Community Policing Grants: COPS Grants Were a Modest Contributor 
to Declines in Crime in the 1990s (14-OCT-05, GAO-06-104).	 
                                                                 
Between 1994 and 2001, the Office of Community Oriented Policing 
Services (COPS) provided more than $7.6 billion in grants to	 
state and local communities to hire police officers and promote  
community policing as an effective strategy to prevent crime.	 
Studies of the impact of the grants on crime have been		 
inconclusive. GAO was asked to evaluate the effect of the COPS	 
program on the decline in crime during the 1990s. GAO developed  
and analyzed a database containing annual observations on crime, 
police officers, COPS funds, and other factors related to crime, 
covering years prior to and during the COPS program, or from 1990
through 2001. GAO analyzed survey data on policing practices that
agencies reportedly implemented and reviewed studies of policing 
practices. GAO assessed: (1) how COPS obligations were		 
distributed and how much was spent; (2) the extent to which COPS 
expenditures contributed to increases in the number of police	 
officers and declines in crime nationwide; and (3) the extent to 
which COPS grants during the 1990s were associated with policing 
practices that crime literature indicates could be effective. In 
commenting on a draft of this report, the COPS Office said that  
our findings are important and support conclusions reached by	 
others. 							 
-------------------------Indexing Terms------------------------- 
REPORTNUM:   GAO-06-104 					        
    ACCNO:   A39537						        
  TITLE:     Community Policing Grants: COPS Grants Were a Modest     
Contributor to Declines in Crime in the 1990s			 
     DATE:   10/14/2005 
  SUBJECT:   Budget obligations 				 
	     Cost effectiveness analysis			 
	     Crime prevention					 
	     Federal aid for criminal justice			 
	     Federal funds					 
	     Financial management				 
	     Funds management					 
	     Grant administration				 
	     Grants to local governments			 
	     Grants to states					 
	     Labor force					 
	     Law enforcement agencies				 
	     Police						 
	     Program evaluation 				 

******************************************************************
** This file contains an ASCII representation of the text of a  **
** GAO Product.                                                 **
**                                                              **
** No attempt has been made to display graphic images, although **
** figure captions are reproduced.  Tables are included, but    **
** may not resemble those in the printed version.               **
**                                                              **
** Please see the PDF (Portable Document Format) file, when     **
** available, for a complete electronic file of the printed     **
** document's contents.                                         **
**                                                              **
******************************************************************
GAO-06-104

     

     * Background
     * Results
     * Concluding Observations
     * Agency Comments and Our Evaluation
     * Appendix I: Objectives, Scope, and Methodology
     * Overview of Our Approach and Methodology
          * Methods Used to Address the Flow of Funds Reporting Objective
          * Methods Used to Address the Effects of COPS Expenditures on
            Officers and Crime
          * Methods to Assess Changes in Policing Practices
     * Database Construction and Samples Used in Our Analyses
          * Data Used in Our Analysis of Obligations and Expenditures
          * Data Used in Our Analysis of Officers and Crime
          * Data Used in Our Analysis of Reported Changes in Policing
            Practices
     * Reliability and Validity of the Data That We Used
     * Appendix II: Background on the COPS Program and Studies of the Impacts
       of COPS Grants on Crime
     * COPS and Other Local Law Enforcement Grants Distributed throughout the
       1990s
     * Debates over whether the COPS Office Met Its Goals for Officers and
       Promoted Community Policing
     * Debates about COPS' Contribution to the Decline i
     * Issues in Assessing the Contribution of COPS Grants to the Decline in
       Crime in the 1990s
     * Appendix III: COPS Grant Obligation and Expenditure Patterns
     * Smaller Agencies Received Larger Amounts of COPS Obligations per Crime
       than Did Larger Ones
          * Most Agencies Had Received Their First COPS Grant by 1996
     * Total COPS Expenditures and Per Capita Expenditures Peaked in 2000,
       and Smaller Agencies Spent More than Larger Ones on a Per Capita Basis
     * COPS Expenditures Amounted to about 1 Percent of All Local Law
       Enforcement Expenditures
     * Appendix IV: COPS Expenditures Led to Increases in Sworn Officers and
       Declines in Crime
     * COPS Expenditures Led to Increases in Sworn Officers above Levels That
       Would Have Been Expected without Them and Were Responsible for about
       88,000 Officer-Years
          * LLEBG Funds Also Contributed to Increases in Officer Strength
     * COPS Expenditures Led to Reductions in Crime through Increases in
       Officers
          * Various Specifications of Our Regressions Yielded Consistent
            Findings about the Effect of COPS Expenditures on Crime
          * Factors other than COPS Expenditures Contributed Larger Amounts
            to the Reduction in Crimes, but COPS Contribution Was in Line
            with COPS Expenditures
     * Appendix V: COPS Expenditures Associated with Policing Practices That
       Crime Literature Indicates Are Effective in Preventing Crime
     * Comparisons of Pre- and Within-COPS Grant Program Levels of Reported
       Policing Practices Show That COPS Grantee Agencies Reported Larger
       Increases than Non-COPS Agencies
          * The Effects of COPS Grants on Agencies' Reported
     * Reported Levels of Policing Practices among COPS Grantees Did Not
       Increase Overall from 1996 to 2000
     * Crime Literature Provides Evidence for Effectiveness of Some Policing
       Practices
     * Appendix VI: Methods Used to Estimate the Effects of COPS Funds on
       Officers and Crime
     * Prior Literature on the Relationship between Officers and Crime
       Addresses Issues Relating to Estimating the Effects of COPS Funds on
       Crime
     * Our Approach to Estimating the Effects of COPS Expenditures on
       Officers and Crime
          * Model of the Effect of COPS Expenditures on the Number of Police
            Officers
          * Model of the Effect of COPS Expenditures on Crime
          * The Implied Relationship between Police Officers and Crime
     * Data Used in Our Analysis
     * Explanation of the Results of Our Analysis
          * The Effect of COPS Expenditures on the Number of Police Officers
          * Effect of COPS Expenditures on Crime
          * The Effects of Different Population Sizes across Agencies
          * Calculations of the Elasticity of Crime with Respect to Officers
          * Estimating the Net Number of Officers Paid for by COPS
            Expenditures
          * Estimating the Number of Crimes Reduced by COPS Expenditures
     * Appendix VII: Methods Used to Assess Policing Practices
     * Methods to Address Changes in Policing Practices
          * Characteristics and Analysis of the Policing Strategies Survey
          * Characteristics and Analysis of the National Evaluation of COPS
            Survey
     * Methods to Review Policing Practices
          * How We Selected Studies
          * How We Reviewed Studies
          * The Research Literature Shows That Some Policing Practices May be
            Effective in Reducing Crime
     * Appendix VIII: Comments from the Department of Justice
     * Appendix IX: GAO Contacts and Staff Acknowledgments
     * GAO Contacts
     * Acknowledgments
     * Bibliography
          * Order by Mail or Phone

                 United States Government Accountability Office

GAO

                    the Judiciary, House of Representatives

October 2005

COMMUNITY POLICING GRANTS

    COPS Grants Were a Modest Contributor to Declines in Crime in the 1990s

GAO-06-104

COMMUNITY POLICING GRANTS

COPS Grants Were a Modest Contributor to Declines in Crime in the 1990s

  What GAO Found

About half of the COPS funds distributed from 1994 through 2001 went to
law enforcement agencies in localities of fewer than 150,000 persons and
the remainder to agencies in larger communities. This distribution roughly
corresponded to the distribution of major property crimes but less so to
the distribution of violent crimes. For example, agencies in larger
communities received about 47 percent of COPS funds but accounted for 58
percent of the violent crimes nationwide. From 1994 through 2001, COPS
expenditures constituted about 1 percent of total local expenditures for
police services.

For the years 1994 through 2001, expenditures of COPS grants by grant
recipients resulted in varying amounts of additional officers above the
levels that would have been expected without the expenditures. For
example, during 2000, the peak year of COPS expenditures by grant
recipients, they led to an increase of about 3 percent in the level of
sworn officers-or about 17,000 officers. Adding up the number of
additional officers in each year from 1994 through 2001, GAO estimated
that COPS expenditures yielded about 88,000 additional officer-years. GAO
obtained its results from fixed-effects regression models that controlled
for pre-1994 trends in the growth rate of officers, other federal
expenditures, and local- and state-level factors that could affect officer
levels.

From its analysis of the effects of increases in officers on declines in
crime, GAO estimated that COPS funds contributed to declines in the crime
rate that, while modest in size, varied over time and among categories of
crime. For example, between 1993 and 2000, COPS funds contributed to a 1.3
percent decline in the overall crime rate and a 2.5 percent decline in the
violent crime rate from the 1993 levels. The effects of COPS funds on
crime held when GAO controlled for other crime-related factors-such as
local economic conditions and state-level policy changes-in its regression
models, and the effects were commensurate with COPS funds' contribution to
local spending on police protection. Factors other than COPS funds
accounted for the majority of the decline in crime during this period. For
example, between 1993 and 2000, the overall crime rate declined by 26
percent, and the 1.3 percent decline due to COPS, amounted to about 5
percent of the overall decline. Similarly, COPS contributed about 7
percent of the 32 percent decline in violent crime from 1993 to 2000.

From 1993 though 1997, agencies that received and spent COPS grants
reported larger changes in policing practices and in the subsets of
practices that focus on solving crime problems or focus on places where
crime is concentrated than did agencies that did not receive the grants.
The differences held after GAO controlled for underlying trends in the
reported use of these policing practices. From 1996 to 2000, there was no
overall increase in policing practices associated with COPS grants. In its
review of studies on policing practices, GAO found that problem-solving
and place-oriented practices can be effective in reducing crime.

                 United States Government Accountability Office

                                    Contents

Letter                                                                   1
                    Background                                              5
                    Results                                                11
                    Concluding Observations                                16
                    Agency Comments and Our Evaluation                     17
Appendix I       Objectives, Scope, and Methodology                     20
Appendix II           Background on the COPS Program and Studies of 
                    the Impacts of COPS Grants on Crime                    39
Appendix III         COPS Grant Obligation and Expenditure Patterns     50
Appendix IV      COPS Expenditures Led to Increases in Sworn        
                    Officers and Declines in Crime                         58
Appendix V       COPS Expenditures Associated with Policing         
                          Practices That Crime Literature Indicates Are
                    Effective in Preventing Crime                          66 
Appendix VI        Methods Used to Estimate the Effects of COPS Funds 
                    on Officers and Crime                                  72 
Appendix VII          Methods Used to Assess Policing Practices         94 
Appendix VIII          Comments from the Department of Justice         108 

Appendix IX GAO Contacts and Staff Acknowledgments

  Bibliography

Tables 
           Table 1: Index of Crimes, 2002, as Reported by the FBI, Excluding
          Arson                                                            21 
            Table 2: COPS Obligations, 1994 through 2001, by COPS Grant    
                       Categories and Types of Grant Programs              24 
             Table 3: Law Enforcement Agencies Reporting to the UCR and in 
                                Our Analysis Dataset                       33 
              Table 4: COPS Grant Obligations 1994-2001, by COPS Grant     
          Program                                                          50 
            Table 5: Percentage Distribution of COPS Obligations and Crime 
                          from 1994 through 2001, by Population Size Group 52 
             Table 6: Per Crime COPS Obligations, by Population Size Group 
                      and Category of Crime, 1994 through 2001             52 
               Table 7: Number of Agencies That Received at Least One COPS 
                   Grant Obligation, 1994-2001, by COPS Grant Program, and 
                           Year of First COPS Obligation                   53 
          Table 8: Percentage of Agencies in GAO's Primary Analysis Sample 
                     That Received at Least One COPS Grant Obligation from 
                        1994 through 2001, by Size of Population Served by 
          Agencies                                                         54 
           Table 9: Estimated Effect of COPS Expenditures on the Number of 
                         Sworn Officers Nationwide in Each Year, 1994-2001 59 
            Table 10: Estimated Percentage Change in Crime Rates from 1993 
                      Levels Due to COPS Expenditures, 1994-2001, by Crime 
          Type Category                                                    61 
          Table 11: Mean Levels of Policing Practices in 1993 and 1997, by 
                       Category of Policing Practices and whether Agencies 
                    Received a COPS Grant between 1994 and 1997            67 
          Table 12: Mean Levels of Policing Practices in 1993 and 1997, by 
                       Size of Agency and whether Agencies Received a COPS 
                            Grant between 1994 and 1997                    68 
                  Table 13: Difference in Mean Levels of Reported Policing 
                                                                 Practices 
                   in 1996 and 2000, by Category of Policing Practices and 
                         Timing of COPS Grant Expenditures                 70 

Table 14: Alternate Specifications of the Relationship between COPS
Expenditures and Crime 79 Table 15: Means and Standard Deviations of
Variables Used in Regression Models 81

Table 16: Parameter Estimates from Regressions of Officers Per Capita on
COPS Hiring Grant Expenditures and Other Outside Funds (Standard Errors in
Parentheses) 83

Table 17: Parameter Estimates from Regressions of Crime Rates on COPS
Hiring Grant Expenditures and Other Outside Funds (Standard Errors in
Parentheses) 84

Table 18: Parameter Estimates from Regressions of Index Crime Rates and
Officers Per Capita on COPS Hiring Grant Expenditures and Other Outside
Funds, by Population Size Category (Standard Errors in Parentheses) 85

Table 19: Elasticities of the Impact of Police Officers on the Crime Rate
86

Table 20: Elasticity of Violent and Property Crime with Respect to
Officers under Alternate Specifications of the Relationship between COPS
Expenditures and Crime 88

Table 21: Estimated Per Capita Effect of COPS Expenditures on the Number
of Officers 89 Table 22: Estimated Per Capita Growth of COPS Expenditures
on Police Officers and Crime from 1993 92 Table 23: Categories of Policing
Practices and Specific Items within Each Category in the Policing
Strategies Survey 98

Table 24: Parameter Estimates from Regressions of Changes in Mean Number
of Policing Practices and Category of Practices between 1993 and 1997 on
whether or Not Agencies Received COPS grant between 1994 and 1997 and on
Per Capita COPS Expenditures between 1994 and 1997 (Standard Errors in
Parentheses) 100

Table 25: Categories of Policing Practices and Specific Items within Each
Category in the National Evaluation of COPS Survey 103

Table 26: Parameter Estimates from Regressions of Changes in Mean Number
of Policing Practices and Category of Practices between 1996 and 2000 on
Whether or Not Agencies Received COPS grant Between 1997 and 2000 and on
Per Capita COPS Expenditures between 1994-1996 and 1997-2000 (Standard
Errors in Parentheses) 105

Figures 
             Figure 1: Total, Violent, and Property Crime Rates per 100,000
                      Persons, as Reported in the Uniform Crime Reports, 1970
           2001                                                            10 
             Figure 2: Estimated Effects of COPS Grant Expenditures on the 
                         Number of Sworn Officers, 1991-2001               13 
             Figure 3: Annual Percentage Changes in the Violent Crime Rate 
                       from 1993: Total Change and Estimated Change Due to 
           COPS Grants                                                     14 
               Figure 4: Reported Levels of Policing Practices in 1993 and 
                                                                   1997 in 
                     Agencies That Did and Did Not Receive COPS Grants, by 
                            Category of Policing Practice                  16 
               Figure 5: Violent Crimes and Violent Crimes Reported to the 
                                                                   Police, 
                 as Reported in the National Criminal Victimization Survey 
                            and Including Homicides from the Uniform Crime 
                                 Reports, 1990-2001                        36 
            Figure 6: Total Index, Violent, and Property Crime Rates per   
                             100,000 Persons, 1990-2001                    46 
             Figure 7: Annual Expenditures of COPS Grant Funds, by Year    55 
              Figure 8: Number of Agencies That Spent COPS Funds, 1994     
           through 2001                                                    56 
               Figure 9: Reported Levels of Policing Practices in 1993 and 
                                                                   1997 in 
                           Agencies That Received and Did Not Receive COPS 
                        Grants, by Size of Population Served               69 

Abbreviations

AHEAD           Accelerated Hiring, Education, and Deployment              
BJA             Bureau of Justice Assistance                               
BJS             Bureau of Justice Statistics                               
COPS            Office of Community Oriented Policing Services             
DLEA            Directory of Law Enforcement Agency                        
DOJ             Department of Justice                                      
FAST            Funding Accelerated for Smaller Towns                      
FBI             Federal Bureau of Investigation                            
FIPS            Federal Information Processing Standards                   
LLEBG           Local Law Enforcement Block Grants                         
MORE            Making Officer Redeployment Effective                      
NCHS            National Center for Health Statistics                      
NVCS            National Crime Victimization Survey                        
OJP             Office of Justice Programs                                 
ORI             originating agency identifier                              
UCR             Uniform Crime Reporting                                    
UHP             Universal Hiring Program                                   
VCCLEA               Violent Crime Control and Law Enforcement Act of 1994 

This is a work of the U.S. government and is not subject to copyright
protection in the United States. It may be reproduced and distributed in
its entirety without further permission from GAO. However, because this
work may contain copyrighted images or other material, permission from the
copyright holder may be necessary if you wish to reproduce this material
separately.

