Determinants of Police Strength in Large U.S. Cities During the 1990s

Abstract

The 1990s represented a unique decade in which to analyze the determinants of police strength in the United States. This decade was a time in which crime initially increased, then substantially decreased. Furthermore, this decade also was characterized by increases in the minority population throughout large American cities. Finally, the 1990s were characterized by increasing police budgets. These realities have direct implications for the competing theories of police growth. This research examines the determinants of police strength in large U.S. cities from 1990 to 2000. A fixed-effects panel analysis was used to assess the number of sworn police officers per 100,000 population. The findings support the resource dependency and social conflict perspectives. Additionally, no support was found for the rational public choice perspective.

Keywords

police police strength conflict theory crime drop police growth

Introduction

The evolution of American police departments has been thoroughly detailed in scholarly research. Researchers have painstakingly documented the origin and historical development of police agencies, particularly in large cities (e.g., Fogelson, 1977; Lane, 1980; Monkkonen, 1981; Walker, 1977). There seems to be a consensus about how American police have changed from a semiprofessional occupation under the control of a local political machine more than a century ago to a well-organized and largely independent organization today (Greene, 2000; Kelling & Moore, 1988; Reiss, 1992; Uchida, 1993). There is a lack of agreement, however, on the factors associated with the development of American police in a democratic society.

The debate over the role and development of American police continues today (see Jacobs & Helms, 1997; Kent & Jacobs, 2005; Nalla, Lynch, & Leiber, 1997; Stucky, 2005). Three specific approaches seek to explain the fluctuation in the number of police officers, namely, the rational public choice theory, social conflict theory (economic and racial), and resource dependency theory. Each approach takes a unique view to explain the growth of American police departments.

Drawing on these three perspectives, the purpose of this study is to advance knowledge on the growth of large American police departments during the 1990s, which represents a very unique period of time for two reasons. First, the 1990s represented the first decade in 40 years in which the crime rate first increased then substantially decreased. This change was especially pronounced in large cities, which are the focus of the current study (Zhao, Scheider, & Thurman, 2002). Similarly, the tumultuous national economy during the 1990s went from a healthy economic situation, to a pronounced downturn, and then ended in unprecedented prosperity. Despite these factors, the number of officers has continued to rise, from 226.5 officers per 100,000 residents in 1990 to 242.8 officers per 100,000 residents in 2000, a net gain of 7.2% (Law Enforcement Officers Killed and Assaulted [LEOKA], 1990, 2000). Second, the decade from 1990 to 2000 represented a time in which both the population of minority residents and the size of police budgets continued to rise while the unemployment rate continued to fall. To date, there has been no empirical study that examines this unique period during the last decade of the 20th century. The sample in this study includes all large cities with a population of more than 150,000 people in the United States when data were available. A two-factor fixed-effects panel analysis is used to estimate the relative contribution of the three competing theories.

Literature Review

After a review of the existing literature, three major perspectives that attempt to explain changes in police strength were identified: rational public choice theory, social conflict theory (economic and racial), and resource dependency theory. Each theory offers a unique approach to explaining the reason for an increase in the size of a police force.

Rational Public Choice Theory

Rational public choice theory posits that the supply of public goods and services is determined by the demands of the citizens similar to the general law of supply and demand. Citizens act rationally in response to escalating crime rates by demanding an increase in municipal services, such as a greater police presence. Bergstrom and Goodman (1973) argued that decisions about the quantities of goods and services provided must be made collectively by the citizenry based on an objective assessment of the external environment. Consequently, a collective electorate, aware of the environment surrounding them, demands the optimal level of governmental services, such as the proper number of police officers employed (Borcherding & Deacon, 1972). In addition, the rational public choice perspective implies that citizens are aware of the crime rates in their communities; thus, municipal police size should be proportional to the reported crime rate (Liska, Lawrence, & Benson, 1981).

Studies using quantitative methods to analyze the increase in police during the second half of the 20th century reported mixed findings concerning this perspective. In a cross-sectional analysis of 91 standard metropolitan statistical areas (SMSAs) in the United States, Jacobs (1979) found that crime rates had a positive and significant effect on police and detectives per capita and all law enforcement per capita in the 1970s. While empirical support does exist for the rational public choice theory, other studies have not reported a relationship between crime rates and police strength (see Jackson & Carroll, 1981; Loftin & McDowall, 1982). For example, Greenberg, Kessler, and Loftin (1985) used a panel design to analyze the determinants of social control in cities with 50,000-plus residents in the United States from 1950 to 1980. The authors found that both violent and property crime rates had no effect on the number of full-time police employees.

Other studies seeking to analyze the rational public choice perspective at different units of analysis have also produced mixed results. For example, Jacobs and Helms (1997) found support for the rational public choice theory at the national level. Their time-series analysis of the number of federal, state, and local police employees in the United States from 1952 to 1991 indicated that an increase in the crime rate resulted in an escalating law enforcement presence in the following year. Nalla (1992) also found support for the rational public choice theory at the national level between 1948 and 1984. In a time series analysis, the crime rate, operationalized as the number of personal and property offenses per 100,000 persons, had a positive and significant effect on police expenditures in the United States (Nalla, 1992). Prior studies that analyzed individual cities, however, have not found support for the rational public choice perspective. For example, McDowall and Loftin’s (1986) time-series analysis of Detroit from 1928 to 1976 indicated that serious crime did not influence the number of police employees.

Although researchers have operationalized and discussed the rational public choice theory in terms of many different types of crime, the central attention of crime control policy is usually paid to violent crime incidents. Some policing scholars are particularly frank about this point. For example, James Q. Wilson (1975) singled out violent street crime as the predominant concern of American society. He argued that “predatory crime, in particular crime committed by strangers on innocent victims, causes the kind of fear that drives people apart from one another and thus impedes or even prevents the formation of meaningful human communities” (Wilson, 1975, p. 5). Following this logic, the current study focuses on the relation between violent crime and police strength.

Social Conflict Theory

Social conflict theory traces the increase in the number of police officers to both economic and racial conflict (Turk, 1969). Although they are both steeped in conflict theory, economic and racial conflict perspectives are often separated because they each have unique measures and associated underlying logic.

