chambers, L.M. (2000). Social disorganization outside the metropolis: An

analysis of rural youth violence. Criminology, 23(2), 81–114. Palacios, W.R. (2003). Where is Mayberry: Community-oriented policing and officers of color. In

J. Joseph & D. Taylor (Eds.), With justice for all: Minorities and women in criminal justice (pp. 65–78). Upper Saddle River, NJ: Pearson.

Paternoster, R., Brame, R., Mazerolle, P., & Piquero, A. (1998). Using the correct statistical test for the equality of regression coefficients. Criminology, 36(4), 859–866.

Police Foundation. (1981). The Newark foot patrol experiment. Washington, DC: Police Foundation. Roth, J.A., & Ryan, J.F. (2000). The national evaluation of the cops program. Washington, DC:

National Institute of Justice. Roth, J.A., Roehl, J., & Johnson, C.C. (2004). Trends in adaptation of community policing. In

W.G. Skogan (Ed.), Community policing: Can it work? (pp. 3–29). Belmont, CA: Wadsworth. Sampson, R. (1987). Urban black violence: The effect of male joblessness and family disruption.

American Journal of Sociology, 93, 348–382. Sampson, R.L., & Groves, W.B. (1989). Community structure and crime: Testing social

disorganization theory. The American Journal of Sociology, 94(4), 774–802. Sampson, R.J., & Wilson, J. (1995). Toward a theory of race, crime, and urban inequality. In J.

Hagan & R. Peterson (Eds.), Crime and inequality (pp. 37–54). Palo Alto, CA: Stanford University Press.

Schnebly, S.M. (2008). The influence of community-oriented policing on crime reporting behavior. Justice Quarterly, 25(2), 223–251.

Schneider, A.L. (1978). The Portland forward records check of crime victims. Washington, DC: Government Printing Office.

Shaw, C.R., & McKay, H.D. (1942). Juvenile delinquency and urban areas. In F.T. Cullen & R. Agnew (Eds.), Criminological theory: Past to present (essential readings) (2nd ed., pp. 104– 110). Los Angeles, CA: Roxbury.

Sherman, D., & Eck, J.E. (2006). Policing for crime prevention. In L.W. Sherman, D.P. Farrington, B.C. Welsh, & D.L. MacKenzie (Eds.), Evidence-based crime prevention (pp. 331–403). New York, NY: Routledge.

Skogan, W.G. (2006). Police and the community in Chicago. New York, NY: Oxford University Press.

Tien, J.M., & Cahn, M.F. (1986). The commercial security field test program: A systematic evaluation of security surveys in Denver, St. Louis, and Long Beach. In D.P. Rosenbaum (Ed.), Community crime prevention: Does it work (pp. 228–250). Beverly Hills, CA: Sage.

Uchida, C.D., Forst, B., & Annan, S.O. (1992). Modern policing and the control of illegal drugs: Testing new strategies in two American cities. Washington, DC: National Institute of Justice.

520 M.A. Sozer and A.V. Merlo

Wang, Y.R. (2006). Does community policing motivate officers at work and how? International Journal of Police Science and Management, 8(1), 67–77.

Weisburd, D., Telep, C.W., Hinkle, J.C., & Eck, J.E. (2010). Is problem-oriented policing effective in reducing crime and disorder? Criminology & Public Policy, 9(1), 139–172.

Weisburd, D., & Eck, J. (2004). What can police do to reduce crime, disorder, and fear? Annals of the American Academy of Political and Social Science, 593, 42–65.

Weisheit, R.A., Wells, L.E., & Falcone, D.N. (1994). Community policing in small town and rural America. Crime & Delinquency, 40(4), 549–567.

Wells, L.E., & Weisheit, R.A. (2004). Patterns of rural and urban crime: A county-level comparison. Criminal Justice Review, 29(1), 1–22.

Worrall, J.L., & Kovandzic, T.V. (2007). COPS grant and crime revisited. Criminology, 45(1), 159–185.

Wycoff, M.A., & Skogan, W.G. (1993). Community policing in Madison: Quality from the inside out. An evaluation of implementation and impact. Technical Report. Washington, DC: Police Foundation.

Yin, R.K. (1986). A synthesis of eleven evaluations. In D.P. Rosenbaum (Ed.), Community crime prevention: Does it work? (pp. 294–308). Beverly Hills, CA: Sage.

Zawitz, M.W., Klaus, P.A., Bachman, R., Bastian, L.D., DeBarry, M.M.Jr., Rand, M.R., & Taylor, B.M. (1993). Highlights from 20 years of surveying crime victims: The national crime victim- ization survey, 1973–1992. Washington, DC: Bureau of Justice Statistics.

Zhao, J., Scheider, M.C., & Thurman, Q. (2003). A national evaluation of the effect of COPS grants on police productivity (arrests) 1995–1999. Police Quarterly, 4, 387–409.

Zhao, J.S., Scheider, M.C., & Thurman, Q. (2002). Funding community policing to reduce crime: Have cops grants made a difference. Criminology and Public Policy, 2(1), 7–32.

Zhao, J., & Thurman, Q.C. (2003). Facilitators and obstacles to community policing in rural Amer- ica. In Q.C. Thurman & E.F. McGarrel (Eds.), Community policing in a rural setting (pp. 29– 34). Cincinnati, OH: Anderson.

Zhao, J., & Thurman, Q.C. (2004). Funding community policing to reduce crime: Have COPS grants made a difference? Washington, DC: Office of Community Oriented Policing Services.

Police Practice and Research: An International Journal 521

Copyright of Police Practice & Research is the property of Routledge and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.