United States Government Accountability Office Washington, DC 20548

October 14, 2005

The Honorable F. James Sensenbrenner, Jr. Chairman Committee on the
Judiciary House of Representatives

Dear Mr. Chairman:

Provisions of the Public Safety Partnership and Community Policing Act of
1994, Title 1 of the Violent Crime Control and Law Enforcement Act of 1994
(VCCLEA), authorized appropriations of $8.8 billion for fiscal years 1995
through 2000 for grants to states and local communities to increase the
hiring and deployment of community police officers and to promote
nationwide the concept of community policing-an approach to policing that
involves the cooperation of law enforcement and community residents in
identifying and developing solutions to crime problems-as an effective
strategy to improve public safety. 1 To administer the grants, in October
1994, the Attorney General created the Office of Community Oriented
Policing Services (COPS).

According to its Director, the COPS Office was responsible for "one of the
greatest infusions of resources into local law enforcement in our nation's
history," 2 and in a report to Congress the former Attorney General linked
increases in COPS-funded officers to declines in crime. By the summer of
2000, the former Attorney General reported, the COPS Office had awarded
more than $7.6 billion in grants to more than 12,000 local law enforcement
agencies-primarily municipal, state, and county police and sheriff's
departments whose officers have general arrest powers-and funded over
105,000 community policing officers. The report claimed that the funded
officers led to an increase in the number of police officers that was well
above what would have been expected in the absence of the passage of

1

P.L. 103-322 (1994), 42 U.S.C. S: 3796dd. The act contained other
provisions to address violent crime, such as those encouraging states to
increase the use of incarceration for violent offenders, enhancing
penalties for gang crimes, and expanding the number of federal offenses
punishable by death.

2

Frazier, Thomas, C., "Introduction from the Director," in Attorney
General, Report to Congress: Office of Community Oriented Policing
Services, Washington, D.C.: U.S. Department of Justice, 2000.

VCCLEA, and it cited research that showed that increased police presence
led to reductions in crime. As evidence that these officers led to
reductions in crime, the report showed that the average number of violent
crimes per police department declined as the number of COPS-funded
officers increased.

A study funded by the COPS Office and released in 2001, which attempted to
control for some of the other factors that could influence crime rates and
also be correlated with the distribution of COPS funds, concluded that
COPS grants contributed to the reduction in crime in the 1990s. 3 You
previously asked us to review this study, and we reported that its
methodological limitations were such that the study's results should be
viewed as inconclusive. 4

In response to our assessment of the results of the study that we
reviewed, you asked us to undertake an independent evaluation of the
impact of COPS grants on the decline in crime that occurred during the
1990s. This report provides results from our evaluation of interrelated
questions about COPS funds, officers, crime, and policing practices.
Specifically, regarding COPS funds: (1) From 1994 through 2001, how were
COPS obligations distributed among local law enforcement agencies in
relation to the populations they served and crimes in their jurisdictions,
and how much of the obligated amounts did agencies spend? Regarding
officers and crime:

(2) To what extent did COPS grants contribute to increases in the number
of sworn officers and declines in crime in the nation during the 1990s?
Regarding policing practices: (3) To what extent were COPS grants during
the 1990s associated with police departments adopting policing practices
that the crime literature indicates could contribute to reductions in
crime?

To address our reporting objectives, we created and analyzed a database
consisting of 12 years of data on local law enforcement agencies for the
years 1990 through 2001. We included data from the Department of Justice's
(DOJ) Office of Justice Programs (OJP) on federal law enforcement grant
obligations and expenditures (including COPS grants);

3

Zhao, J., and Q., Thurman. A National Evaluation of the Effect of COPS
Grants on Crime from 1994 to 1999. Report submitted to the Office of
Community Oriented Policing Services, Washington, D.C.: U.S. Department of
Justice, December 2001.

4

We reported our review of this study in GAO, Technical Assessment of Zhao
and Thurman's 2001 Evaluation of the Effects of COPS Grants on Crime,
GAO-03-867R (Washington, D.C.: June 13, 2003).

Page 2 GAO-06-104 Community Policing Grants

the Federal Bureau of Investigation's (FBI) Uniform Crime Reporting (UCR)
Program on crime and sworn officers; the Department of Commerce's Bureau
of Economic Analysis on local economic conditions such as employment rates
and per capita income; and the National Center for Health Statistics and
U.S. Census Bureau's estimates of demographic characteristics of local
populations-such as their age, race, and gender composition. The UCR crime
data that we used are data on crimes reported to or known by the police
and reported to the UCR Program. The crimes in the UCR are based on the
FBI's crime index. The index crimes include the violent crimes of murder
and non-negligent manslaughter, forcible rape, robbery, and aggravated
assault as well as the property crimes of burglary, larceny-theft, motor
vehicle theft, and arson. The FBI reports that there is limited reporting
of arson offenses to the UCR Program by law enforcement agencies. We
therefore excluded arson crimes from our analysis. 5

Prior to developing and using the database to address our objectives, we
assessed the reliability of each data source, and in preparing this
report, we used only the data that we found to be sufficiently reliable
for the purposes of our report. We also assessed possible biases in our
estimates of the effects of COPS funds on crime arising from our use of
UCR data on reported crimes. We concluded from our analysis that our
estimates of the impacts of COPS funds are likely to understate the effect
of COPS funds on crime. (See app. I for a more detailed discussion of our
approach, methods, and database construction.)

To describe how COPS grant funds were distributed and spent, we analyzed
data on COPS obligations to and expenditures by local law enforcement
agencies, comparing them with several characteristics of the agencies that
received COPS funds, such as population size and crime levels.

To assess the possible relationships between COPS expenditures and changes
in the number of officers and rates of crime, we analyzed data on the
agencies that reported complete crime and officer data for at least 1 year
from 1990 through 2001 using a two-stage regression model of these
relationships. In the first stage, we estimated the relationship between
the

Because of the limited reporting of arson, the FBI also excludes arson
from its published tables containing national estimates of index crimes.
See Federal Bureau of Investigation, Crime in the United States, Uniform
Crime Reports, Washington, D.C.: U.S. Department of Justice, published
annually.

Page 3 GAO-06-104 Community Policing Grants

variation in the timing and amount of agencies' expenditures on COPS
grants that were for hiring officers and changes in the number of
officers. In the second stage, we estimated the relationship between
changes in COPS expenditures and changes in crime rates using
fixed-effects regression models. We used the results from these two sets
of regressions to calculate the amount of the change in crime (from the
second stage) due to changes in officers (from the first stage). As the
relationship between officer levels and crime rates may reflect a complex
causal relationship, we used COPS hiring grants as a statistical
instrument to help to identify the relationship between officers and
crime. In both sets of regression models, we used agency and year fixed
effects to control for unobserved sources of variation among agencies
(within a given year) and within agencies (over time). We also included
variables to measure agencies' pre-1994 trends in the growth of crime
rates and officers. These controls allowed us to compare agencies that had
similar, pre-COPS trends in these variables, thereby reducing further the
differences among agencies that are not due to COPS expenditures. To
control for economic factors that may be related to crime-such as
employment and income- we included measures of local economic conditions,
and to control for changes in the composition of local populations that
could be correlated with crime, we included measures of age and race
composition of local populations. Finally, to control for changes in
state-level practices that could affect crime rates, such as changes in
state incarceration rates or state sentencing policy, we included
state-by-year fixed effects in our regression models. (See app. VI for
additional details about our regression models.)

To address the relationship between COPS grants and changes in policing
practices, we analyzed data from two surveys of nationally representative
samples of local law enforcement agencies on policing practices that they
reportedly implemented in various years from 1993 to 2000. The first
survey-the Policing Strategies Survey-was administered in 1993 and 1997 to
provide information on the development and implementation of community
policing. 6 The second survey-the National Evaluation of

The first survey was the National Survey of Community Policing Strategies,
and it was administered in 1993 and 1997. The Police Foundation
administered the 1993 wave of the survey, and ORC Macro International,
Inc. and the Police Executive Research Forum administered the 1997 wave of
the survey. Both surveys used the same sampling frame. In the remainder of
this report, we refer to the two waves of this longitudinal survey as the
Policing Strategies Survey.

Page 4 GAO-06-104 Community Policing Grants

COPS Survey-was administered in 1996 and 2000. 7 We identified and
analyzed practices that are associated with problem-solving,
place-oriented approaches to policing, community collaboration efforts,
and the use of crime analysis. We assessed changes in the levels of
reported practices between agencies that spent COPS grants over particular
periods with those that did not receive or spend COPS grant funds. To
control for the underlying trends in reported policing practices, we
estimated fixed-effects regression models of the effects of COPS grants
expenditures on changes in reported policing practices. To assess the
possible relationship between policing practices and crime, we analyzed
systematic reviews of the effectiveness of policing practices in reducing
crime to identify the types of policing practices that have been judged to
be effective in preventing crime. (See app. VII for details about the
surveys and our analytic methods.)

In addition, we reviewed relevant economic and criminological literatures
that addressed issues related to estimating models of the effects of
federal grant funds on crime rates. We spoke with officials at the
Department of Justice about the operation of the COPS program, and we also
spoke with researchers about our approach and methods. We reviewed our
approach and methods with a group of experts in the field of policing and
crime. The group consisted of criminologists, economists, statisticians,
and criminal justice practitioners, and was convened for us by the
National Research Council of the National Academies to enable participants
to offer their individual views as experts in the field.

We conducted our work between January 2004 and August 2005 in accordance
with generally accepted government auditing standards.

Established in October 1994 by the Attorney General to implement the

administration of community policing grants under VCCLEA, the Office of
Community Oriented Policing Services announced it first grant programs in
1994. Prior to its establishment, in December 1993 the Department of

7

The second survey was the National Evaluation of the COPS Program survey,
which was conducted by the National Opinion Research Corporation for the
Urban Institute in its evaluation of the implementation of the COPS
program. It was a nationally representative sample of law enforcement
agencies that were contacted in 1996 and again in 2000. In the remainder
of this letter, we refer to this second survey as the National Evaluation
of COPS Survey.

Page 5 GAO-06-104 Community Policing Grants

Justice awarded community policing grants to hire officers under the
Police Hiring Supplement. 8

The COPS Office distributed grants in a variety of program funding
categories. Hiring grants, which required agencies to hire new officers
and at the same time to indicate the types of community policing
strategies that they intended to implement with the grants, was the
largest COPS grant program category in terms of the amounts of grant funds
obligated. 9 The hiring grants paid a maximum of $75,000 per officer over
a 3-year period (or at most 75 percent of an officer's salary) and
generally required that local agencies cover the remaining salary and
benefits with state or local funds. From 1994 through 2001, more than $4.8
billion in COPS obligations (or about 64 percent of COPS obligations over
this period) were in the form of hiring grants. The Making Officer
Redeployment Effective (MORE) grant program, which provided funds to law
enforcement agencies to purchase equipment, hire civilians, and redeploy
existing officers to community policing was the second largest COPS grant
program, obligating more than $1.2 billion. Additional COPS grant programs
provided funds for specific innovations in policing and for a variety of
other purposes.

Each year the COPS Office was required to distribute half of the grant
funds to agencies in communities whose populations exceeded 150,000

8

The Police Hiring Supplement Program was established by the Supplemental
Appropriations Act of 1993 (P.L. 103-50 (1993)). The grants made under
this program were funded by DOJ's Bureau of Justice Assistance. In this
report, when we refer to COPS grants, we include both the grants made
under the Police Hiring Supplement and the community policing grants
authorized under VCCLEA.

9

Hiring programs authorized under VCCLEA and administered by the COPS
office included the Phase I program, which funded qualified applicants who
had applied for the Police Hiring Supplement but were denied because of
the limited funds available; COPS AHEAD (Accelerated Hiring, Education,
and Deployment) for municipalities with populations of 50,000 and above;
and COPS FAST (Funding Accelerated for Smaller Towns) for towns with
populations below 50,000. In June 1995, Phase I, COPS AHEAD, and COPS FAST
were replaced by the Universal Hiring Program.

persons and half of the grant funds to agencies in communities with
populations of 150,000 or fewer persons. 10

In the applications for hiring grants, the COPS Office requested agencies
to indicate the types of community policing practices that they planned to
implement with their grants. The various practices related to community
policing included practices such as identifying crime problems by looking
at records of crime trends and analyzing repeat calls for service, working
with other public agencies to solve disorder problems, locating offices or
stations within neighborhoods, and collaborating with community residents
by increasing officer contact with citizens and improving citizen
feedback. These types of policing practices also corresponded with general
approaches to policing. For example, problem-solving policing practices
may rely on crime analysis tools to help to identify crime problems and
develop solutions to them. Place-oriented practices attempt to identify
locations where crime occurs repeatedly and to implement procedures to
disrupt these recurrences of crime. By collaborating with community
residents, agencies attempt to improve citizen feedback about crime
problems and effectiveness of policing to address these problems.

In 2000, DOJ reported that COPS-funded officers helped to reduce crime and
reported that the drop in crime that occurred after 1994 was more than
what would have been expected in the absence of the passage of VCCLEA and
the creation of the COPS Office. 11 The report suggested that COPS had
achieved its goal of funding 100,000 officers, and through increases in
officers and the practice of community policing, the COPS program was
credited with reducing crime. However, while COPS may have funded 100,000
officers, it was not apparent that the funded officers had resulted in new
officers having been hired. Researchers at the Urban Institute reported in
2000, for example, their estimates that by 2003, the COPS program would
have raised the level of police on the street by the

10

Of funds available in any fiscal year, up to 3 percent could have been
used for technical assistance or for evaluations or studies carried out or
commissioned by the Attorney General. The requirement to allocate the
funds by size of agency population applies to the remaining funds in any
fiscal year (42 U.S.C. S: 3793 (a)(11)(B)). In addition, COPS had to meet
a national coverage requirement to ensure that no state received less than
0.5 percent of total funding.

11

Attorney General, Report to Congress: Office of Community Oriented
Policing Services, Washington, D.C.: U.S. Department of Justice, 2000.

equivalent of 62,700 to 83,900 full-time equivalent officers. 12 They also
indicated that it was unclear whether the program would ever increase the
number of officers on the street at a single time by 100,000. 13

The COPS Office-funded study of the effect of COPS grants on crime in over
6,000 communities from 1995 through 1999 that had received COPS grants
concluded that COPS grants were effective in reducing crime. 14 The study
also reported that COPS grants that encouraged agencies to implement a
variety of innovative strategies to improve public safety had larger
impacts on reducing violent and property crime than did other COPS grant
types. 15 However, a study released by the Heritage Foundation, which was
based upon an analysis of county-level data, was unable to replicate the
findings of the COPS Office-funded study. 16 Specifically, the Heritage
study found no effect of COPS hiring grants on crime rates, but it did
find that the COPS grants for specific problems-such as gangs, domestic
violence, and illegal use of firearms by youth-were associated with
reductions in crime. In addition, we questioned whether the sizes of the
effects of COPS grants on crime that were reported in the COPS
Office-funded study were large enough to be significant in a practical
sense and whether they could accurately represent the expected returns on
the investment of billions of dollars. 17

12

Roth, Jeffrey, et al., National Evaluation of the Implementation of the
COPS Program, Washington, D.C.: National Institute of Justice, August
2000.