Economic conflict theory

Economic conflict theorists argue that the law is not administered impartially for the benefit of the majority in society. On the contrary, differences in economic status allow those who possess resources to subjugate those who do not possess resources (see Chambliss, 1976; Chambliss & Seidman, 1982; Jacobs, 1979; Quinney, 1974, 1975). The best way to maintain the status quo is through a strong police force that controls the economically disadvantaged class (Jacobs, 1979; Turk, 1969). In addition to possessing economic resources, those with power are easily able to gain political influence (Chambliss & Seidman, 1982). Economic conflict theorists thus propose that increased formal control mechanisms, such as the police, are a result of the wishes of those with power who attempt to advance their own interests. Researchers have commonly used the unemployment rate to measure economic conflict with the expectation that high-status individuals will view the unemployed as a threat to the existing social order (Kent & Jacobs, 2005).

The results of research seeking to link economic inequality to police strength have been mixed. In a cross-sectional study of all SMSAs with a population of greater than 250,000 in 1960 and 1970, Jacobs (1979) found that economic inequality had a positive and significant relationship with police and detectives per capita and all law enforcement personnel per capita. Recently, Kent and Jacobs (2005) utilized the natural log of the unemployment rate to assess the effect of economic inequality on police strength.¹ In their panel analysis of 125 cities in the United States with populations of at least 100,000 residents over three waves from 1980 to 2000, Kent and Jacobs (2005) found that the unemployment rate was a significant and positive predictor of police strength. The historical schism between those with economic status and those without economic status has been supported through prior qualitative research. Silver (1966) found that the police were initially intended to protect those with property, particularly in the South in the 19th century. One hundred years later, the National Advisory Commission on Civil Disorders (1968) chronicled the gulf between the decision makers in local governments and the residents of the ghettos in the city. In essence, the impasse between the economically privileged and the economically unprivileged has been an enduring theme over time.

While considerable support does exist for the relationship between economic inequality and police strength, other studies have been forced to rely on suspect indicators of economic conflict or have found no support for the theory. Chamlin (1990), for example, was forced to utilize the unemployment rate for the entire United States in a longitudinal analysis of police expenditures in Chicago because data of that nature were not available for that specific city.² Other research on the relationship between economic inequality and police strength has either found no significant relationship (Greenberg et al., 1985) or a significant and negative relationship (Nalla, 1992) that contradicts the underpinnings of the theory.

Racial conflict theory

While the economic conflict model argues that class differences result in an increased police presence, racial conflict theorists argue that police strength is the result of a greater presence of racial and ethnic minorities (Turk, 1969). An increasingly visible minority population can be perceived as a crime problem (Blalock, 1967). As this perceived criminal threat from racial and ethnic groups becomes more salient, racial conflict theorists state that the majority White population often makes successful political demands for an additional police presence to quell the fear (Jacobs, 1979; Liska et al., 1981). The actual measured crime rate thus has a less substantial effect on the number of police officers than the perceived threat of criminal activity, as influenced by the population of racial and ethnic minorities.

Empirical research has indicated a fairly consistent relationship between the size of the minority population and the strength of the police (see Chamlin, 1990; Greenberg et al., 1985; Huff & Stahura, 1980; Jackson & Carroll, 1981; Jacobs, 1979; Kent & Jacobs, 2005; Liska et al., 1981; Sharp, 2006). In their panel study of large U.S. cities between 1980 and 2000, Kent and Jacobs (2005) found a significant and positive relationship between racial composition, as measured by the natural log of the percentage of the Black population, and the strength of police. In addition, they found that the relationship between the percentage of African Americans and the per capita police officers has become stronger in the recent two census years (1990 and 2000) (Kent & Jacobs, 2005). These findings indicate that the link between the size of the minority population and the size of the police force may still be enduring despite the progress in race relations since the civil rights era (Kent & Jacobs, 2005). Other researchers have also found a significant relation between the African American population and police strength. In a time-series analysis of Milwaukee, Brandl, Chamlin, and Frank (1995) found that the percentage of Black residents had a significant and positive effect on the number of patrol officers.

While the aforementioned two studies focused specifically on the African American population, there has been debate about how to best operationalize the minority population. Liska et al. (1981), for example, focused on the total percentage of non-White residents. Other researchers have focused on the population of Hispanic residents (e.g., Jackson & Carroll, 1981) or the total percentage of minority residents (e.g., Nalla et al., 1997). In a thorough review and analysis of the relationship between minority populations and police strength, Sever (2001) found that the percentage of Black residents had the most consistent effect on police strength. Following the logic of Sever (2001), the current study focuses primarily on the relation between the African American population and police strength during the 1990s.

Resource Dependency Theory

Resource dependency theory states that the number of police officers is contingent on the amount of available financial resources within an organization (Donaldson, 1995; Pfeffer & Salancik, 1978). The theory holds that external resources and the ability to change are vital for organizations to survive and those resources that are so pivotal for survival and change are also scarce and limited (Pfeffer & Salancik, 1978). This theory remains popular in management science today (Casciaro & Piskorski, 2005; Pfeffer & Salancik, 2003). Applying the resource dependency perspective to the growth of the number of police officers, the theory posits that change in police personnel is highly dependent on the availability of budgetary resources.

Prior research has mostly supported the resource dependency perspective at numerous different units of analysis. At the city level, Greenberg et al. (1985) analyzed revenue per capita as a determinant of police growth in their panel study of police employees in 259 U.S. cities between 1950 and 1980. The authors found the city revenue per capita variable had a positive relationship with police strength until the 1970 to 1980 panel (Greenberg et al., 1985). Research analyzing individual cities has also supported the resource dependency perspective. Nalla et al. (1997) found that police expenditures in Phoenix, Arizona, had a positive and significant effect on police expenditures in the following year. McDowall and Loftin (1986) used multiple indicators of revenue in analyzing the determinants of police growth in Detroit between 1928 and 1976. The variable measuring per capita revenues generated by the city of Detroit had a positive and significant effect on the number of police officers (McDowall & Loftin, 1986).

The current study hopes to add to the body of knowledge about the determinants of police strength by studying large American police departments. In addition, the current research analyzes the determinants of police strength during the 1990s, which represents the first decade in 40 years in which the crime rate first increased and then substantially decreased. To date, this is the only study that focuses specifically on the 1990s when seeking to assess the determinants of police strength in American police departments.

Methodology

Sources of Data

The data used in this analysis were derived from five principal sources. The first source was obtained through the Police Employee (LEOKA) data, which is one of four types of files found in the Uniform Crime Reports (UCR). These data are intended to provide information about law enforcement officers killed or assaulted in the line of duty. In addition, the Police Employee data provide more general information about the number of officers and the ratio of sworn police officers per population for police departments throughout the nation. The Police Employee annual data for the period from 1991 to 2000 were downloaded from the Inter-University Consortium for Political and Social Research (ICPSR) Web site (http://www.icpsr.umich.edu). A second type of data was obtained through the Annual Finance Survey of City Government (U.S. Census Bureau, 1992). The survey of local government finance officials asked questions concerning a wide variety of financial situations in a city government with respect to revenues, expenditures, debts incurred, and assets (cash, capital possessions, and security holdings). A unique feature of this annual financial survey was that all cities were required to complete the survey and submit it to the U.S. Census Bureau annually. In this study, financial information such as total expenditures for each city for the period from 1991 to 2000 was downloaded from the Census Bureau’s Web site (http://www.census.gov/govs/).

The third source of data is the UCR published annually by the Federal Bureau of Investigation. UCR data reflect a nationwide effort to collect criminal activity data from approximately 17,000 state, county, and municipal law enforcement agencies voluntarily reporting crimes that have been brought to their attention. These data on “crimes known to the police” contained in the UCR database for the period from 1990 to 1999 were also downloaded from the ICPSR Web site. The fourth type of data gathered is the annual unemployment rate data for the period from 1991 to 2000, obtained from the Bureau of Labor Statistics Web site, which contains annual information about cities with more than 25,000 residents (http://www.bls.gov). Finally, demographic information at the city level is also included from the 1990 and 2000 United States Census. The 1990 and 2000 data were obtained from the DVD disks published by the U.S. Census Bureau.

Variables in the Analysis

Dependent variable

The dependent variable for this study is police strength, measured as the number of sworn police officers per 100,000 residents. Following prior research, a ratio of the number of sworn officers per population was utilized (see Jacobs & Helms, 1997; Kent & Jacobs, 2005; Nalla et al., 1997).

Independent variables

Five independent variables were utilized to represent the rational public choice theory, conflict theory, and resource dependency theory. The violent crime rate represents the rational public choice theory.³ It reflected the sum of the incidents of four specific crimes against persons (murder, rape, robbery, and aggravated assault), divided by each city’s population and multiplied by 100,000. The authors selected this variable because violent crimes were the most salient type of crimes with respect to media coverage and public discussion, and they served as the centerpiece of the policy propositions in criminal justice and in the marketplace of political ideas and election campaign debates (Currie, 1985; Walker, 1985; Wilson, 1975). The violent crime rate variable was lagged for 1 year in the analysis, as was the common practice in the analysis of panel data to tap into the simultaneous relationship between crime and police strength (e.g., Marvell & Moody, 1996). For example, the 1990 violent crime rate was used to predict police strength in 1991. Such a lag effect was presumed throughout the analysis to follow.

Three variables were included to represent the conflict theory. First, the unemployment rate and per capita income were included as the measures of economic conflict in this study. There are a variety of different ways to measure economic inequality (see Alker & Russett, 1964). The unemployment rate was chosen as the measure of economic inequality in this study, because it is plausible that the economically and politically elite will view a growing population of individuals without work as a threat to the existing social structure (Chambliss, 1964; Kent & Jacobs, 2005). Furthermore, an increasing number of unemployed individuals may cause resentment against the underclass and subsequently encourage demands for a stronger police force (Kent & Jacobs, 2005). The time-varying city-level unemployment rates were obtained from the Bureau of Labor Statistics. Per capita income has long been used as a measure of poverty level as well as an indicator of the economic health of an area in prior studies (e.g., Marvell & Moody, 2001; Worrall, 2004). A per capita measure rather than the median family income was used to adjust for differences across cities and family size (Ross & Sawhill, 1975). To adjust for inflation, the annual consumer price index (CPI) was used to convert the nominal dollars of per capita income from 1991 to 2000 to constant 2000 dollars. The CPI, reported by the Bureau of Labor Statistics, is the most commonly used indicator of inflation.

Second, following previous research, the percentage of the population that is Black was used to represent racial conflict (see Kent & Jacobs, 2005; Sever, 2001). The percentage Black variable was calculated at the city level by adding the number of Black residents and dividing by the total number of residents. Data were obtained from the U.S. Census Bureau Web site. Because the U.S. Census is taken on a decennial basis, the estimated percentages for the missing years were calculated by using linear interpolation.

Finally, the annual city expenditure per resident was used as a measure of the resource dependency theory. City expenditure is defined as “all amounts of money paid out by a government during its fiscal year. . . . Expenditure includes amounts spent by all agencies, boards, commissions, or other organizations categorized as dependent on the government concerned” (U.S. Census Bureau, 1992). It was measured as the total amount of annual city expenditure (in dollars) divided by the city’s population for the given year, adjusted by the annual CPI.

Control variables

Five control variables designed to account for the socioeconomic health of cities also were included. The social disorganization perspective (Shaw & McKay, 1942) provides a framework for the inclusion of these variables in the current analysis (see also Bursik, 1988; Smith & Jarjoura, 1988). Recently, Garland (2000) argued that social disorganization increases demands for order. The crime rate may not capture all elements of social disorganization, but a greater perception of these threatening acts may still increase anxiety and lead to greater demands for a stronger police presence (Garland, 2001; Tyler & Boeckman, 1997). For this reason, it is necessary to control for various aspects of social disorganization that may not be captured by the crime rate. In this study, social disorganization is measured by five variables: the percentage of single-parent households, percentage of young people between the ages of 15 to 24, the percentage of home ownership, the population density (number of people in a square mile), and the percentage of people having lived at the same address for 5 years or more before the census. Researchers have consistently included variables measuring social disorganization in previous studies of police strength (see Kent & Jacobs, 2005; Nalla et al., 1997). Because census data on these measures were available only for every decade, linear interpolation was used to compute the missing values of these socioeconomic variables from 1991 through 1999 (Kovandzic, Sloan, & Vieraitis, 2002; Worrall & Kovandzic, 2007).