13

In a 2002 report, the Urban Institute researchers updated their estimates
of the number of officers due to COPS and reported an estimate of a
permanent increase of between 69,100 and 92,200 officers post-2005, taking
into account post-grant attrition of officers. Koper, Christopher, et al.,
Putting 100,000 Officers on the Street: A Survey-Based Assessment of the
Federal COPS Program, Washington, D.C.: The Urban Institute, 2002.

14

Zhao and Thurman, 2001.

15

The authors of the COPS Office-funded study revised their report to take
into account criticism presented by reviews, and in 2004, they released
their final report on the effect of COPS grants on crime. In their final
report, they updated their findings through 2000, and their results were
comparable to what they reported in their initial report. Zhao, J., and Q.
Thurman, Funding Community Policing to Reduce Crime: Have COPS Grants Made
a Difference from 1994 to 2000? Report submitted to the Office of
Community Oriented Policing Services, U.S. Department of Justice, July
2004.

16

Muhlhausen, D., Do Community Oriented Policing Services Grants Affect
Violent Crime Rates (Washington, D.C.: The Heritage Foundation, May 25,
2001).

17GAO Technical Assessment of Zhao and Thurman's 2001 Evaluation of the
Effects of COPS Grants on Crime, GAO-03-867R (Washington, D.C.: June 13,
2003).

Assessing the contribution of COPS funds to the decline in crime during
the 1990s presents challenges for evaluators. Nationwide, crime rates
began to decline in about 1991, before the COPS program announced its
first grant programs in 1994 (fig. 1). Hence the factors other than COPS
grants that were responsible for precipitating the decline in crime could
have continued to influence its decline throughout the 1990s. Researchers
have pointed to a number of factors that could have precipitated the
decline in crime, including increased use of prison as a punishment for
violent crimes, improved economic conditions, and the subsiding of
violence that accompanied the expansion of drug markets. To the extent
that any of these factors are correlated with the distribution of COPS
grants, they could be responsible for impacts that have been attributed to
COPS grants.

Figure 1: Total, Violent, and Property Crime Rates per 100,000 Persons, as
                Reported in the Uniform Crime Reports, 1970-2001

COPS grant period

1970 1975 Calendar year

1980 1985 1990 1994 2001

Index crime rate

Property crime rate

Source: GAO analysis of Uniform Crime Report data as reported on the
Bureau of Just ice Statistics online analysis of Uniform Crime Report
data. Prepared by the National Archive of Criminal Justice Data. Data
available at http://.bjsdata.ojp.usdoj.gov/dataonline/

Prior studies of the impact of COPS grants on crime have correlated COPS
funds with crime rates, controlling for other factors that could influence
crime rates. The authors of the prior studies describe various mechanisms
by which COPS grants may affect crime, but their statistical models do not
explicitly take these mechanisms into account in estimating the effects of
the grants. By identifying and explicitly modeling mechanisms through
which COPS funds could affect crimes-such as increasing the number of
sworn officers on the street who are available for patrolling places or
contributing to changes in policing practices that may be effective in
preventing crime-the possibility of a spurious relationship between inputs
(such as COPS funds) and outcomes (such as crime) can be minimized. (For
additional background information, see app. II.)

Results

Our analysis showed that from 1994 through 2001, COPS obligated more than
$7.32 billion to 10,680 agencies for which we were able to link Office of
Justice Programs financial data on COPS obligations to the records of law
enforcement agencies. 18 About $4.7 billion (or 64 percent) of these
obligations were in the form of hiring grants. About half of these funds
went to agencies serving populations of 150,000 or fewer persons and about
half was distributed to agencies serving populations of more than 150,000
persons. This distribution roughly corresponds to the distribution of
index crimes across the two size categories of jurisdictions. However, in
relation to violent crimes, the share of COPS funds distributed to larger
jurisdictions was smaller than the share of violent crimes that they
contributed to the national total. For example, agencies serving
populations of more than 150,000 persons contributed about 58 percent of
all violent crimes reported to the UCR during this time period while
receiving about half of all COPS funds. To be specific, the smallest
agencies-those serving populations of fewer than 10,000 persons- received
an average of $1,573 per violent crime reported to UCR. Agencies serving
populations of more than 150,000 persons received about $418 in COPS funds
per violent crime.

By the end of 2001, the COPS grantee agencies in our sample had spent
about $5 billion (or 68 percent of the $7.3 billion obligated to them)
from 1994 through 2001. Annually, the total amount of COPS expenditures
made by grantees increased each year from 1994 until 2000, and then
declined, while the number of agencies that drew down COPS funds peaked in
1998 at about 7,600 and declined to about 6,000 in 2001. From 1994 through
2001, a total of about 10,300 agencies spent COPS funds. The maximum
number of agencies spending funds in any given year occurred during 1998,
when about 7,600 agencies spent funds. From 1998 through 2000, the amount
of COPS expenditures per person in the jurisdiction served by an

18

The amount obligated to these agencies was 96.1 percent of the $7.6
billion total in COPS obligations reported in the Office of Justice
Programs financial data.

Page 11 GAO-06-104 Community Policing Grants

agency increased from about $4 to about $4.80. COPS expenditures amounted
to an annual average of about 1 percent of total expenditures for police
services by local law enforcement agencies from 1994 through 2001. This
contribution varied by year. For example, in 1999 and 2000, COPS
expenditures were about 1.5 percent of total local police expenditures.
(See app. III for a further discussion of COPS obligations and
expenditures.)

For the years 1994 through 2001, we infer from our estimates that COPS
hiring grant expenditures contributed to increases in sworn officer levels
above the levels that would have been expected without these funds. The
additional number of sworn officers stemming from these funds varied over
the years, and it increased from 1994 though 2000 and declined in 2001
(fig. 2). For example, for 1997 we estimate that COPS funds contributed
about 14,000 additional officers in that year-or about 2.4 percent of the
total number of sworn officers nationwide-and for 2000, COPS funds
contributed about 17,000 additional officers-or about 3 percent of the
total number of sworn officers nationwide. For all years from 1994 through
2001, we estimate that COPS expenditures paid for a total of about 88,000
additional officer-years over this entire period, where the total number
of officer-years equals the sum of the number of officers due to COPS
grant expenditures in each year. An officer-year refers to the number of
officers in a given year that we could attribute to COPS expenditures, and
the additional officers in a given year attributable to COPS expenditures
represent a net addition to the stock of sworn officers. 19 Using the
results from our regression estimates of the effects of COPS expenditures
on the level of sworn officers, we set the values for COPS expenditures to
zero to predict the level of officers absent COPS funds. The difference
between this number and the actual number of sworn officers yields the
number of officers due to COPS expenditures. Our analysis also shows that
apart from the COPS hiring and COPS MORE grants, other COPS grant types
did not have a significant effect on officer strength. (See app. IV for
more detailed information about the results of our analysis of COPS
expenditures on officers.)

An officer-year is not equivalent to the total number of officers or
full-time officer equivalents hired as a result of COPS funds; nor is it
equivalent to the total number of officers funded by COPS grants. Across
years, the COPS funds may have paid for the same person. In counting
officer-years, this person would be counted one time for each year in
which we estimated that COPS funds paid for the position.

Page 12 GAO-06-104 Community Policing Grants

Figure 2: Estimated Effects of COPS Grant Expenditures on the Number of
Sworn Officers, 1991-2001

Pre-COPS program period COPS program period

Estimated number of officers 620,000

600,000

580,000

560,000

540,000

520,000

500,000

480,000 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001

    Calendar year

Actual number of sworn officers

Predicted number of sworn officers per capita in the absence of COPS grant
expenditures (if COPS grant expenditures were equal to zero)

Source: GAO analysis of Uniform Crime Report, Office of Justice Programs,
National Center for Health Statistics, U.S. Census Bureau, and Bureau of
Economic Analysis data.

We estimate that the COPS grant expenditures contributed to the reduction
in crime in the 1990s independently of other factors that we were able to
take into account in our analysis. Other factors that could have
contributed to the reduction in crimes in the 1990s that we took into
account included federal law enforcement expenditures other than COPS
grants, local economic conditions and changes in population composition,
and changes in state-level policies and practices that could be correlated
with crime, such as incarceration and sentencing policy. Specifically,
from our model of the effect of changes in sworn officers on crime, we
estimate that a 1 percent increase in the number of sworn officers per
capita would lead to a 0.4 percent reduction in the total number of index
crimes. Through their effects on changes in officers in a given year, COPS
expenditures led to varying amounts of declines in crime rates over the
years from 1994 through 2001. For example, the 2.4 percent increase in
sworn officers due to COPS expenditures in 1997 was responsible for about
a 1.1 percent decline in the total index crime rate from 1993 to 1997,
while the roughly 3 percent increase in officers due to COPS expenditures
in 2000 was responsible for about 1.3 percent decline in the total index
crime rate from 1993 to 2000. Put into another context, the total crime
rate declined from 5,904 per 100,000 persons in 1993 to 4,367 per 100,000
persons in 2000, or by about 26 percent. Of this 26 percent drop, we
attribute about 5 percent to the effect of COPS. From our analysis of
violent crimes, we estimated that declines in the violent crime rate due
to COPS expenditures also varied with the level of officers due to COPS
funds. The declines in violent crime rates attributable to COPS increased
from about 2 percent in 1997 to 2.5 percent in 2000, where both of the
amounts of decline attributable to COPS expenditures are based upon
comparisons with the 1993 violent crime rate (fig. 3). We further estimate
that at its peak in 1998, COPS accounted for about a 1.2 percent decline
in the property crime rate.

Figure 3: Annual Percentage Changes in the Violent Crime Rate from 1993:
Total Change and Estimated Change Due to COPS Grants Percent change in
violent crime rate since 1993 0 -5 -10 -15 -20 -25 -30

                    1994 1995 1996 1997 1998 1999 2000 2001

      Calendar year

Estimated percentage change in violent crime rates due to COPS grants

Total percentage change in violent crime rates from 1993

Source: GAO analysis of Uniform Crime Report, Office of Justice Programs,
National Center for Health Statistics, U.S. Census Bureau, and Bureau of
Economic Analysis data.

Our estimates of the impacts of COPS expenditures on the broad categories
of crime are supported by our findings from our crime-type-specific
regression models. We find significant reductions due to COPS expenditures
for the crimes of murder and non-negligent manslaughter, robbery,
aggravated assault, burglary, and motor vehicle theft. Our analysis of
larceny indicates that while the relationship between COPS funds and
larceny is a negative one, it is not statistically significant, nor is the
effect of COPS on rape statistically significant. Further, we estimated
the effects of COPS grants on the total crime rate under various
assumptions about lags between the receipt of COPS grants and expenditures
of COPS funds. The estimates for the amount of the decline in the total
crime rate that we report here-for example, the 1.3 percent of the decline
in crime from 1993 to 2000-are among the smallest effects that we
estimated from our various models. Under different assumptions about lags
associated with COPS expenditures, the amount attributable to COPS could
be as high as 3.2 percent. Interestingly, the 1.3 percent decline in the
index crime rate that we attribute to COPS expenditures in 2000 is on the
same order of magnitude as the contribution of COPS expenditures to total
local spending on police. In 2000, for example, COPS expenditures
accounted for about 1.5 percent of total local police expenditures. We
further find that factors other than COPS expenditures account for the
majority of the decline in the crime rate. (See app. IV for more detailed
information about the results of our analysis of COPS expenditures on
crime.)

Our regression analysis of the Policing Strategies Survey data for 1993
and 1997 indicate that receipt of a COPS grant and the amount of per
capita COPS expenditures by agencies were associated with increases in the
agencies' reported use of problem-solving and place-oriented policing
practices but not crime analysis and community collaboration policing
practices (fig.4). According to our review studies of the effectiveness of
policing practices, problem-solving and place-oriented practices are among
those that the crime literature indicates may be effective in reducing
crime. With problem-solving practices, police focus on specific problems
and tailor their strategies to them. Place-oriented practices include
efforts to identify the locations where crime repeatedly occurs and to
implement procedures to disrupt these recurrences of crime. Crime analysis
includes the use of tools such as geographic information systems to
identify crime patterns. Community collaboration includes attempts to
improve or enhance citizen feedback about crime problems and the
effectiveness of policing efforts to address them. In our regressions, we
controlled for the underlying trends in the reported adoption of policing
practices, agency characteristics, and local economic conditions.

Concluding Observations

Figure 4: Reported Levels of Policing Practices in 1993 and 1997 in
Agencies That Did and Did Not Receive COPS Grants, by Category of Policing
Practice

    Mean number of policing practices

7

6

Problem-solving: COPS grantee

5

Problem-solving: Non-COPS agency

Place-oriented: COPS grantee

4

3

Place-oriented: Non-COPS agency

2

1

0

1993 1997

Calendar year

Source: GAO analysis of Policing Strategies Survey and Office of Justice
Programs financial data.

Our analysis of the National Evaluation of COPS Survey-which measured
practices in 1996 and again in 2000-showed that while COPS grantee
agencies increased their reported use of all policing practices combined,
these changes were not statistically significant in regressions that
controlled for the underlying trends in the reported adoption of policing
practices, agency characteristics, and local economic conditions. (See
app. V for more detailed information about the results of our analysis of
COPS expenditures and policing practices.)

While we find that COPS expenditures led to increases in sworn police
officers above levels that would have been expected without these
expenditures and through the increases in sworn officers led to declines
in crime, we conclude that COPS grants were not the major cause of the
decline in crime from 1994 through 2001. Other factors-which other
researchers have attempted to sort out-combined to contribute more to the
reduction in crime than did COPS expenditures. This is not surprising, as
COPS expenditures-while a large federal investment in local law
enforcement-made a comparatively small contribution to local law
enforcement expenditures for policing.

Nevertheless, our analysis shows that COPS grant expenditures did reduce
crime during the 1990s. Our models isolate the effects of COPS

                       Agency Comments and Our Evaluation

expenditures from the effects of other factors. We cannot identify another
variable that is correlated with changes in COPS expenditures, officers,
and crime rates in local communities that would explain away our findings.
Thus, we conclude that the results of our model are sound. Further, our
results do not address whether the COPS program met its goals of putting
100,000 officers on the street-and the evidence suggests that while it
funded more than 100,000 officers, it may have fallen short of achieving
this goal. Still, through the increases in officers that we attribute to
COPS expenditures, we find that COPS grants affected crime rates.
Therefore, as a demonstration of whether a federal program can affect
crime through hiring officers and changing policing practices, the
evidence indicates that COPS contributed to declines in crime above the
levels of declines that would have been expected without it.

Our work cannot identify an optimum number of officers needed by any
individual agency to maximize the effect of officers on reducing crime,
nor can it identify the specific types of practices that agencies should
adopt in particular settings. It is highly likely that there is indeed a
point where additional officers no longer affect crime. The numbers of
additional officers hired as a result of COPS were relatively small
compared with the sizes of individual police agencies, and these small
increases led to commensurate reductions in crime rates. Given resource
constraints and competing priorities at all levels of government, it is
probably unlikely that most police agencies would have the resources
available to hire large enough numbers of officers to go past the point of
diminishing returns.

We provided a draft of this report to the Attorney General for comment on
September 13, 2005. In its written comments, the Office of Community
Oriented Policing Services (COPS) drew upon information from both this
report and our prior correspondence on the effects of COPS grants on
crime. They said that we were careful and diligent in our research, and
that our findings support conclusions reached by others and correspond
with what local law enforcement leaders report. The COPS Office also
expanded upon some of our main findings, which they characterized
correctly. In their comments, the COPS Office introduced data and opinions
about potential effects of the COPS grants that were beyond the scope of
our work. We therefore cannot corroborate these statements.

For example, in discussing our findings about the effects of COPS grants
on sworn officers, the COPS Office introduced data about officers derived
from the MORE technology grants and reports that 42,058 (or 36 percent) of
the 118,397 officers that the COPS Office has funded to date are derived
from the MORE grants. Our work does not corroborate either of these
figures. We point out in Appendix VI that our estimates of a total of
88,000 additional officer-years takes into account the effects of MORE
grant expenditures.

In their comments on our finding about changes in policing practices that
resulted from COPS, the COPS Office points out that the aggregate counts
of policing practices that we used in our analysis provide only a
superficial measure of the level of community policing taking place. We
acknowledged this point in appendix VII, but chose not to speculate on the
extent to which police departments increased the amount of problem solving
or other policing practices they engaged in. By speculating that agencies
may have increased the quantity of a specific activity, the COPS Office
provides only one view of what may have happened. Another view, proffered
by policing researchers, is that there is little evidence to suggest that
problem-solving policing was implemented with sufficient rigor in enough
departments to have contributed to declines in violent crime during the
1990s. As they point out, problem-solving activities may have increased,
and they may have contributed to declines in crime, "but we simply do not
know." 20

We are sending copies of this report to other interested congressional
committees and the Attorney General. We will also make copies available to
others upon request. In addition, the report will be available at no
charge on GAO's Web site at http://www.gao.gov.

If you or your staff have any questions concerning this report, please
contact Laurie Ekstrand at (202) 512-8777 or by e-mail at
[email protected] or Nancy Kingsbury at (202) 512-2700 or by e-mail at

Eck, John, and Edward Maguire, "Have Changes in Policing Reduced Violent
Crime? An Assessment of the Evidence." In Blumstein, Alfred, and Joel
Wallman (eds.), The Crime Drop in America, Cambridge: Cambridge University
Press, 2000: p. 245.

Page 18 GAO-06-104 Community Policing Grants

[email protected]. Contact points for our Offices of Congressional
Relations and Public Affairs may be found on the last page of this report.
Key contributors to this report are listed in appendix IX.

Sincerely yours,

Laurie E. Ekstrand, Director Homeland Security and Justice Issues

Nancy R. Kingsbury, Managing Director Applied Research and Methods

Appendix I: Objectives, Scope, and Methodology

Overview of Our Approach and Methodology

In response to a request from F. James Sensenbrenner, Jr., Chairman,
Committee on the Judiciary, House of Representatives, this report provides
the findings of our evaluation of the impact of Community Oriented
Policing Services (COPS) grants on the decline in crime that occurred
during the 1990s. Our objectives were to address interrelated questions
about COPS funds, officers, crime, and policing practices. Specifically,
regarding COPS funds: (1) From 1994 through 2001, how were COPS
obligations distributed among local law enforcement agencies in relation
to the populations they served and crimes in their jurisdictions, and how
much of the obligated amounts did agencies spend? Regarding officers and
crime: (2) To what extent did COPS grants contribute to increases in the
number of sworn officers and declines in crime in the nation during the
1990s? Regarding policing practices: (3) To what extent were COPS grants
during the 1990s associated with police departments adopting policing
practices that the crime literature indicates could contribute to
reductions in crime?

To address our reporting objectives, we analyzed a database consisting of
12 years of data from 1990 through 2001 on local law enforcement agencies.
To create this database-our primary analysis database-we obtained data
from several sources, and we organized the data as a panel dataset in that
it contained information on multiple law enforcement agencies over
multiple years. For each agency, we obtained data on COPS and other
federal law enforcement grant obligations and expenditures from the
Department of Justice's (DOJ) Office of Justice Programs (OJP), and data
on index crimes and the number of sworn officers from the Federal Bureau
of Investigation's (FBI) Uniform Crime Reporting (UCR) Program. Index
crimes include the violent crimes of murder and non-negligent
manslaughter, forcible rape, robbery, and aggravated assault, as well as
the property crimes of burglary, larceny-theft, motor vehicle theft, and
arson. 1 As shown in table 1, in 2002, property crimes constituted 88
percent of the 11,877,218 index crimes. Among violent crimes, robberies
accounted for 3.5 percent of all index crimes, and aggravated assaults
accounted for 7.5 percent.

1

We excluded arson from our analyses because according to the FBI, there is
limited reporting of arson offenses to the UCR Program by law enforcement
agencies. Also because of the limited reporting of arson by law
enforcement agencies, the FBI does not include estimates for arson in its
published tables that contain offense estimates, including its table 1,
which reports its estimates of index crimes for the nation as a whole.

Page 20 GAO-06-104 Community Policing Grants

    Table 1: Index of Crimes, 2002, as Reported by the FBI, Excluding Arson

Percentage of Crime category Number index crimesa

                        Index crimesb 11,877,218 100.0%

                        Violent crimesc 1,426,325 12.0%

               Murder and non-negligent manslaughter 16,204 0.1%

                           Forcible rape 95,136 0.8%

                              Robbery 420,637 3.5%

                        Aggravated assault 894,348 7.5%

                       Property crimesd 10,450,893 88.0%

                            Burglary 2,151,875 18.1%

                         Larceny theft 7,052,922 59.4%

                      Motor vehicle theft 1,246,096 10.5%

Source: Table 1 of Federal Bureau of Investigation, Crime in the United
States, 2002, Uniform Crime Reports, Washington, D.C.: Department of
Justice. Printed annually.

Note: Although arson is part of the crime index, the FBI does not estimate
the number of arson crimes for the nation as a whole, and consequently, it
does not include an estimate for arson crimes in its table 1 of Crime in
the United States.

a

Percentages for specific types of crime within a category may not add up
to category totals because of rounding.

b

Sum of violent and property crimes.

c

Sum of murder and non-negligent manslaughter, forcible rape, robbery, and
aggravated assault.

d

Sum of burglary, larceny theft, and motor vehicle theft.

We obtained data on some of the factors that the research literature on
crime suggests are related to changes in crime. From the Department of
Commerce's Bureau of Economic Analysis, we obtained data on local economic
conditions-including employment rates and per capita income-and from the
National Center for Health Statistics and the U.S. Census Bureau-we
obtained data on demographic variables-including the percentage of the
population aged 15 to 24, and the racial and gender composition of the
population.

We also analyzed data from two surveys of nationally representative
samples of police departments on the policing practices that they
reportedly implemented in various years from 1993 to 2000. We refer to the
first survey as the Policing Strategies Survey, and it was administered in
1993 and again in 1997. 2 We refer to the second survey as the National
Evaluation of COPS Survey, as it was completed as part of the Urban
Institute's national evaluation of the implementation of the COPS program,
and we used the data from the surveys that were administered in 1996 and
2000. 3 The multiple administrations of each survey allowed us to analyze
changes in policing practices. Using agency and year identifiers, we
matched and merged data from our primary analysis database with the
agency-level records in each of the surveys.

Prior to developing and analyzing our database, we assessed the
reliability of each data source, and in preparing this report, we used
only the data that we found to be sufficiently reliable for the purposes
of our report.

In addition, to identify policing practices that are considered to be
effective in preventing crime, we analyzed reviews of research and
evaluation literature. We also reviewed relevant economic and
criminological literatures that addressed issues related to estimating
models of the effects of federal grant funds on crime rates. We spoke with
officials at the Department of Justice about the operation of the COPS
programs, and we also spoke with researchers about our approach and
methods. We reviewed our approach and methods with a group of experts in
the field of policing and crime. The group consisted of criminologists,
economists, statisticians, and criminal justice practitioners, and was
convened for us by the National Research Council of the National Academies
to enable participants to offer their individual views as experts in the
field.

We conducted our work between January 2004 and August 2005 in accordance
with generally accepted government auditing standards.

2

Rosenthal, Arlen M., and Lorie Fridell. National Survey of Community
Policing Strategies Update, 1997, and Modified 1992-1993 Data [Computer
file]. Inter-university Consortium for Political and Social Research
(ICPSR) version. Calverton, Maryland: ORC Macro International, Inc.
[producer], 2002. Ann Arbor, Michigan: ICPSR [distributor], 2002. In the
remainder of this report, we refer to the two administrations of this
longitudinal survey as the Policing Strategies Survey.

3

The second survey was conducted by the National Opinion Research Center
for the Urban Institute in 1996 and 2000 as part of the National Institute
of Justice-funded implementation evaluation of the COPS program. See Roth,
J., et al., National Evaluation of the COPS Program-Title I of the 1994
Crime Act, Washington, D.C.: National Institute of Justice, August 2000.
In the remainder of this report, we refer to the two administrations of
this second longitudinal survey as the National Evaluation of COPS Survey.

Methods Used to Address the Flow of Funds Reporting Objective

To address our first objective, we analyzed OJP financial system data on
grant obligations and expenditures and UCR data on the size of populations
served by agencies and crimes occurring within the jurisdictions of the
agencies that reported crime to the UCR. We used the OJP financial data to
compute the amount of COPS funds obligated by COPS grants and the amount
expended by local police agencies during the period from 1994 through
2001. To describe the overall COPS funding trends by grant type, we
analyzed the universe of agencies in the OJP data that received any
federal law enforcement grant during the period from 1990 through 2001,
regardless of whether or not the agency received a COPS grant during the
period and regardless of whether we were able to link the data from these
agencies to records in the UCR. For the years from 1990 through 2001, the
OJP data show that 13,332 agencies received any federal law enforcement
grant. For analyses of COPS funds by agency population sizes and for
comparisons of funding levels with levels of violent and total index
crime, we limited our analysis to the sample of agencies whose crime and
population data we were able to link to the OJP data. This resulted in a
sample of 11,187 agencies in our primary analysis database. These 11,187
agencies accounted for 86 percent of the reported crimes in the UCR data
that we received from the FBI.

The COPS Office distributed grants in a variety of programs. To describe
the amounts of COPS obligations and expenditures, we organized the COPS
grant programs into four broader categories of grants, and we reported our
results at the level of these broader categories. These four categories
include: Hiring, Making Officer Redeployment Effective (MORE), Innovative,
and Miscellaneous grants, and the specific grant programs within each
category, along with obligated amounts from 1994 through 2001 for each
grant program and category, are shown in table 2.

Table 2: COPS Obligations, 1994 through 2001, by COPS Grant Categories and
Types of Grant Programs

Obligations

Amount COPS grant category and (in billions Percentage types of grant
programs of dollars) of total

                    Total, all grant programs $7.616 100.0%

                       Hiring grant programs $4.863 63.9%

                      Police Hiring Supplement $0.143 1.9%

                            COPS Phase I $0.184 2.4%

                               AHEAD $1.245 16.4%

                               FAST $1.234 16.2%

                     Universal Hiring Program $2.055 27.0%

                            MORE grants $1.262 16.6%

                     Innovative grant programs $0.418 5.5%

                    Advancing Community Policing $0.034 0.5%

                              COPS 311 $0.005 0.1%

               Distressed Neighborhoods Pilot Program $0.112 1.5%

Community Policing to Combat Domestic Violence $0.070 0.9%

                       Anti-Gang Initiatives $0.011 0.1%

                        Integrity Initiative $0.018 0.2%

                     Methamphetamine Initiative $0.089 1.2%

                    Problem Solving Partnerships $0.038 0.5%

                 School-Based Partnership Programs $0.031 0.4%

                 Youth Firearm Violence Initiative $0.009 0.1%

                   Miscellaneous grants programs $1.073 14.1%

                          COPS in Schools $0.533 7.0%

                    Demonstration Sites Program $0.005 0.1%

                           Miscellaneous $0.132 1.7%

                         Technology Grants $0.207 2.7%

               Regional Community Policing Initiative $0.084 1.1%

                   Small Community Grant Program $0.013 0.2%

                        Tribal Grant Program $0.098 1.3%

Source: GAO analysis of Office of Justice Programs financial data.

In our analysis, we compared the distribution of COPS obligations with the
distribution of crimes contributed by agencies serving populations of

Page 24 GAO-06-104 Community Policing Grants

Methods Used to Address the Effects of COPS Expenditures on Officers and
Crime

150,000 or fewer persons and those serving more than 150,000 persons. We
used UCR population to identify agency size and crimes. The UCR population
may not reflect the population that agencies provided on the applications
for COPS grants. Our analysis of the distribution of COPS funds describes
the extent to which the distribution of funds is related to agency size-as
measured by populations served-and the distribution of violent crimes.

To assess the effects of COPS expenditures on the number of sworn officers
and crime, we developed and estimated a two-stage regression model of
these relationships. In the first stage, we estimated the relationship
between per capita COPS expenditures and per capita sworn officer rates in
the agencies included in our sample. The per capita measures were based
upon the UCR population for the jurisdiction covered by an agency. In the
second stage, we estimated the relationship between changes in per capita
COPS expenditures and changes in crime rates per 100,000 persons. As the
relationship between officer levels and crime rates may reflect a complex
and interrelated causal relationship, we used COPS hiring grants as an
instrument to help to identify the relationship between officers and
crime. To use COPS hiring grant expenditures as an instrument for sworn
officers, we made use of the fact that, unlike the purposes of other COPS
grant types, the purpose of hiring grants was limited to hiring officers.
Given the number of officers, variation in hiring grant expenditures
should be uncorrelated with other changes in crime. From our regression
results, we calculated the elasticity of crime with respect to officers or
the effect of a 1 percent change in the levels of officers on the
percentage change in crime. To assess the robustness of our results, we
estimated several specifications of our crime rate regression and
calculated the elasticities of crime with respect to officers for each
specification. We estimated these equations separately for each type of
index crime. We compared the range of our estimated elasticities with
those in the published literature on officers and crime. To estimate COPS'
contribution to the national decline in crime, we projected our sample
results to the nation as a whole by weighting our results by the ratio of
the total population in the United States to the population in the sample
of agencies included in our analysis.

In our regression models of the effects of COPS grant expenditures on
officers and crime, we organized our primary analysis database as a panel
dataset, and we limited our analysis to the 4,509 law enforcement agencies
serving populations of 10,000 or more persons and that reported complete
crime data for at least 1 year from 1990 through 2001. The number of
agencies that reported complete crime data and served populations of
10,000 or more persons varied over time, as in 1990 about 23 percent of
all agencies in the UCR data that we received from the FBI met these
criteria, and in 2001 about 21 percent did. However, these agencies also
reported the majority of crimes to the UCR. From 1990 through 2001, these
agencies reported between 86.8 percent and 88.8 percent of all index
crimes in the UCR data that we received from the FBI. Because of data
concerns with agencies serving populations of fewer than 10,000 persons,
we omitted these agencies from our analysis.

We used fixed-effects regression models to estimate the relationships
among COPS expenditures, officers, and crime. Given that we included
agencies based on the completeness of their crime data in each year, and
agencies provided complete crime data in different numbers of years over
our 1990 through 2001 analysis period, our models used an unbalanced panel
approach. In all of our models, we expressed expenditures, officers, and
crime in per capita amounts. The fixed-effects models provide estimates of
the amount of change in our dependent variables-the per capita sworn
officer rate and the per capita crime rates-that can be attributed to
changes in the per capita COPS hiring grant expenditures, controlling for
other factors that could also contribute to changes in the per capita
sworn officer rate. Our models included agency and year fixed effects to
control for unobserved differences between agencies and changes over time
within agencies in factors that could contribute to declines in crime. We
introduced state-by-year fixed effects into our regressions to control for
factors occurring at the state level-such as changes in incarceration or
state sentencing practices-that could affect crime rates. Further, we
included in our models variables that classify each agency in categories
based upon their pre-1994 trends in the growth of officers and crime.
These growth cell variables allow us to make comparisons between agencies
that were similar in their pre-COPS program trends but that varied in the
timing and amount of COPS expenditures. Finally, we included in our models
measures of other federal law enforcement grant programs that also
provided funds to state and local law enforcement agencies for hiring
officers and other crime-prevention purposes. Specifically, we included
measures of the per capita expenditures on Local Law Enforcement Block
Grants, 4 which local governments could use to hire law enforcement
officers, pay overtime,

Local Law Enforcement Block Grant Program (LLEBG), as authorized by the
Omnibus Consolidated Rescissions and Appropriations Act of 1996. (P.L.
104-134.)