There were 112 cities in 1990 that reached the population of 150,000 or more residents. Among these municipalities, some 85 cities had complete data for police strength, violent crime rates, unemployment rate, per capita income, percentage of Blacks, annual municipal expenditures, and five control variables. Data for these 85 cities were used for the analysis.

Statistical Model Specification

This study used a panel design (longitudinal data, cross-sectional time series data, or multiple time-series design), which has long been considered the preferred method for the study of causation. For example, Campbell and Stanley (1967, pp. 55-57) referred to panel models as “excellent quasi-experimental designs, perhaps the best of the more feasible designs.” Still other researchers argued that panel techniques were essential to the causal analysis of correlation findings derived from cross-sectional studies (e.g., Hsiao, 2003; Stimson, 1985).

More specifically, a two-factor fixed-effects panel analysis was employed to analyze the effects of rational public choice, social conflict, and resource dependency variables on police strength in 85 large cities in the United States from 1991 to 2000.⁴ A fixed-effects panel model allows the analysis to control for unobserved systematic (nonrandom) variation (Hsiao, 2003; see also Kent & Jacobs, 2005, p. 740). The “two-factor” approach accommodates a geographic component represented by the cities in which police agencies reside and a time-specific component represented by the 10 years of data for each of the 85 cities. By investigating the “first factor” (the geographic component) through the inclusion of a cross-sectional dummy variable for each city in which the police agencies reside, the difference in police strength caused by unobserved variance occurring in each city was estimated. Each city was allowed to have its own intercept but shared the slope coefficients with the other cities, a total of 84 city dummies (Marvell & Moody, 1995; Pindyck & Rubinfeld, 1998). Similarly, the “second factor” (the time-specific component) involved the inclusion of year dummy variables (nine year dummy variables) that permitted the authors to control for those unknown factors (omitted variables) affecting police strength in these large cities that were not accounted for by the other independent and socioeconomic variables. The two-factor, fixed-effects model has the following symbolic form:

Y_{it} = α + β_{1} {CRIME}_{i (t - 1)} + β_{2} {UNEMPLOY}_{it} + β_{3} {INCOME}_{it} + β_{4} {BLACK}_{it} + β_{5} {EXPENDITURES}_{it} + β_{6} {COMM}_{it} + γ_{2} W_{2 t} + γ_{3} W_{3 t} + . . . + γ_{N} W_{Nt} + δ_{2} Z_{i2} + δ_{3} Z_{i3} + . . . + δ_{T} Z_{iT} + ε_{it}

where

\begin{matrix} W_{it} = 1 & for i th individual, i = 2, . . ., N \\ 0 & otherwise \end{matrix}

\begin{matrix} Z_{it} = 1 & for t th time period, t = 2, . . ., T \\ 0 & otherwise, \end{matrix}

where Y_it is the number of sworn police officers per 100,000 residents for city i at year t. γ_NW_Nt is the fixed effects for city i to be estimated. δ_T Z_iT is the fixed effects for year t. CRIME_i(t-1) is the total violent crime rate for city i at year t – 1, representing the rational public choice theory. UNEMPLOY_it, INCOME_it, and BLACK_it represent the three measurements of social conflict theory for city i at year t. EXPENDITURES_it is the city expenditure per resident, while COMM_it are the five control variables representing community characteristics. These symbols β_i (i = 1,2,3,4,5,6) represented the associated coefficients to be estimated. Finally, ε_it is the error term. All statistical diagnoses, analyses and estimations were carried out in STATA.⁵

Findings

Figure 1 presents a graphic display of changes over time for the dependent variable and the five primary independent variables. Among these 85 cities, police strength began increasing in 1992, and the police ratio reached its highest point in 1999. More specifically, police strength went up from 221.06 in 1992 to 249.65 in 1999, a relative increase of 12.93% and an absolute growth of 28.59 officers per 100,000 population. The violent crime rate demonstrates an opposite trend. It started dropping in 1993, and the downward trend continued until 1999 (lagged one year in Figure 1). The violent crime rate decreased from its peak level of 1,594.94 per 100,000 population in 1993 to 964.14 in 1999 in these 85 large cities, consistent with the national crime drop during the same period of time. While the unemployment rate exhibited a sharp decrease over the period of the 1990s, per capita income steadily went up from $18,205.06 (in 2000 dollars) in 1991 to $19,993.94 in 2000. The percentage of the population that is Black remained nearly constant among these 85 cities with a slight difference of 1.32% between 1991 and 2000. Finally, city expenditure per resident exhibited a steady increase over the course of the decade. City expenditure per capita increased from $1,963.18 (in 2000 dollars) in 1991 to $2,290.65 in 2000.

Figure 1.

Tracking Police Strength, Public Rational Choice, Social Conflict, and Resource Dependency Between 1991 and 2000

The descriptive statistics shown in Table 1 represent the average of 10-year means of each variable used in the panel study analysis. The grand mean of police strength in the 85 cities was 235.16 sworn officers per 100,000 population. In terms of the independent variables, the average violent crime rate for the 85 cities was 1,295.53 incidents per 100,000 residents. The average total unemployment rate was 6.09% of the population in the 85 cities. In addition, per capita income was $19,099.50 (in 2000 dollars) on average, and 23.09% of the residents living in these 85 cities identified themselves as Black. Finally, the average city expenditure for the 85 cities in the analysis was $2,082.93 per resident during the period of 10 years.

Table 1.

Descriptive Statistics

Variables	Mean	SD	Min.	Max.
Dependent variable
Police strength (per 100,000)^a	235.16	98.46	100	760
Independent variables
Rational public choice
Violent crime rate (per 100,000)	1,295.53	812.63	199.82	7,323.24
Economic & racial conflict
% of unemployment	6.09	2.79	1.5	17.9
Per capita income	19,099.50	3,868.26	12,152	34,556
% of Black	23.09	19.40	.73	81.15
Resource dependency
City expenditure per resident	2,082.93	1,496.01	533.25	10,767.73
Control variables
% of single-parent households	36.03	10.84	14.15	64.30
% of young (ages 15-24)	16.03	2.42	11.20	24.53
% of home ownership	51.61	8.99	23.14	66.36
Population density	4,577.67	3,937.42	741.51	26,403.03
Living in the same house for 5+ years (residential stability)	47.89	5.66	36.00	64.09

Note: Number of cities = 85; number of observations = 850. SD = standard deviation.

The police strength variable was initially calculated as the number of sworn officers per 1,000 population. To ensure that the coefficients in the multivariate analysis would not be overly miniscule, the ratio per 1,000 was multiplied by 100 to arrive at the number of sworn officers per 100,000 population.

The demographic variables are based on two waves of census data, using linear interpolation to fill in the intervening years. The demographic variables show that 36.03% of the households were headed by single parents. In addition, the percentage of young people, ages 15 to 24, was 16.03, and more than half the population (51.61%) resided in owner-occupied dwellings. The average density of these cities between 1991 and 2000 was 4,577.67 residents per square mile. Finally, the percentage of residents living in the same house was 47.89. A complete list of the means of all variables for 10-wave data is included in Appendix A.

The results from the multivariate analysis are reported in Table 2. The initial theory attempting to explain changes in police strength is rational public choice, which posits that increases in violent crime will cause increases in police strength. The current analysis does not support the rational public choice perspective. More specifically, the violent crime rate does not have a significant impact on police force strength in the 85 cities during the 1990s.⁶

Table 2.

The Effects of Violent Crime Rate, Economic and Racial Conflict, and City Expenditures on Police Strength (1991-2000): Two-Factor Fixed-Effects Panel Model

Variables	Coefficient^a	Robust SE	t-Value
Rational public choice
Violent crime rate	−.003	.002	−1.22
Economic & racial conflict
% of unemployment	−1.26	.80	−1.56
Per capita income	−.004	.001	−2.75**
% of Black	5.54	1.08	5.11***
Resource dependency
City expenditure per resident	.02	.003	5.88***
Control variables
% of single-parent households	.10	.85	.12
% of young (ages 15-24)	−1.20	1.82	−.65
% of home ownership	1.39	.95	1.46
Population density	.03	.01	3.04**
Residential stability	−4.12	.96	−4.31***
Year 1992^b	−1.04	2.80	−.37
Year 1993	3.02	3.32	.91
Year 1994	3.26	3.87	.84
Year 1995	9.38	4.70	2.00*
Year 1996	11.90	5.58	2.13*
Year 1997	11.48	6.50	1.77
Year 1998	14.12	7.44	1.90
Year 1999	15.85	8.46	1.87
Year 2000	6.88	9.23	.75
σ_α^c	85.25
σ_εit	15.24
ρ		.97
Adjusted R²	.98
F values	23.97***

Note: SE = standard error.

The coefficients were unstandardized.

Year of 1991 was the reference group. The coefficients of 84 city dummy variables were not reported.

σ_α was the panel-level standard deviation. σ_εit was the standard deviation of ε_it (error term). ρ reflected the fraction of variance due to fixed effects.

p < .05. **p < .01. ***p < .001.

The next theory, social conflict, is measured through three variables. The first two variables, the unemployment rate and per capita income, represent the economic conflict theory. The analysis indicates that the unemployment variable is not a significant predictor of police strength, while per capita income is negatively correlated with the dependent variable. The coefficient for the per capita income variable suggests that an increase in $1 per resident in the annual income led to a decrease of about .004 sworn police officers per 100,000 in large U.S. cities over the period 1990 to 2000. In other words, every $1,000 increase per resident annually will cause a decrease of 4 sworn officers per 100,000 population. This finding lends support to the economic conflict theory. The third variable, percentage of Black population, representing the racial conflict theory, manifests highly significant effects. An increase in 1% of Black population is associated with an increase in 5.54 police officers per 100,000 residents.⁷ The significance of the Black population on police strength offers support for the racial conflict theory.

The final theory attempting to explain changes in police strength is resource dependency, which holds that police strength is influenced by municipal financial considerations. The findings lend strong support for the resource dependency theory. An increase of $100 in city expenditures per resident will result in an increase of 2 officers per 100,000 residents.

In addition to the findings regarding the three theories, two demographic predictors in the model manifested significant effects. An increase of 100 residents per square mile results in an increase of 3 sworn officers per 100,000 residents. In addition, the measure of mobility has a significant and negative effect on the police strength variable. An increase of 1% of individuals living in the same house results in a decrease of 4.12 officers per 100,000 residents.

Overall, the R-squared for the equation is extremely high at .98. This indicates that about 98% of the variance in the dependent variable, police strength, is explained by the model. An advantage of using fixed-effects panel analysis is that the relationship between the dependent and independent variables is estimated both within the groups (change in one city over time) and between groups (change among cities over time). It is essential to note that the independent and control variables explained slightly more than half of the variance in the dependent variable (R-squared 68%). The rest of the 30% was contributed by the nine year dummy variables and the 84 city-specific dummy variables. Standardized coefficients were obtained to better assess the relative importance of per capita income, percentage of Black population, city expenditure per resident, and demographic variables (see Appendix B).