Page 26 GAO-06-104 Community Policing Grants

Methods to Assess Changes in Policing Practices

purchase equipment, as well as several other purposes. Because of data
limitations, we were unable to track amounts of the Edward Byrne Memorial
State and Local Law Enforcement Assistance (Byrne Formula Grant Program) 5
grants that went to local agencies. Byrne Formula Grant funds could be
used to provide for personnel, equipment, training, technical assistance,
and information systems, among other purposes. In addition to the formula
grant program, there was also a Byrne discretionary grant program, and we
included measures for these grants.

In appendix VI, we provide the details about the specific models that we
estimated and our methods for calculating elasticities of the relationship
between changes in officers and changes in crime rates.

To assess whether COPS funds contributed to changes in policing practices,
we analyzed data from the Policing Strategies and National Evaluation of
COPS surveys, two nationally representative surveys of local law
enforcement agencies that asked about the types of policing practices that
the agencies reported implementing in various years. In each survey, chief
executives or their designees were presented a list of policing practices
and asked to indicate whether their agency implemented the practice. We
classified items in the surveys into four categories of policing practices
corresponding to general approaches to policing identified in the criminal
justice literature: problem-solving practices, place-oriented practices,
community collaboration activities, and crime analysis activities.
Problem-solving practices call for police to focus on specific problems
and tailor their strategies to the identified problems. Place-oriented
practices include attempts to identify the locations where crime occurs
repeatedly and to implement procedures to disrupt these recurrences of
crime. Community collaboration practices include improving citizen
feedback about crime problems and the effectiveness of policing efforts to
address these problems. Crime analysis includes the use of tools such as
geographic information systems to identify crime patterns. These tools may
help an agency support other practices for preventing crime, such as
problem-solving and place-oriented practices.

5

42 U.S.C. S: 3750 et seq. The Byrne Formula Grant Program was a variable
pass-through grant program administered by the Bureau of Justice
Assistance. States were required to pass through to local jurisdictions
amounts of funding based upon a variable pass-through formula.

Page 27 GAO-06-104 Community Policing Grants

For each agency in a survey, we created a summary index of the number of
such practices that agencies reportedly implemented in the years in which
the surveys were administered. We then compared mean levels of reported
practices between groups of agencies that participated in the COPS program
and those that did not participate in the program.

We used the data from the Policing Strategies Survey to make pre- and
within-COPS program comparisons of changes in reported policing practices
in 1993 and in 1997. Levels of reported practices among agencies that
received COPS grants were compared with levels among agencies that were
not funded by COPS grants over this period. We used the National
Evaluation of COPS Survey to compare levels of practices in 1996 and 2000
between groups of agencies that received COPS grants and those agencies
that were not funded by COPS over this period. In appendix VII, we provide
additional details about the surveys and our methods for analyzing the
survey data.

To assess changes in reported practices in relation to participation in
the COPS program, we estimated separate regression models of the effects
of the receipt of a COPS grant and per capita COPS expenditures on changes
in reported policing practices, controlling for various characteristics of
agencies and underlying trends in the reported adoption of policing
practices.

To identify policing practices that may be effective in reducing crime, we
analyzed six studies that provided summaries of research on the
effectiveness of policing practices and activities on reducing crime. We
chose to review studies that reviewed research, rather than reviewing all
of the original studies themselves, because of the volume of studies that
have been conducted on the effectiveness of policing practices. (See app.
VII for a list of the studies that we reviewed and additional details on
policing practices and crime.)

To construct our primary analysis database, which consisted of 12 years of

Database

data from 1990 through 2001 for law enforcement agencies that reported at
Construction and least 1 complete year of crime data to the FBI's Uniform
Crime Reporting

Program, we obtained data from several sources and merge-matched
information from these various sources to the level of the local law
Analyses enforcement agency. The sources of data that we used to compile
the

annual observations from 1990 through 2001 on local police departments

included:

     o Office of Justice Programs Financial Data-Annual data on the
       obligation and expenditures on each grant awarded by OJP. Obligations
       refer to the funds that are expected to be paid on a grant, and
       expenditures refer to the grant funds that have been paid to a
       recipient. Because OJP and the COPS Office share data on awards, the
       OJP data also included COPS grant obligation and expenditure amounts.
       We used data on grant obligation amounts to and annual amounts
       expended by each recipient of a community-oriented policing (or COPS)
       grant, 6 and annual amounts of other federal local law enforcement
       grants expended both by agencies that received COPS funds and those
       that did not. We used information about place codes and OJP vendors to
       link these data to our other sources.
     o The UCR-Annual data files on the number of crimes and sworn officers
       reported by each agency to the UCR. The data on sworn officers
       represent the reported number of full-time officers in each agency on
       October 31 of each year. We analyzed the number of sworn officers per
       10,000 persons in the covered jurisdiction. We analyzed data on the
       violent crimes of murder and non-negligent manslaughter, forcible
       rape, robbery, and aggravated assault, and the property crimes of
       burglary, larceny-theft, and motor vehicle theft. We analyzed the
       crime rate per 100,000 persons in the covered jurisdiction for each
       type of crime, as well as the rates for all index crimes, violent
       crimes, and property crimes. We used the originating agency identifier
       (ORI) variable and place codes to link crime and officer data to other
       data

7

sources.

6

In this report, COPS grants refer both to DOJ grants awarded through the
Police Hiring Supplement Program and the COPS Office's community policing
grants authorized under the Violent Crime Control and Law Enforcement Act
of 1994.

7

We used Federal Information Processing Standards codes (or FIPS codes),
which identify named population places and are issued by the National
Institute of Standards and Technology.

Page 29 GAO-06-104 Community Policing Grants

Data Used in Our Analysis of Obligations and Expenditures

     o Bureau of Economic Analysis (U.S. Department of Commerce)- Annual
       county-level estimates of per capita income and employment for each
       year from 1990 through 2001. We included in our analysis of officers,
       crime, and policing practices, measures of economic factors that are
       related to crime, such as the employment-to-population ratio and per
       capita income. We linked these data to agency-level data using place
       codes. Local economic conditions within each county are applied to
       each agency within a county.
     o National Center for Health Statistics (NCHS) and U.S. Census Bureau-
       Annual estimates of the United States resident population for each
       county from 1990 through 2001. Data obtained include population totals
       and population breakdowns by gender, race, and age. Under a
       collaborative arrangement with the U.S. Census Bureau and with support
       from the National Cancer Institute, NCHS prepared postcensal
       population estimates for 2000 through 2001. The Census estimates of
       county population from 1990 through 1999 are updated to take into
       account these postcensal estimates. We included in our analysis of
       officers, crime, and policing practices measures of demographic
       factors that are related to crime, such as the percentage of total
       population in the 15-to-24 age group-an age group associated with high
       crime rates-and the racial composition of populations. We linked these
       data to agency-level data using place codes.
     o Law Enforcement Agency Identifiers Crosswalk (Bureau of Justice
       Statistics)-The crosswalk file provides geographic and other
       identification information for each record included in either the
       Federal Bureau of Investigation's Uniform Crime Reporting Program
       files or in the Bureau of Justice Statistics (BJS) Directory of Law
       Enforcement Agencies (DLEA). The main variables each record contains
       are the UCR originating agency identifier number, agency name, mailing
       address, Census Bureau's government identification number, and Federal
       Information Processing Standards (FIPS) state, county, and place
       codes. We utilized FIPS codes to merge records from the crosswalk with
       OJP financial data and then used agency ORI codes to merge the
       crosswalk and financial data with crime data from the UCR.

To report on COPS obligations and expenditures, we first analyzed the
amounts reported in OJP financial data before we merged the financial
information onto the agency-level crime records in the UCR. In the OJP
data, each record represents either an obligation or an expenditure
amount, and an agency appears in the database each time it has either an

Page 30 GAO-06-104 Community Policing Grants

Data Used in Our Analysis of Officers and Crime

obligation or an expenditure. The total amount of obligations for COPS
grants for the 1990- through 2001- period in the OJP data was $7.62
billion.

Second, we linked the OJP financial data to agency information in the BJS
crosswalk file. We used agency identifying information in the OJP
financial data-such as FIPS state, county, and place codes-to link OJP
records with agencies in the crosswalk file. This resulted in our
identifying 13,332 agencies that had at least one record of an obligation
in the OJP financial data. Of these, 10,680 (or 80 percent) received at
least one COPS grant, and among the agencies that received COPS grants,
the total amount of COPS obligations was $7.32 billion (or 96 percent of
all COPS obligation amounts).

Third, to describe the distribution of obligations relative to agency
population and crime, we selected agencies that reported complete crime
data-12 months of crime data within a given year-in at least 1 year from
1990 through 2001, and we merged their records onto the records of the
agencies for which we had OJP financial information. This last group
contained 11,187 agencies, and 8,819 (or 78.8 percent) of these agencies
received at least one COPS grant. The total amount of COPS obligations
among these agencies was $6.01 billion (or 79 percent of the total amount
of COPS obligations from 1994 through 2001).

To analyze the impacts of COPS expenditures on officers and crime, we
started with the UCR data and included in our samples agencies that met
specific criteria. First, we identified and included agencies that
reported at least 1 year of complete crime data-that is, 12 months of
crime data in a given year-to the UCR from 1990 through 2001, and we
included agencies only in the years in which they provided complete crime
data.

Second, we excluded from our analysis agencies that the UCR classifies as
"zero-population" agencies. To avoid double counting of citizens within
geographic areas, the UCR program assigns population counts only to the
primary law enforcement agency within each jurisdiction. Consequently,
transit police, park police, university police, and similar agencies that
are contained within these jurisdictions are assigned a value of zero for
population. Because of the fact that jurisdictions among zero-population
agencies overlap with primary agencies, calculation of precise per capita
crime rates for these nonprimary agencies is problematic. Many state
police agencies also enforce laws among populations that are policed by
other local agencies, which also makes problematic calculating per capita
crime rates for state police agencies. Additionally, given that state
police agencies often have multiple substations in varied locations
throughout the state, the correct allocation of the proportion of federal
dollars to each substation is unknown. As a result, we excluded
zero-population and state police agencies from our analysis. Further, we
included in our analysis agencies whose crime records we were able to
merge-match and link with OJP financial data about COPS and other federal
law enforcement grant expenditures, as well as link with Bureau of
Economic Analysis and Census data on economic and population
characteristics.

Overall, we identified 13,133 agencies that provided complete crime data
for at least 1 year from 1990 through 2001, that were not zero-population
agencies, and that we were able to link to our other data sources. For
example, in 1990, we found 10,160 agencies out of 17,608 that met our
conditions. These 10,160 agencies represented 57.7 percent of the agencies
that were included in the 1990 data that we obtained from the FBI, but
they contained 93.2 percent of the crimes included in the 1990 data. That
the agencies that we included in our sample in 1990 represented about 58
percent of all agencies but 93 percent of all crimes indicates that most
of the agencies that we omitted with our criterion of providing complete
crime data within a year were small agencies that reported relatively
small amounts of crime to the national total. For 2001, the 9,733 agencies
that reported complete crime data and were not zero-population agencies
represented 49.1 percent of all agencies in the UCR data in 2001 and
covered 94.8 percent of all crimes (table 3).

In our analysis of officers and crime, we further limited our sample to
agencies that covered populations serving 10,000 or more persons. Complete
crime data for agencies serving populations of fewer than 10,000 persons
were missing for a large percentage of agencies, and we determined that
the data for these smaller agencies were unreliable for the purposes of
this report. In 1990, we found 4,051 of agencies serving populations of
10,000 or more persons, which represented 23 percent of the agencies
included in the data that we received from the UCR for 1990 but also
represented 86.8 percent of the crimes (table 3).

Table 3: Law Enforcement Agencies Reporting to the UCR and in Our Analysis
Dataset

                                      Year

                               Database 1990 2001

                 Uniform Crime Report data provided by the FBI

                 Number of agencies 17,608 100.0% 19,820 100.0%

          Number of index crimesa 13,962,575 100.0% 11,092,578 100.0%

 Agencies in the UCR data that reported complete crime data in at least 1 yearb

                  Number of agencies 12,168 69.1% 11,802 59.5%

            Number of index crimes 13,456,345 96.4% 10,902,718 98.3%

GAO primary analysis dataset-agencies reporting complete crime data in at
least 1 year and not zero population agenciesb

                  Number of agencies 10,160 57.7% 9,733 49.1%

            Number of index crimes 13,010,329 93.2% 10,520,533 94.8%

Percentage of population in UCR data covered by

c c

agencies 86.6% 84.5%

GAO dataset used in the analysis of officers and crime-from the primary
analysis dataset, agencies serving populations of 10,000 and more persons

                   Number of agencies 4,052 23.0% 4,247 21.4%

            Number of index crimes 12,113,789 86.8% 9,797,096 88.3%

Percentage of population in UCR data covered by

c c

agencies 76.6% 76.8%

Source: GAO analysis of Uniform Crime Report data.

aThe number of index crimes reported in the data that we received from the
FBI is less than the number of index crimes that appears in Crime in the
United States: Uniform Crime Reports. For example, for 2001, Crime in the
United States reported a total of 11,876,669 index crimes, while the data
that we received from the UCR reported 11,092,578 index crimes. The totals
reported in Crime in the United States are the estimated total numbers of
index crimes in the nation. These totals are based upon the crime reports
that the FBI receives from individual agencies and upon the methodology
that the FBI uses to estimate crimes occurring in agencies that did not
submit complete crime reports. The data that we received from the FBI
contain the crimes actually reported by law enforcement agencies to the
FBI.

b

"Complete crime data" means that an agency reported 12 months of crime
data in a given year.

c

Not applicable.

Data Used in Our Analysis To assess changes in reported policing
practices, we analyzed data from of Reported Changes in two separate
surveys of nationally representative samples of local law Policing
Practices enforcement agencies. The surveys asked key officials at
agencies about the types of policing practices that they reportedly used.
Both surveys

Page 33 GAO-06-104 Community Policing Grants

Reliability and Validity of the Data That We Used

consisted of two administrations or waves of observations on the agencies
in their respective samples. The first survey, the National Survey of
Community Policing Strategies (or Policing Strategies Survey), was
administered in 1993 and again in 1997. A total of 1,269 agencies in the
1993 and 1997 samples responded to both waves of the survey. We limited
our analysis to the 1,188 agencies that had complete data on each of the
policing practices items that we included in our analysis and that we were
able to link to our larger database on crime, officers, money, and
economic conditions. These agencies amounted to about 94 percent of the
agencies that responded to both waves of the survey. For comparability
with our analysis of the effects of COPS grants on officers and crime, we
limited our analysis to the sample of agencies that served jurisdictions
with populations of 10,000 or more persons.

The second survey, which we call the National Evaluation of COPS Survey,
was conducted by the National Opinion Research Center for the Urban
Institute in its national evaluation of the implementation of the COPS
program. 8 Of the 1,270 agencies that responded to both the 1996 and 2000
administrations of the survey, we were able to link the data from 1,067
agencies to our larger database on crime, officers, money, and economic
conditions. We restricted our analysis to agencies that served
jurisdictions having populations of 10,000 or more persons, and we
excluded from our analysis state police agencies and other special police
agencies. (See app. VII for more information about the sample of agencies
that we analyzed.)

Prior to developing our database, we assessed the reliability of each data
source. To assess the reliability of the various data sources, we (1)
performed electronic testing for obvious errors in accuracy and
completeness; (2) reviewed related documentation, including data
dictionaries, codebooks, and published research reports that made use of
the data sources; and (3) worked closely with agency officials to identify
any data problems. When we found discrepancies (such as nonpopulated
fields or what appeared to be data entry errors) we brought them to the
agencies' attention and worked with them to correct the discrepancies
before conducting our analyses. We determined that the data were
sufficiently reliable for the purposes of our report.

8

See Roth, Jeffrey., et al., National Evaluation of the COPS Program-Title
I of the 1994 Crime Act.

Page 34 GAO-06-104 Community Policing Grants

In our regression analysis of the effects of COPS expenditures on crime,
we use the UCR reported crime rates as our dependent variables. Crimes
reported to the UCR are those brought to the attention of law enforcement
agencies and subsequently reported to the UCR, or reported crimes.
Reported crimes are a subset of all crimes committed, which is the sum of
reported crimes plus crimes that are not reported to the police. Our
ultimate interest, however, lies in determining whether COPS expenditures
affected the crime rate for all crimes, whether or not they were reported
to the UCR. This raises issues related to analyzing reported crimes to
learn about all crimes.