Discussion and Conclusion

The purpose of this article was to assess the relative impact of three competing theories on police employment in the last decade of the 20th century. The results suggest that the 1990s proved to be a very unique decade for analyzing and discussing law enforcement trends in the United States. More specifically, four observations warrant discussion. First, despite the fact that the 1990s represented the first decade in 40 years characterized by a precipitous drop in crime, the variable measuring the percentage of Black population remained one of the strongest predictors of police strength. This finding supports the notion that decisions about police strength are based more on the perception of danger (i.e., percentage of Black population) than on objective measures of danger (i.e., crime rates). Garland (2001, p. 102), for example, reaches a similar conclusion with his assertion that law enforcement has become more “racialized” in its attempts to control populations that are perceived to be dangerous (see also Blalock, 1967). Sharp (2006) expanded by arguing that this trend seems to be static across many diverse cities, regardless of whether they are Southern or Northern or traditionally liberal or conservative. Given that the 85 large cities in this analysis represented many different geographies and political ideologies, this research further bolsters the underlying tenets of racial conflict theory.

The current analysis also demonstrates that the percentage of Black residents is the most important minority population to consider. In Appendix C, variables measuring the percentage of Hispanic residents along with the percentage of all other minority residents were included in the analysis along with the percentage of Black residents. The results indicate that the percentage of Black residents is the only measure of minority population that has a positive and significant effect on police strength. These findings are similar to Sever’s (2001) conclusion that the percentage of Black residents has the most consistent effects on measures of police strength among all indicators of the minority population. Given the visible and often cantankerous immigration debate that is taking place in the United States, it will be interesting to see if future research detects a stronger positive association between the percentage of Hispanic population and police strength, especially considering the current economic downturn that plagues the United States.

Our second observation concerns the application of resource dependency theory in police research. The findings suggest that change in police organizations largely depends on the availability of resources, which is similar to Lindbloom’s (1959) long-ago observation that the nature of organizational change is incremental. It is speculated that the increase in police strength was partially due to the role of the Office of Community Oriented Policing Services (COPS) in the 1990s, with the infusion of $8.8 billion to fund law enforcement agencies through enhancement of their community policing capabilities. From the point of time when the Crime Act was passed in September of 1994 to August 10, 2001, the COPS Office estimates that more than 83,000 new officers were on the street, which accounted for more than 11% of law enforcement employment in the country (Zhao et al., 2002, p. 11). During a roughly similar time period from September of 1994 to 2002, more than $7 billion of the initially allocated $8.8 billion was awarded to law enforcement agencies (Zhao et al., 2002, p. 9).

As Donaldson (1995) pointed out, organizational innovations are closely associated with the availability of external resources. In the 1990s, a situation existed in which the Crime Act and the COPS Office facilitated funding for law enforcement agencies throughout the country. This was especially true for cities in this sample, as the COPS Office funded every city with a population of more than 150,000 residents. This level of funding could help to catalyze organizational innovation and change in the form of new police officers; thus, a strong association existed in the 1990s between police expenditures, backed by grants from the COPS Office, and police strength, as measured through the ratio of officers to residents. There is a precedent for the association between federal subsidies and change in law enforcement agencies. In the 1970s, the Law Enforcement Assistance Administration (LEAA) subsidized a great deal of research on the police, which included the Kansas City Preventive Patrol Experiment. Ultimately, this research had a significant effect on police reform efforts, including the shift to community- and problem-oriented policing.

On the other side, the results did not lend support for the rational public choice perspective. Violent crime did not have a significant relationship with police strength in the 85 cities during the 1990s. Although this result may initially seem counterintuitive, it may simply be a by-product of the fact that although the average ratio of officers to citizens increased during the 1990s, the crime rate fell precipitously after 1993. As pointed out earlier, the crime drop was the first after more than three decades of continuous increases in the nation’s crime rate (1960s to 1992). The current study thus concludes that increases in police strength during the 1990s have little to do with changes in all measures of the crime rate after controlling for other demographic factors.

The third observation is related to the significant contribution of variables that measure the social disorganization perspective, including population density, single-parent households, and people living at the same residence for at least 5 years. Garland (2000) initially argued that demands for order in a community may be a result of social disorganization. Kent and Jacobs (2005) expounded on this point by arguing that although certain threatening acts that are caused by social disorganization may not be captured by official crime rates, they still may be powerful enough to cause residents to clamor for more police protection. For example, this analysis found that population density has a positive and significant effect on police strength. An increasingly dense population may exacerbate conflict between individuals living in closer proximity. Although these disputes may not be criminal in nature, they certainly have the potential to increase perceptions that more order is needed in an increasingly dense city or neighborhood. Just as Garland (2000) argued, social disorganization, in this case in the form of population density, may increase demands for more law and order in the form of more police officers. Additionally, this reality may not be captured through traditional crime rates.

The final observation concerns the advantage of making use of panel data analysis. Particularly, the importance of year dummy and city dummy variables is the ability to capture additional variance that cross-sectional methods cannot achieve. Panel data models have long been considered one of the best designs for the study of causation next to a purely random experiment. Lempert (1966, p. 130), for example, stated that panel designs are research designs “par excellence.” In addition, the two-factor fixed-effects approach is superior to the random-effects approach because it is able to capture unobserved systematic variance in an equation (Hsiao, 2003).

In conclusion, some of the limitations of this study need to be acknowledged. First, the current study only focuses on large cities with more than 150,000 residents. Consequently, the results of this study are not generalizable to all cities or law enforcement agencies in the United States. Second, as Sever (2001) argued, quantitative studies of police strength are limited because many important variables cannot be included in the models. Variables capturing the fear of crime felt by the residents, the presence of interest groups, and the race of police bureaucrats are difficult to gather when using a sample of multiple cities (Sever, 2001). Third, the crime drop that was so apparent in 1990s has extended into the first part of the current century as well. It might be interesting to determine whether the same relationships found in the current study of the 1990s extend and apply to the 2000s. The symmetry in results from the 1990s to the early 2000s may be complicated by the fact that COPS funding has steadily eroded since 2000. This reality may convolute the influence of resources on police strength and present interesting questions for further research on this subject. Finally, this topic seems to be conducive to studies using qualitative methods and data. Interviews with city or governmental officials about why they made certain decisions about police staffing issues would provide an excellent supplement to the many quantitative studies of police strength. Content analyses of newspapers or local news broadcasts could also elucidate how the environment in certain cities dictates decisions about the law enforcement presence.