Because data on all crimes-reported and unreported-committed within local
jurisdictions are unavailable in national data systems, we use the data on
reported crimes. The nature of the relationship between reported crimes
and all crimes therefore determines whether the results of our analysis of
COPS expenditures on reported crime would lead to biased estimates of the
effects of COPS expenditures on all crimes. Under certain circumstances,
it is possible that our analysis of the effects of COPS on the reported
crime rate could lead to overestimates of the effect of COPS on the crime
rate for all-reported plus unreported-crimes. This would lead us to
overstate the effect of COPS in reducing crime.

Several conditions could lead to overestimates of the effects of COPS
expenditures on reducing crime. If the reported crime rate and the crime
rate for all crimes diverge, we would attribute to COPS a larger reduction
in crime than is warranted. If these crime rates diverge, the reported
crime rate would either decline at a faster rate or increase at a slower
rate than the rate for all crimes, and our analysis of the effects of COPS
on the reported crime would reveal either larger declines or smaller
increases than would occur if we had data on the rate for all crimes. A
divergence between the reported crime rate and rate for all crimes could
arise for either or both of two reasons: Citizens do not report all of the
crimes they experience to the police, or the police do not record and send
to the UCR all of the crimes that citizens report to them.

To assess whether citizens decreased the rate at which they reported
crimes to the police, we reviewed data from the National Crime
Victimization Survey (NCVS). These data are drawn from a nationally
representative sample of households and are gathered independently of the
police agencies that report crime to the UCR. They therefore provide a
measure of crime that is independent of the reporting practices of police
agencies. Respondents in the NCVS are asked about their experiences as
victims of crimes. If respondents were victims of crime, they are asked if
they or others reported the criminal victimization to the police. Using
the NCVS data, it is possible to assess whether the rate at which citizens
report crimes to the police has changed over time. These data show that
during the 1990s, victims generally increased the rate at which they
reported crimes to the police. As figure 5 shows, the decline in violent
crime over the decade was steeper for all crimes reported in the survey
than for the violent crimes reported to the police. Consequently, because
the rates diverged rather than converged, victims' practices of reporting
of crime to the police during the 1990s are not likely to lead us to
overestimate the effects of COPS grants on the crime rate.

Figure 5: Violent Crimes and Violent Crimes Reported to the Police, as
Reported in the National Criminal Victimization Survey and Including
Homicides from the Uniform Crime Reports, 1990-2001

Pre-COPS program period COPS program period

  Estimated number of officers per 10,000 persons

22.5

22.0

21.5

21.0

20.5

20.0

19.5

19.0

18.5

18.0

1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 Calendar year

Predicted number of sworn officers per capita in the absence of COPS grant
expenditures (if COPS grant expenditures were equal to zero)

Actual number of sworn officers per capita

Source: GAO analysis of Uniform Crime Report, Office of Justice Programs,
National Center for Health Statistics, U.S. Census Bureau, and Bureau of
Economic Analysis data.

For police recording practices to lead to overestimates of the effects of
COPS grants on crime, it would be necessary for the agencies that received
COPS grants to decrease the rate at which they recorded and reported
crimes to the UCR. Research on police recording practices suggests that
agencies are unlikely to underreport serious crimes, such as murder, rape,
robbery, and aggravated assault. Other studies found, second, that as
police agencies adopt computer technology and become more sophisticated in
recording crimes, they became more likely to increase the rate at which
they included all citizen-reported crimes to the UCR. 9 As COPS MORE
grants provided funds for technology-such as laptop computers in police
cars-that would have increased the level of sophistication within
agencies, COPS grantee agencies would be more likely to report a larger
percentage of the crimes that citizens drew to their attention.
Consequently, changes in police reporting practices that stem from COPS
grants and lead to increases in police reporting of crimes to the UCR are
likely to lead us to underestimate the magnitude of effects of COPS grants
on reducing crime.

Two other conditions that could affect our estimates include the
following:

(1) Criminals who commit the crimes that are not reported to the police
are unresponsive to the effects of COPS expenditures, and (2) as the
number of police increase, the number of reported crimes increases,
independently of the true crime rate.

If criminals who commit crimes that go unreported to the police are
unresponsive to police presence, then we would overestimate the effects of
COPS on crime only if criminals changed their behavior to victimize more
persons who would be unlikely to report crimes to the police. This appears
to be an unlikely occurrence, as the NCVS data show a convergence between
the total number of criminal victimizations, especially for violent
crimes, and the number of crimes reported to the police.

Second, if the size of the police force systematically affects the
willingness of victims to report crime to the police or a police
department's likelihood of recording and reporting to the UCR crime
victims' reports, then these changes could lead to biased estimates of the
impact on the crime rate. However, if changes in reporting behaviors
occurred as the result of the COPS program, the likely impact on our
estimates of the effect of COPS

Lynch, James P., "Exploring the Sources of Non-response in the Uniform
Crime Reports." Paper presented at the Annual Meetings of the American
Society of Criminology Research Conference, November 19, 2003.

Page 37 GAO-06-104 Community Policing Grants

grants on crime through their effects on the number of officers is that we
would underestimate the effects of the grants on crime. 10

Given these considerations, our analysis of the effects of COPS
expenditures on crime is more likely to underestimate than overestimate
the effect of COPS funds on changes in the true crime rate.

Levitt uses three methods to estimate the bias associated with changes in
reporting practices in efforts to estimate the effects of changes in the
size of the police force on crime rates. He concludes that ignoring this
effect will lead researchers to understate the benefits associated with
increases in the size of the police force. See Levitt, Steven D., "The
Relationship between Crime Reporting and Police: Implications for the Use
of Uniform Crime Reports," Journal of Quantitative Criminology, Vol. 14,
No. 1,1998: pp. 61-81.

Appendix II: Background on the COPS Program and Studies of the Impacts of COPS
Grants on Crime

COPS and Other Local Law Enforcement Grants Distributed throughout the
1990s

Established in October 1994 by the Attorney General to implement the
administration of community policing grants under the Violent Crime
Control and Law Enforcement Act (VCCLEA) of 1994, 1 the Office of
Community Oriented Policing Services announced its first grant program in
November 1994. Prior to its establishment, in December 1993 the Department
of Justice began making community policing grants to state and local law
enforcement agencies that the COPS Office monitored. In 1993, DOJ awarded
community policing grants under the Police Hiring Supplement Program,
which was established by the Supplemental Appropriations Act of 1993 (P.L.
103-50 (1993)). The grants made under this program were funded by DOJ's
Bureau of Justice Assistance. 2

Two goals of the COPS Office were to advance community policing by
providing funding for 100,000 community policing officers and to promote
the practice of community policing, an approach to policing that involves
the cooperation of law enforcement and the community in identifying and
developing solutions to crime problems. COPS attempted to achieve these
goals by providing law enforcement agencies with grants to hire officers,
purchase equipment, and implement innovative policing practices.

According to our analysis of Office of Justice Programs data, from 1994
through 2001, the COPS Office distributed more than $7.6 billion in
grants. Grants were made in a variety of grant program funding categories.
Table 2 in appendix I contains more information about these funding
categories. The largest amount of COPS grant funds obligated-about $4.8
billion, or 64 percent of the total-was in the form of hiring grants.
These grants required agencies to hire new officers and at the same time
to indicate the types of community policing strategies that they intended
to implement. Hiring grants paid a maximum of $75,000 per officer over a
3-year period (or at most 75 percent of an officer's salary) and generally
required that local agencies cover the remaining salary and benefits with
state or local funds. Hiring programs authorized under VCCLEA and
administered by the COPS Office included the Phase I program, which funded
qualified applicants who had applied for the Police Hiring Supplement but
were denied because of the limited funds available; COPS AHEAD
(Accelerated Hiring, Education, and Deployment) for municipalities with
populations

1

P.L. 103-322 (1994), 42 U.S.C. S: 3796dd.

2

In this report, when we refer to COPS grants, we include both the grants
made under the Police Hiring Supplement and the community policing grants
authorized under VCCLEA.

Page 39 GAO-06-104 Community Policing Grants Appendix II: Background on
the COPS Program and Studies of the Impacts of COPS Grants on Crime

50,000 and above; and COPS FAST (Funding Accelerated for Smaller Towns)
for towns with populations below 50,000. In June 1995, Phase I, COPS
AHEAD, and COPS FAST were replaced by the Universal Hiring Program.

The next largest grant category was the Making Officer Redeployment
Effective (MORE) grant program, which provided funds to law enforcement
agencies to purchase equipment and hire civilians, with the goal of
expanding the amount of time spent on community policing. COPS obligated
more than $1.3 billion-or about 17 percent of total obligations-as MORE
grants. Additional COPS grant programs provided funds for specific
innovations in policing. For example, the Distressed Neighborhoods Pilot
Project grants provided funds to communities with high levels of crime or
economic distress to hire officers and implement a variety of strategies
to improve public safety, and the Methamphetamine Initiative provided
funds to state and local agencies to support a variety of enforcement,
intervention, and prevention efforts to combat the methamphetamine
problem. About $418 million-or about 5.5 percent of the total-was
obligated under these innovative grant programs. The COPS Office also
provided grants for a variety of other purposes, including funding to meet
the community policing training needs of officers and representatives of
communities and local governments (through a network of Regional Community
Policing Institutes), and grants to law enforcement agencies to hire and
train school resource officers to help prevent school violence and improve
school and student safety (the COPS in Schools Program). Over $1
billion-or about 14 percent of total obligations-was obligated among these
miscellaneous grant programs.

In each year, the COPS Office was required to distribute half of the grant
funds to agencies in communities whose populations exceeded 150,000
persons and half of the grant funds to agencies in communities with
populations of 150,000 or fewer persons. 3

During the 1990s, other federal law enforcement grant programs also
provided funds to state and local law enforcement agencies for hiring

Of funds available in any fiscal year, up to 3 percent were available for
use for technical assistance or for evaluations or studies carried out or
commissioned by the Attorney General. The requirement to allocate the
funds by size of agency population applies to the remaining funds in any
fiscal year (42 U.S.C. S: 3793 (a)(11)(B)). In addition, the COPS Office
had a national coverage requirement to ensure that no state received less
than 0.5 percent of total funding.

Page 40 GAO-06-104 Community Policing Grants Appendix II: Background on
the COPS Program and Studies of the Impacts of COPS Grants on Crime

officers and other crime prevention purposes. The Edward Byrne Memorial
State and Local Law Enforcement Assistance (Byrne Formula Grant Program) 4
was a variable pass-through grant program administered by the Bureau of
Justice Assistance (BJA). According to our analysis of data that we
obtained from OJP, from 1990 through 2001, the Byrne Formula Grant Program
distributed between $415 million and $520 million in grants. States were
required to pass through to local jurisdictions amounts of funding based
upon a variable pass-through formula. Byrne Formula Grant funds could be
used to provide for personnel, equipment, training, technical assistance,
and information systems, among other purposes. According to an evaluation
of the Byrne formula grant program, about 40 percent of Byrne subgrant
funds-the amounts passed through the states to local jurisdictions-were
for multijurisdictional task forces. 5 In addition to the formula grant
program, there also was a Byrne discretionary grant program. According to
an official at the Bureau of Justice Statistics (BJS), a large percentage
of the Byrne discretionary funds were targeted for specific programs.

The Local Law Enforcement Block Grant (LLEBG) Program was also
administered by BJA. 6 The LLEBG grant funds amounted to about an average
of $475 million per year from 1996 through 2000. According to BJS
officials, these funds were allocated by a formula based upon violent
crimes as reported in FBI's crime index. LLEBG funds were available to
local governments for hiring law enforcement officers, paying overtime,
purchasing equipment, as well as several other purposes. According to the
Urban Institute's evaluation of the implementation of the COPS program,
agencies that received COPS grants reported using both Byrne and LLEBG
funds to support their transitions to community policing. 7

Additional grant programs that provided funds to local law enforcement
agencies included the Juvenile Accountability Incentive Block Grants,

4

42 U.S.C. S: 3750 et seq.

5

Dunworth, Terence, Peter Haynes, and Aaron J. Saiger, National Assessment
of the Byrne Formula Grant Program, Washington, D.C.: National Institute
of Justice Research in Brief, June 1997.

6

Local Law Enforcement Block Grant Program, as authorized by the Omnibus
Consolidated Rescissions and Appropriations Act of 1996 (P.L. 104-134).

7

Roth, Jeffrey A., et al., National Evaluation of the COPS Program-Title I
of the 1994 Crime Act, Washington, D.C.: National Institute of Justice
Research Report, August 2000.

Appendix II: Background on the COPS Program and Studies of the Impacts of
COPS Grants on Crime

Debates over whether the COPS Office Met Its Goals for Officers and
Promoted Community Policing

Weed and Seed Grants, and several Office on Violence Against Women grants,
according to a BJS official.

The amount of COPS funding was more than sufficient to fund the federal
portion for 100,000 officers. According to the Attorney General's report,
from 1994 through 2000, the COPS Office awarded more than 30,000 grants to
over 12,000 law enforcement agencies and funded more than 105,000
community policing officers. 8 However, a research report by the Heritage
Foundation questioned how effective the COPS Office had been in putting
100,000 officers on the street. 9 The study analyzed trends in the number
of officers and concluded that the COPS program had not added 100,000
additional officers above historic trends. In its review of the COPS
Office's performance for the fiscal year 2004 budget, the Office for
Management and Budget (OMB) indicated that by 2002, COPS grants funding
was sufficient for almost 117,000 officers, a number that exceeded the
program's original commitment to fund 100,000 officers. 10 At the same
time, OMB acknowledged that fewer than 90,000 officers had been hired or
redeployed to the street. OMB reported that the COPS Office counted 88,028
COPS-funded officers on duty as of August 2002-or about 75 percent of
funded officers. In their report of October 2002 on the COPS program,
researchers at the Urban Institute updated earlier estimates of
COPS-funded officers. 11 They projected that over the years from 1994
through 2005, COPS-funded officers would add between 93,400 and 102,700
officers to the nation's communities on a temporary basis, but that not
all of these officers would be available for service at any one point in
time. They further estimated that the permanent impact of COPS, after
taking into account postgrant attrition of officers and civilians, would
be between 69,100 and 92,200 officers.

8

Attorney General of the United States, Report to Congress: Office of
Community Oriented Policing Services, Washington, D.C.: U.S. Department of
Justice, September 2000.

9

Davis, Gareth, et al., "The Facts about COPS: A Performance Overview of
the Community Oriented Policing Services Program," Washington, D.C.: The
Heritage Foundation, September 25, 2000.

10

Executive Office of the President, Performance and Management Assessments:
Budget of the United States Government, Fiscal Year 2004, Washington,
D.C.: White House, 2003.

11

Koper, Christopher S., et al., Putting 100,000 Officers on the Street: A
Survey-Based Assessment of the Federal COPS Program, Washington, D.C.: The
Urban Institute, October 2002.

Appendix II: Background on the COPS Program and Studies of the Impacts of
COPS Grants on Crime

In addition to promoting the hiring of officers, the COPS Office sought to
promote community policing. COPS hiring grant applications asked agencies
to report the types of practices that they planned to implement with their
grants, such as identifying crime problems by looking at records of crime
trends and analyzing repeat calls for service, working with other public
agencies to solve disorder problems, locating offices or stations within
neighborhoods, and collaborating with community residents by increasing
officer contact with citizens and improving citizen feedback. In 2000, the
Attorney General reported that 87 percent of the country was served by
departments that practiced community policing. 12

Studies that have addressed the extent to which the COPS Office grants
caused the spread of community policing suggest that COPS grants
accelerated the adoption of these practices but did not launch the spread
of community policing. The Police Foundation's study of community policing
practices during 1993-1 year before the COPS Office began making
grants-indicated that the practice of community policing was fairly
widespread, especially in larger police departments. 13 The Police
Foundation researcher reported that 47 percent of the agencies surveyed in
1993 reported that they either were in the process of adopting or had
adopted community policing, but that 86 percent of municipal agencies with
more than 100 sworn personnel were either in the process of implementing
or had implemented community policing. In their evaluation of the
implementation of the COPS program, Urban Institute researchers credited
COPS with promoting community policing, but the researchers concluded that
COPS funds seemed to have fueled movements that were already accelerating
rather than have caused the acceleration. In a later report, they pointed
out that for large agencies, the problem-solving practices that they
examined were already widespread by 1995, and almost no COPS grantees
reported adopting problem-solving practices for the first time between
1998 and 2000. 14

Some of the types of practices that agencies planned to implement with
their COPS grants correspond with approaches to policing that recent

12

Attorney General of the United States, Report to Congress, 2000.