Footnotes

Appendix

Appendix C

The Effects of Violent Crime Rate, Economic and Racial Conflict, and City Expenditures on Police Strength (1991–2000): Two-Factor Fixed-Effects Panel Model (three minority groups are included; percentage White is the reference group)

Variables	Cofficient^a	Robust SE	t–Value
Rational public choice
Violent crime rate	–.003	.003	–1.08
Economic & racial conflict
% of unemployment	–1.74	.80	–2.18*
Per capita income	–.008	.002	–4.03***
% of Black	4.22	1.15	3.65***
% of Hispanic	–3.09	.81	–3.81***
% of Other^b	.37	.68	.54
Resource dependency
City expenditure per resident	.02	.003	5.76***
Control variables
% of single-parent households	.78	.96	.82
% of young (ages 15–24)	.89	1.80	.49
% of home ownership	2.84	1.12	2.54*
Population density	.03	.01	3.94***
Residential stability	–4.61	.98	–4.71***
Year 1995^c	14.35	5.31	2.70**
Year 1996	18.15	6.40	2.84**
Year 1997	18.95	7.52	2.52*
Year 1998	22.82	8.66	2.64**
Year 1999	25.82	9.89	2.61**
σ_α^d	102.79
σ_εit	14.99
ρ		.98
Adjusted R²	.98
F values	22.07***

Note: SE = standard error.

The coefficients were unstandardized.

% of other racial/ethnic population included percentage of Asian and Pacific Islander, and percentage of American Indian and Alaska Native.

Year of 1991 was the reference group. Of the year dummy variables, five were statistically significant and reported here. The coefficients of 84 city dummy variables were not reported.

σα was the panel-level standard deviation. σε_it was the standard deviation of ε_it (error term). ρ reflected the fraction of variance due to fixed effects.

p < .05; **p < .01; ***p < .001.

Notes

References

Alker

Jr. Russett

B. M.

(1964). On measuring inequality. Behavioral Sciences, 9, 207-218.

Bergstrom

T. C.

Goodman

R. P.

(1973). Private demands for public goods. American Economic Review,63, 280-298.

Berry

W. D.

Feldman

(1985). Multiple regression in practice. Beverly Hills, CA: Sage.

Blalock

(1967). Towards a theory of minority group relations. New York: Capricorn.

Borcherding

T. E.

Deacon

R. T.

(1972). The demand for the services of non-federal governments. American Economic Review, 62, 891-901.

Brandl

Chamlin

Frank

(1995). Aggregation bias and the capacity for formal crime control: The determinants of total and desegregated police force size in Milwaukee, 1934-1987. Justice Quarterly,12, 543-562.

Bursik

R. J.

Jr. (1988). Social disorganization and theories of crime and delinquency. Criminology, 26, 519-551.

Campbell

D. T.

Stanley

J. C.

(1967). Experimental and quasi-experimental designs for research. Chicago, IL: Rand McNally.

Casciaro

Piskorski

M .J.

(2005). Power imbalance, mutual dependence, and constraint absorption: A closer look at resource dependency theory. Administrative Science Quarterly, 50, 167-199.

10.

Chambliss

W. J.

(1964). A sociological analysis of the law of vagrancy. Social Problems, 12, 67-77.

11.

Chambliss

W. J.

(1976). Functional and conflict theories of crime. In Chambliss

W. J.

Mankoff

(Eds.), Whose law? What order? (pp. 1-28). New York: John Wiley.

12.

Chambliss

W. J.

Seidman

(1982). Law, order, and power. Reading, MA: Addison-Wesley.

13.

Chamlin

M. B.

(1990). Determinants of police expenditures in Chicago, 1904-1958. Sociological Quarterly,31, 485-494.

14.

Cohen

West

S. G.

Aiken

L. S.

(2003). Applied multiple regression/correlation analysis for the behavioral sciences (3rd ed.). Mahwah, NJ: Lawrence Erlbaum.

15.

Currie

(1985). Confronting crime: An American challenge. New York, NY: Pantheon.

16.

Donaldson

(1995). American anti-management theories of organization: A critique of paradigm proliferation. Cambridge, UK: Cambridge University Press.

17.

Drukker

D. M.

(2003). Testing for serial correlation in linear panel-data models. STATA Journal,3, 168-177.

18.

Fisher

J. C.

Mason

R. L.

(1981). The analysis of multicollinear data in criminology. In Fox

J. A.

(Ed.), Methods in quantitative criminology (pp. 99-125). New York, NY: Academic Press.

19.

Fogelson

(1977). Big-city police. Cambridge, MA: Harvard University Press.

20.

Garland

(2000). The culture of high crime societies. British Journal of Criminology, 49, 347-375.

21.

Garland

(2001). The culture of control. Chicago: University of Chicago Press.

22.

Greenberg

D. F.

Kessler

R. C.

Loftin

(1985). Social inequality and crime control. Journal of Criminal Law and Criminology, 76, 684-703.

23.

Greene

J. R.

(2000). Community policing in America: Changing the nature, structure, and function of the police. In Horney

(Ed.), Policies, processes, and decisions of the criminal justice system (pp. 299-370). Washington, DC: National Institute of Justice.

24.

Hausman

(1978). Specification tests in econometrics. Econometrica,46, 1251-1271.

25.

Hsiao

(2003). Analysis of panel data (2nd ed.). New York: Cambridge University Press.

26.

Huff

C. R.

Stahura

J. M.

(1980). Police employment and suburban crime. Criminology, 17, 461-470.

27.

Jackson

P. I.

Carroll

(1981). Race and the war on crime: The sociopolitical determinants of municipal expenditures in 90 non-Southern U.S. cities. American Sociological Review,46, 290-305.

28.

Jacobs

(1979). Inequality and police strength: Conflict theory and coercive control in metropolitan areas. American Sociological Review, 44, 913-925.

29.

Jacobs

Helms

R. E.

(1997). Testing coercive explanations for order: The determinants of law enforcement strength over time. Social Forces, 75, 1361-1392.

30.