13

Wycoff, Mary Ann, Community Policing Strategies: A Comprehensive Analysis,
Washington, D.C.: The Police Foundation, November 1994.

14

Johnson, Calvin C., and Jeffrey A. Roth, The COPS Program and the Spread
of Community Policing, 1995-2000. Washington, D.C.: The Urban Institute,
June 2003.

Page 43 GAO-06-104 Community Policing Grants Appendix II: Background on
the COPS Program and Studies of the Impacts of COPS Grants on Crime

Debates about COPS' Contribution to the Decline in Crime in the 1990s

reviews of policing practice suggest are effective in preventing crime. 15
For example, our review of policing practices indicates that
problem-solving policing and place-oriented policing practices-such as
those in which officers attempt to identify the locations where crime
occurs repeatedly and to implement procedures to affect crime-are among
the types of practices that research has demonstrated to be effective in
preventing crime. These practices were among the types that agencies could
implement with their COPS grants.

In 2000, the Attorney General reported that COPS-funded officers helped to
reduce crime. 16 The Attorney General's report to Congress asserted that
the drop in crime that occurred after 1994 was more than would have been
expected in the absence of the passage of VCCLEA and the creation of the
COPS Office. As evidence of the impact of COPS grants on crime, it
proffered the inverse relationship between increases in the per agency
number of police officers and decreases in the per agency levels of
violent crimes.

Studies of the impact of COPS grants on crime that attempted to take into
account factors other than just the underlying trends in crime were
released in 2001. A COPS Office-funded study examined the impact of COPS
grants on local crime rates in over 6,000 communities from 1995 through
1999. 17 Analyzing changes in crime rates in communities that had received
COPS grants, the study concluded that COPS hiring grants were effective in
reducing crime and that COPS grants for innovative policing practices had
larger impacts on reducing violent and property crime than did other types
of COPS grants. However, a study released by the Heritage Foundation,
which was based upon the analysis of county-level data, was

15

Skogan, Wesley, and K. Frydl, "The Effectiveness of Police Activities in
Reducing Crime, Disorder, and Fear," in Skogan, W., and K. Frydl, (eds.)
Fairness and Effectiveness in Policing: The Evidence, Washington, D.C.:
National Academies Press, pp. 217-251, 2004.

16

Attorney General of the United States, Report to Congress, 2000.

17

Zhao, J., and Q. Thurman, A National Evaluation of the Effect of COPS
Gants on Crime from 1994 to 1999. Report submitted to the Office of
Community Oriented Policing Services, Washington, D.C.: U.S. Department of
Justice, December 2001. In 2004, Zhao and Thurman released a revised
report on the impacts of COPS grants on crime covering the years from 1994
through 2000. In their 2004 report, the estimated effects of hiring grants
were larger and the estimated effects of innovative grants were smaller
than they reported in 2001.

Appendix II: Background on the COPS Program and Studies of the Impacts of
COPS Grants on Crime

Issues in Assessing the Contribution of COPS Grants to the Decline in
Crime in the 1990s

unable to replicate the findings of the COPS-funded study. 18
Specifically, the Heritage study found no effect of COPS hiring grants on
crime rates, but it found that grants for specific problems-such as gangs,
domestic violence, and illegal use of firearms by youth-were associated
with reductions in crime. In addition, our review of the COPS-funded study
found that its methodological limitations were such that the study's
results should be viewed as inconclusive. 19

The inconclusiveness of the findings of studies was reflected in OMB's
assessment of the performance of the COPS program. According to OMB,
although the COPS Office used evaluation studies to assess whether its
grants had an impact on crime, the results of the findings were
inconclusive, and OMB rated the COPS program as "Results Not Demonstrated"
in 2004 using its Program Assessment Rating Tool (PART).

Assessing whether COPS funds contributed to the decline in crime during
the 1990s is complicated by many factors. Nationwide, the decline in crime
began before 1993, which was before the COPS program made its first
grants. According to the FBI's data on index crimes-the violent crimes of
murder, rape, aggravated assault, and robbery and the property crimes of
burglary, larceny, and motor vehicle theft-the decline in the overall
index crime rate, as well as the property and violent crime rates started
as early as 1991 or 1992 (fig. 6). 20

18

Muhlhausen, David. Do Community Oriented Policing Services Grants Affect
Violent Crime Rates (Washington, D.C.: The Heritage Foundation, May 25,
2001).

19

Our review of this study was reported in GAO, Technical Assessment of Zhao
and Thurman's 2001 Evaluation of the Effects of COPS Grants on Crime,
GAO-03-867R (Washington, D.C.: June 13, 2003).

20

Although arson is included in the crime index, the FBI reports that it
excludes arson crimes from its estimates of national crime totals because
of limited reporting of arson by law enforcement agencies to the UCR.

 Appendix II: Background on the COPS Program and Studies of the Impacts of COPS
                                Grants on Crime

 Figure 6: Total Index, Violent, and Property Crime Rates per 100,000 Persons,
                                   1990-2001

Pre-COPS grant period COPS grant period

Index and property crime rates per 100,000 population Violent crime rate
per 100,000 population 100,000

100,000

7,000

6,300

5,600

4,900

4,200

  3,500

800

600

400

200

  0 1990 1994 1995 2000 2001 Calendar year

Index crime rate Property crime rate

Source: GAO analysis of Uniform Crime Report data.

As COPS grants cannot be the cause of the start of the decline in crime
rates, the other factors that led to the decline in the crime rate could
also have affected the decline in crime during the period that the COPS
Office made its grants. Factors such as a downturn in handgun violence,
the expansion of imprisonment, a steady decline in adult violence, changes
in drug markets, and expanding economic opportunities are among those
suggested as related to the decline in crime-especially violent crime-in

Appendix II: Background on the COPS Program and Studies of the Impacts of
COPS Grants on Crime

the 1990s. To the extent that these factors also are correlated with the
disbursement of COPS funds, this increases the challenges involved in
isolating the effects of COPS grants.

Other federal funds for local law enforcement could also have contributed
to expanding the number of police officers and contributed to declines in
crime. If the distribution of non-COPS funds such as LLEBG and Byrne
grants is correlated with that of COPS funds, and if research does not
take these funds into account, a study could attribute some of the effect
on crime of these other grant funds to COPS grants.

COPS grants were distributed in ways that make rigorous evaluations of
their causal impacts difficult to implement. Receipt of a COPS grant was
not randomly assigned; therefore, it is difficult to determine whether the
agencies that received grants are the same ones that, in the absence of
the grant, would have experienced reductions in crime. The amount of
funding certain agencies receive may also relate to the agency's ability
to combat crime. For example, certain police chiefs may be more capable
than others at acquiring funds and also more up-to-date on policing
methods. This underlying capacity of an agency to organize policing,
rather than the receipt of a particular grant, would then be the cause of
a crime decline as opposed to a particular grant. Additionally, COPS
grants were fairly widespread throughout police departments and the nation
as a whole. This distribution of grants leaves relatively few unfunded
agencies to serve as comparison groups against which to assess the
performance of the agencies that received COPS grants. The roughly 12,000
agencies that the former Attorney General reported received COPS grants by
2000 represent about 61 percent of the agencies that reported crime to the
Uniform Crime Reports.

The mechanisms by which COPS funds could affect crime have not been
explicitly examined. For example, the two prior studies that we cited did
not examine whether COPS grants potentially affect crime through changes
in police officers or through changes in policing practices, both of which
may have been affected by COPS funds. Additional officers may affect crime
by increasing police presence, by increasing arrests that lead to
incapacitation of offenders, or by deterring offenders by increasing the
likelihood of capture. Changes in policing practices toward
problem-solving or place-oriented practices that focus police resources on
recurring crime problems could also lead to reductions in crime.

Appropriate methodologies from research on crime have been developed to
address issues that could confound efforts to assess the impacts of

Appendix II: Background on the COPS Program and Studies of the Impacts of
COPS Grants on Crime

COPS grants on crime rates. For example, if COPS grants are to affect
crime through their impacts on the number of officers, then isolating the
effects of increases in officers on crime presents a challenge in
assessing the direction of the relationship between officers and crime. If
additional officers are hired in response to increases in crime rates,
then it could appear that crime causes officers. Alternatively, if
additional officers lead to reductions in crime below the levels that they
would have been without the officers, then it would appear that officers
caused changes in crime. To isolate the causal effect of COPS grants,
researchers employ the use of instruments for causal variables. One
suggestion in the research literature for an instrument for police
officers is COPS hiring grants. 21 To the extent that COPS hiring grants
buy only officers, COPS hiring grants can be used as an instrument for the
actual number of police officers and therefore be used to estimate the
relationship between crime and police officers in a way that takes into
account the possibility of this simultaneous relationship.

Second, particular forms of statistical models take advantage of
information about the variation in the amount and timing of COPS grants
among agencies to assess how changes in the number of sworn officers and
crime rates are associated with these two sources of variation. These
fixed-effects regression models use a panel of data-or repeated
observations on the same units, in this case, police agencies, over
several time periods-to assess the effects of changes in the number of
sworn officers and crime rates that are associated with variation in the
timing and amount of COPS grant expenditures. These regression methods
also allow for the introduction of controls for unobserved preexisting
differences between units (agencies) and differences over time within
units. Incorporating each agency's underlying trajectories (or growth rate
trends) in crime rates and sworn officers into the modeling of the effects
of COPS funds allow for explicit comparisons within groups of agencies
sharing similar trajectories, which helps to control for potential biases
associated with preexisting trends. 22 By identifying and explicitly
modeling the mechanisms through which a program could have its
effects-such as COPS funds leading to increases in the number of officers
and their effects

21

Evans, William N. and Emily Owens. "Flypaper COPS," College Park,
Maryland: University of Maryland. Available online at
www.bsos.umd.edu/econ/evans/wpapers/Flypaper%20COPS.pdf, 2005.

22

This methodology was implemented by Evans and Owens (2005).

Page 48 GAO-06-104 Community Policing Grants Appendix II: Background on
the COPS Program and Studies of the Impacts of COPS Grants on Crime

on crime-the possibility of a spurious relationship between inputs (such
as COPS funds) and outcomes (such as crime) can be minimized.

Appendix III: COPS Grant Obligation and Expenditure Patterns

Smaller Agencies Received Larger Amounts of COPS Obligations per Crime
than Did Larger Ones

This appendix addresses how COPS obligations were distributed among local
law enforcement agencies in relation to the populations they served and
the crimes in their jurisdictions. It also addresses how much of the
obligated amounts agencies spent. Specifically, it covers (1) the amount
of COPS obligations between 1994 and 2001, (2) the distribution of grant
funds to larger and smaller agencies relative to total index and violent
crimes, (3) the number of agencies in our sample that received COPS
grants, (4) the amounts of COPS expenditures, and (5) the amount of these
expenditures relative to total local law enforcement expenditures.

Our analysis showed that from 1994 through 2001, COPS obligated more than
$7.32 billion to 10,680 agencies for which we were able to link OJP
financial data on COPS obligations to the records of law enforcement
agencies. 1 As shown in table 4, about $4.7 billion (or 64 percent) of
these obligations were for hiring grants. Equipment and redeployment
grants made under the MORE category of grants amounted to about $1.2
billion (or about 17 percent) of total obligations.

        Table 4: COPS Grant Obligations 1994-2001, by COPS Grant Program

                    COPS grant program category Obligations

Amount Percentage (in billions of dollars)a of totalb

                         Total, all grants $7.32 100.0%

                           Hiring grants $4.69 64.1%

                            MORE grants $1.22 16.7%

                          Innovative grants $0.42 5.7%

                        Miscellaneous grants $1.00 13.7%

Source: GAO analysis of Office of Justice Programs financial data.

Note: Table 2 in appendix I identifies the specific grant programs that we
classified into these four categories of grants.

a

Amounts for each grant program category may not add up to total because of
rounding.

b

The percentages may not add up to 100 percent because of rounding.

As shown in table 5, from 1994 through 2001, slightly more than half of
the COPS obligations in the sample of agencies for which we were able to
link

1

The amount obligated to these agencies was 96.1 percent of the $7.6
billion total in COPS obligations reported in the OJP financial data.

Page 50 GAO-06-104 Community Policing Grants

OJP financial data to the records of agencies that reported crime and
population to the FBI's Uniform Crime Reporting Program 2 went to those
agencies serving populations of 150,000 or fewer persons and slightly less
than half went to those agencies serving populations of more than 150,000
persons, roughly consistent with the requirements of COPS authorizing
legislation. 3

The largest agencies-those serving populations of 150,000 or more
persons-accounted for more than half of all violent crimes reported to the
UCR. Specifically, in our sample, these agencies accounted for about 58
percent of all violent crimes reported in the UCR from 1994 through 2001.
Their share of all violent crimes declined slightly from 60 percent from
1994 through 1997 to 57 percent from 1998 through 2001. These agencies
received about 47 percent of all COPS obligations, a share that is
disproportionately small relative to their contribution to all violent
crimes. However, as shown in table 5, the amount of COPS obligations going
to agencies serving populations of 150,000 or fewer persons and those
serving populations of more than 150,000 persons was about equal to the
distribution of all index crimes occurring within these agencies.

2

The population data that we used in our analysis came from the UCR, and
they may not reflect the population information that agencies submitted to
the COPS Office on their applications.

3

Each year, the COPS Office was required to allocate half of its grant
funds in each year to agencies serving populations of 150,000 or fewer
persons and half to agencies covering populations of more than 150,000
persons.

Page 51 GAO-06-104 Community Policing Grants

Table 5: Percentage Distribution of COPS Obligations and Crime from 1994 through
                         2001, by Population Size Group

Percentage of Population size group (number total COPS Percentage of total
Percentage of all Percentage of all of persons) obligations crimes violent
crimes property crimes

                         Fewer than 10,000 15% 7% 5% 7%

                   10,000 to fewer than 25,000 13% 11% 8% 12%

                   25,000 to fewer than 50,000 11% 12% 9% 13%

                       50,000 to 150,000 15% 22% 19% 22%

                  Subtotal (150,000 or fewer) 54%a 52% 41% 54%

                       More than 150,000 47%a 48% 58% 46%

                           Total 100% 100% 100% 100%

Source: GAO analysis of Office of Justice Programs financial and Uniform
Crime Report data.

aThe subtotal for agencies serving 150,000 or fewer persons and those
serving populations of more than 150,000 may not add to 100 percent
because of rounding.

Table 6 shows that law enforcement agencies serving the smallest
populations received the largest amounts of COPS obligations on a per
crime basis. For example, agencies serving populations of fewer than
10,000 persons received, on average, $1,573 per violent crime reported
from 1994 through 2001. By comparison, agencies serving populations of
more than 150,000 persons received $418 per reported violent crime.

Table 6: Per Crime COPS Obligations, by Population Size Group and Category
of Crime, 1994 through 2001

     Population size group Ratio of total COPS obligations to total crimes

Violent Property All index crimes crimes crimes

                       Fewer than 10,000 $146 $1,573 $160

                    10,000 to fewer than 25,000 $78 $844 $86

                    25,000 to fewer than 50,000 $61 $625 $68

                         50,000 to 150,000 $47 $404 $53

                         More than 150,000 $67 $418 $80

                               Total $69 $525 $79

Source: GAO analysis of Office of Justice Programs financial and Uniform
Crime Report data.

Note: Ratios are computed as COPS obligations over the number of each type
of crime. The ratio for all index crimes is not weighted by the
contribution of violent and property crimes to the total.