Judge

Hill

R. C.

Griffiths

W. E.

Lutkepohl

Lee

T. C.

(1988). Introduction to the theory and practice of econometrics (2nd ed.). New York: John Wiley.

31.

Kelling

G. L.

Moore

M. H.

(1988). From political to reform to community. In Greene

J. R.

Mastrofski

(Eds.), Community policing: Rhetoric or reality (pp. 3-25). New York: Praeger.

32.

Kent

S. L.

Jacobs

(2005). Minority threat and police strength from 1980 to 2000: A fixed-effects analysis of nonlinear and interactive effects in large U.S. cities. Criminology, 43, 731-759.

33.

Kovandzic

Sloan

Vieraitis

(2002). Unintended consequences of politically popular sentencing policy: The homicide promoting effects of “Three Strikes” in U.S. cities (1980-1990). Criminology and Public Policy,1, 399-424.

34.

Lane

(1980). Urban police and crime in nineteenth-century America. In Tonry

Morris

(Eds.), Crime and justice: An annual review of research (pp. 1-44). Chicago: University of Chicago Press.

35.

Lempert

(1966). Strategies of research design in the legal impact study: The control of plausible rival hypotheses. Law and Society Review,1, 111-132.

36.

Lindbloom

(1959). The science of muddling through. Public Administration Review, 21, 78-88.

37.

Liska

A. E.

Lawrence

J. J.

Benson

(1981). Perspectives on the legal order: The capacity for social control. American Journal of Sociology,87, 413-426.

38.

Loftin

McDowall

(1982). The police, crime, and economic theory: An assessment. American Sociological Review, 47, 393-401.

39.

Marvell

Moody

(2001). The lethal effects of three-strikes laws. Journal of Legal Studies,30, 89-106.

40.

Marvell

Moody

C. E.

(1995). The impact of enhanced prison terms for felonies committed with guns. Criminology,33, 247-282.

41.

Marvell

Moody

C. E.

(1996). Specification problems, police levels, and crime rates. Criminology, 34, 609-646.

42.

McDowall

Loftin

(1986). Fiscal politics and the police: Detroit, 1928-1976. Social Forces, 65, 162-176.

43.

Monkkonen

E. H.

(1981). Police in urban America, 1860-1920. Cambridge, UK: Cambridge University Press.

44.

Nalla

M. K.

(1992). Perspectives on the growth of police bureaucracies, 1948-1984: An examination of three perspectives. Policing and Society, 3, 51-61.

45.

Nalla

M. K.

Lynch

M. J.

Leiber

M. J.

(1997). Determinants of police growth in Phoenix, 1950-1988. Justice Quarterly, 14, 115-143.

46.

National Advisory Commission on Civil Disorders. (1968). Report. New York: Bantam.

47.

Pfeffer

Salancik

G. R.

(1978). The external control of organizations: A resource dependency perspective. New York: Harper and Row.

48.

Pfeffer

Salancik

G. R.

(2003). The external control of organizations: A resource dependency perspective (2nd ed.). Stanford, CA: Stanford University Press.

49.

Pindyck

R. S.

Rubinfeld

D. L.

(1998). Econometric models and economic forecasts (4th ed.). New York: McGraw-Hill/Irwin.

50.

Quinney

(1974). Critique of the legal order. Boston: Little, Brown.

51.

Quinney

(1975). Criminology. Boston: Little, Brown.

52.

Reiss

(1992). Police organizations in the twentieth century. In Tonry

Morris

(Eds.), Modern policing (pp. 51-97). Chicago: University of Chicago Press.

53.

Ross

Sawhill

(1975). Time of transition: The growth of families headed by women. Washington, DC: Urban Institute.

54.

Sever

(2001). The relationship between minority populations and police force strength: Expanding our knowledge. Police Quarterly,4, 28-68.

55.

Sharp

E. B.

(2006). Policing urban America: A new look at the politics of agency size. Social Science Quarterly, 87, 291-307.

56.

Shaw

C. R.

McKay

H. D.

(1942). Juvenile delinquency and urban areas. Chicago: Chicago University Press.

57.

Silver

(1966). The demand for order in civil society: A review of recent developments. In Bordua

(Ed.), The police: Six sociological essays (pp. 1-24). New York: John Wiley.

58.

Smith

D. A.

Jarjoura

G. R.

(1988). Social structure and criminal victimization. Journal of Crime and Delinquency, 26, 358-386.

59.

Stimson

(1985). Regression in space and time: A statistical essay. American Journal of Political Science,29, 914-947.

60.

Stucky

T. D.

(2005). Local politics and police strength. Justice Quarterly, 22, 139-169.

61.

Turk

(1969). Criminality and legal order. Chicago: Rand McNally.

62.

Tyler

T. R.

Boeckman

R. J.

(1997). Three strikes and you are out, but why? Law and Society Review, 31, 237-265.

63.

Uchida

C. D.

(1993). The development of the American police: An historical overview. In Dunham

R. G.

Alpert

G. P.

(Eds.), Critical issues in policing: Contemporary readings (2nd ed., pp. 16-32). Prospect Heights, IL: Waveland Press.

64.

U.S. Census Bureau. (1992). Government finance and employee classification manual. Retrieved May 11, 2009, from http://www.census.gov/govs/www/class_ch8.html.

65.

Walker

(1977). A critical history of police reform: The emergence of professionalism. Lexington, MA: Lexington Books.

66.

Walker

(1985). Sense and nonsense about crime: A policy guide. Monterey, CA: Brooks/Cole.

67.

Wilson

J. Q.

(1975). Thinking about crime. New York: Random House.

68.

Wooldridge

J. M.

(2002). Econometric analysis of cross section and panel data. Cambridge, MA: MIT Press.

69.

Worrall

J. L.

(2004). The effect of three-strikes legislation on serious crime in California. Journal of Criminal Justice,32, 283-296.

70.

Worrall

J. L

Kovandzic

T. V.

(2007). COPS grants and crime revisited. Criminology, 45, 159-190.

71.

Zhao

Scheider

M. C.

Thurman

(2002). Funding community policing to reduce crime: Have COPS grants made a difference? Criminology and Public Policy, 2, 7-32.