Most Agencies Had Received Their First COPS Grant by 1996

As shown in table 7, of the 10,680 agencies included in our analysis, just
under half (49 percent) had received at least their first COPS grant by
1995, and 71 percent had received at least their first grant by 1996. Of
the 9,845 agencies that received at least one COPS hiring grant, 53
percent had received their first hiring grant by 1995, and 73 percent had
done so by 1996.

Table 7: Number of Agencies That Received at Least One COPS Grant
Obligation, 1994-2001, by COPS Grant Program, and Year of First COPS
Obligation

  Year of first At least one Miscellaneous COPS grant COPS grant Hiring grants
                      MORE grants Innovative grants grants

                               1994 241 241 0 0 0

                             1995 4,989 4,988 0 3 1

                         1996 2,319 1,965 1,394 265 255

                            1997 825 750 624 200 17

                            1998 910 941 231 234 18

803 605 1,010 131 1,339

                             2000 241 141 216 3 678

                            2001 352 214 378 13 476

Total number of agencies 10,680 9,845 3,853 849 2,784

Source: GAO analysis of Office of Justice Programs financial and Uniform
Crime Report data.

Note: The sum of agencies across specific COPS program categories does not
equal the total number of agencies that received at least one COPS grant
because some agencies may have received more than one type of COPS grant
in the same year.

We estimated that about 67 percent of the agencies that reported complete
crime data to the UCR for at least 1 year from 1990 through 2001 received
a COPS grant by 2001. 4 The percentages of agencies that received COPS
grants varied by the size of agencies, as measured by the size of the
population in the jurisdictions served by the agencies. As table 8 shows,
as the population served by the agencies increased, the percentage of
agencies that received a COPS grant also increased. Among the largest
agencies-those serving populations of more than 150,000 persons-about 95
percent received a COPS grant. By comparison, among agencies serving
populations of fewer than 10,000 persons, about 61 percent in our sample
of agencies received at least one COPS grant.

4

According to our definition, an agency reports complete crime data if its
reports to the UCR contain crime data for all 12 months within a year.

Page 53 GAO-06-104 Community Policing Grants

Total COPS Expenditures and Per Capita Expenditures Peaked in 2000, and
Smaller Agencies Spent More than Larger Ones on a Per Capita Basis

Table 8: Percentage of Agencies in GAO's Primary Analysis Sample That
Received at Least One COPS Grant Obligation from 1994 through 2001, by
Size of Population Served by Agencies

Size of population served by agencies Number of Percentage receiving (number of
                    persons) agencies at least 1 COPS grant

                         Fewer than 10,000 7,940 60.6%

                    10,000 to fewer than 25,000 2,673 76.2%

                    25,000 to fewer than 50,000 1,127 81.7%

                          50,000 to 150,000 702 85.2%

                          More than 150,000 185 94.6%

                       Total, all agenciesa 13,133 67.2%

Source: GAO analysis of Office of Justice Programs financial and Uniform
Crime Report data.

Note: GAO's primary analysis sample consists of 13,133 agencies that
reported at least 12 months of crime data in at least 1 year from 1990
through 2001. (See app. I.)

a

The sum of the agencies in each population size group does not add up to
the total of 13,133 because data on the size of the population served were
missing for 506 agencies. Among these 506 agencies, 276, or 54.5 percent,
received at least one COPS grant.

By 2001, agencies had drawn down about $5 billion in COPS funds (or
roughly 68 percent of all obligations awarded from 1994 through 2001). As
figure 7 shows, total COPS expenditures increased annually from 1994 to
2000. Total expenditures exceeded $900 million per year in each year from
1998 through 2001, and in 2000, they exceeded $1 billion. COPS hiring
grant expenditures totaled $3.5 billion (or roughly 70 percent of the
roughly $5 billion in hiring grant obligations made from 1994 through
2001). Hiring grant expenditures peaked in 1998-exceeding $690 million-and
declined slightly in 1999 and 2000.

Figure 7: Annual Expenditures of COPS Grant Funds, by Year

Expenditures (in millions of dollars) $1,250

  1994 1995 1996 1997 1998 1999 2000 2001 Calendar year

All COPS grants expenditures

Hiring grant expenditures

Other COPS grants expenditures

Source: GAO analysis of Office of Justice Programs financial and Bureau of
Justice Statistics crosswalk data.

The number of agencies that spent COPS funds peaked in 1998 and declined
thereafter, as figure 8 shows. In 1998, more than 7,500 agencies were
spending COPS funds. However, by 2001, the number had fallen to about
6,000.

Figure 8: Number of Agencies That Spent COPS Funds, 1994 through 2001

Number of agencies 12,000

  1994 1995 1996 1997 1998 1999 2000 2001 Calendar year

Ever spent COPS funds

Spent COPS funds during the current year

Source: GAO analysis of Office of Justice Programs financial and Bureau of
Justice Statistics crosswalk data.

COPS expenditures per population in the jurisdictions that spent funds-
per capita expenditures-also increased as the total amount of COPS
expenditures increased. Total per capita COPS expenditures peaked in 2000
at $5.6 per person. Hiring grant expenditures per capita similarly peaked
at $4.8 per person in 2000. The per capita expenditure amounts varied by
size of agency, as smaller agencies generally spent more on a per capita
basis than did larger agencies. Agencies serving populations of fewer than
10,000 persons spent about twice as much COPS grant monies on a per capita
basis than did the larger agencies. For example, per capita COPS
expenditures for agencies serving fewer than 10,000 persons averaged $6.6
as compared with about $3.4 for agencies serving populations of more than
150,000 persons.

COPS Expenditures Amounted to about 1 Percent of All Local Law Enforcement
Expenditures

From 1994 through 2001, COPS expenditures amounted to about 1 percent of
total local expenditures for nationwide police services, based upon BJS
data on criminal justice expenditures and our analysis of OJP data on COPS
grant expenditures. 5 From 1994 through 2001, total local expenditures for
police services increased from about $46 billion to $72 billion. During
the years from 1998 through 2000, when COPS expenditures neared and then
exceeded $1 billion per year, the contribution of COPS expenditures to
local police expenditures increased to about 1.5 percent of total local
expenditures for police services.

5

This includes total expenditures for jurisdictions with agencies that
received COPS grants as well jurisdictions with agencies that did not
receive COPS grants.

Appendix IV: COPS Expenditures Led to Increases in Sworn Officers and Declines
in Crime

COPS Expenditures Led to Increases in Sworn Officers above Levels That
Would Have Been Expected without Them and Were Responsible for about
88,000 Officer-Years

This appendix addresses our second reporting objective, which has two
parts: determining the extent to which COPS grant expenditures contributed
to increases in the number of sworn officers in police agencies, and
determining the extent to which COPS grant expenditures led to reductions
in crime through their effects on sworn officers.

We found that COPS hiring grants were significantly related to increases
in sworn officers above levels that would have been expected without the
expenditures, after controlling for economic conditions in the counties in
which agencies were located, population composition, and preexisting
trends in agencies in the growth rate of sworn officers. Further, the
effects of COPS hiring grants were consistent across several different
regression models, including those that controlled for state-level factors
that could affect the size of local police forces-such as state-level
differences in the amount of funding provided to local departments.
Overall, the parameter estimates from our models indicate that each
$25,000 in COPS hiring grant expenditures was associated with roughly an
additional 0.6 officers in any given year. 1 With the exception of MORE
grants, no other types of COPS grant expenditures were associated with
increases in officers.

Using the results from our regression models, we calculated for each year
from 1994 through 2001 the number of sworn officers nationwide that would
have been on the street absent the COPS expenditures in each year. The
difference between this amount and the actual level of sworn officers
yielded the number of officers due to COPS expenditures in a given year.
The number of officers due to COPS increased from 84 in 1994 to 17,387 in
2000, and then declined to 12,226 in 2001 (table 9). The increase and
decrease in the number of officers due to COPS followed the pattern of
COPS expenditures, which peaked in 2000 and then declined (see fig. 7 in
app. III). Adding up the number of officers due to COPS in each year
across the years from 1994 through 2001, we arrive at a total of about
88,000 sworn officer-years due to COPS expenditures.

From 1997 through 2000, when COPS expenditures neared or exceeded $1
billion per year, we estimated that the expenditures led to increases in

1

Our estimate of the effect of COPS expenditures on officers is consistent
with those in the research conducted by Evans and Owns, who used COPS
hiring grants to estimate the relationship between changes in sworn
officers and crime. They estimated that each $25,000 in COPS hiring grant
expenditures produced an additional 0.7 of an officer in a given year.
Evans and Owens, "Flypaper COPS," 2005.

Page 58 GAO-06-104 Community Policing Grants Appendix IV: COPS
Expenditures Led to Increases in Sworn Officers and Declines in Crime

sworn officers of between 2.4 percent and 2.9 percent above levels
expected without them. In years prior to 1997, and in 2001, when COPS
expenditures were lower, the percentage of officers due to COPS
expenditures were lower than occurred from 1997 through 2000.

Table 9: Estimated Effect of COPS Expenditures on the Number of Sworn
Officers Nationwide in Each Year, 1994-2001

Estimated number of officers Percentage of total number of Year due to
COPS expenditures officers in the United States

84 0.02%

1,916 0.35%

8,639 1.55%

13,897 2.42%

17,630 3.02%

16,415 2.72%

17,387 2.91%

12,226 2.05%

                         Total, officer-yearsa 88,195 b

Source: GAO analysis of Uniform Crime Report, Office of Justice Programs
financial, Bureau of Economic Analysis, National Center for Health
Statistics, and U.S. Census Bureau data.

aThe total number of officer-years due to COPS expenditures is the sum of
the number of officers due to COPS in each year. We call this total the
number of officer-years due to COPS expenditures. It is not directly
comparable with estimates of the number of sworn officers on the street as
a result of COPS funds or with estimates of the number of officers funded
by the COPS Office.

b

Not applicable.

An officer-year is the number of officers in a given year that were
associated with COPS expenditures. According to this measure, an
individual officer-or person-might be included in our counts of officers
due to COPS in several years. Therefore, our estimate of the total number
of officer-years arising from COPS expenditures is not equivalent to the
number of officers that the COPS Office reportedly funded, nor does it
represent an estimate of the total number of officers as a result of COPS
grants. For a given year, however, our estimate represents the number of
COPS-funded officers on the street. (For additional details on the methods
we used to estimate the effects of COPS expenditures on officers, see app.
VI.)

Appendix IV: COPS Expenditures Led to Increases in Sworn Officers and
Declines in Crime

LLEBG Funds Also Contributed to Increases in Officer Strength

In addition to our findings of the effects of COPS expenditures on the
level of sworn officers, we found that Local Law Enforcement Block Grants
expenditures also contributed to increases in officers above levels
expected without them. Our finding about LLEBG grants effects on sworn
officers is consistent with interview and survey responses reported by
Urban Institute researchers in their evaluation of the implementation of
the COPS program. 2 In their interviews with police chiefs, they found
that the chiefs reported that they used LLEBG to supplement COPS funds.
LLEBG grants could be used for a variety of purposes in addition to
funding officers.

COPS Expenditures Led to Reductions in Crime through Increases in Officers

Estimating the impact of COPS expenditures on changes in crime rates
through their effects on the number of sworn officers, we found that COPS
expenditures were associated with declines in crime rates for total,
violent, and property crimes, as compared with their baseline levels in
1993, the year prior to the distribution of COPS grants. The amounts of
decline in crime rates varied among crime types and across years. The
variation in the decline in crime rates in various crime types arose from
our estimates of the effects of changes in officers on crime rates, and
the variation over time within crime types arose from the variation in
COPS expenditures. For example, for the total crime rate, we found that
the impact of COPS peaked in 1998, as for that year, we estimated that
COPS led to a reduction in the total crime rate of almost 1.4 percent from
the level of crime in 1993. From 1999 and 2000, COPS expenditures of
between $920 million and about $1 billion led to reductions in the total
crime rate of about 1.3 percent, again, as compared with the 1993 level.
In years prior to 1998 and in 2001, when COPS expenditures were lower than
their levels in 1998 through 2000, the declines in total crime arising
from COPS expenditures also were less than 1.3 percent (table 10).

Similarly, for violent and property crimes, we found that the amount of
decline associated with COPS expenditures varied from year to year, and
for both of these crime categories, the largest decline in crime occurred
during 1998. COPS expenditures led to a decline in violent crime of almost

2.6 percent in 1998, compared with violent crime levels in 1993. For 1999
and 2000, COPS expenditures led to about a reduction of about 2.4 percent
in violent crime, from the 1993 level. For property crimes, the impact of

2

Roth, Jeffrey, et al. National Evaluation of the COPS Program, 2000.

Appendix IV: COPS Expenditures Led to Increases in Sworn Officers and
Declines in Crime

COPS expenditures from 1998 through 2000 was between 1.1 percent and

1.2 percent, as compared to the 1993 level (table 10).

Table 10: Estimated Percentage Change in Crime Rates from 1993 Levels Due
to COPS Expenditures, 1994-2001, by Crime Type Category

                                 Crime category

                Year Total crimes Violent crimes Property crimes

-.01 -.01 -.01

-.16 -.29 -.13

-.70 -1.29 -.60

-1.11 -2.05 -.95

-1.39 -2.57 -1.19

-1.28 -2.36 -1.10

-1.34 -2.48 -1.15

-0.93 -1.73 -.80

Source: GAO analysis of Uniform Crime Report, Office of Justice Programs
financial, Bureau of Economic Analysis, National Center for Health
Statistics, and U.S. Census Bureau data.

Note: All estimates of the magnitude of the impact of COPS on crime are
made with respect to the level of crime in 1993, the baseline year for our
comparisons. The crime rates in 1993 in the data in our sample were as
follows: the total crime rate was 5,904 per 100,000 persons; the violent
crime rate was 846 per 100,000 persons; and the property crime rate was
5,058 per 100,000 persons.

Our estimates of the impact of COPS expenditures on crime through their
effects on the number of officers represent the effects of COPS
expenditures on crime net of the effects of other factors that we
controlled for in our model-including changes in economic conditions,
population composition, and pre-COPS program trends in police agencies'
growth rate of sworn officers and growth rate in crime. By controlling for
pre-COPS program growth rates in officers and crime, we made comparisons
between agencies within population size categories that had similar growth
rates in officers and crime but which differed on the timing and amount of
COPS expenditures. In addition, through the use of state-by-year fixed
effects, we controlled for state-level factors that could affect crime
rates, such as changes in sentencing policy or state incarceration.

As our estimates of the impact of COPS expenditures on crime come, in
part, from our estimates of the effects of changes in officers on crime,
we compared our estimates of the effect of changes in officers on changes
in crime with estimates of these effects that appear in recent research.
We found that each 1 percent increase in sworn officers was associated
with about a 0.4 percent decline in total crime, about a 0.8 percent
decline in violent crime, and a slightly less than 0.4 percent decline in
property

Appendix IV: COPS Expenditures Led to Increases in Sworn Officers and
Declines in Crime

Various Specifications of Our Regressions Yielded Consistent Findings
about the Effect of COPS Expenditures on Crime

crime. Our estimates of this relationship-the elasticity of crime with
respect to officers-is consistent with estimates that appear in recent
literature of the effects of changes in police officers on changes in
crime rates. Others report elasticities that are similar to ours. For
example, in a study that used COPS granted officers to estimate the effect
of increases in officers on crime, the authors reported an estimated
elasticity for violent crime of -0.99 (a 1 percent increase in officers
led to a 0.99 percent decline in violent crimes) and a property crime
elasticity of -0.26. 3 In another paper that used electoral cycles to
estimate the effect of increases in officers on crime, the author provides
a set of elasticities under different model specifications. 4 The
elasticity for property crimes was calculated to be about -0.3, and the
elasticity for violent crimes was about -1.0. (See app. VI for more
information on the methods that we used to calculate our elasticities and
to estimate the impact of COPS expenditures on crime.)

While we found that COPS expenditures were associated with reductions in
total crime and the violent and property crime categories, when we
examined the effects of COPS expenditures on specific types of index
crimes, we found significant reductions in murder, robbery, aggravated
assault, burglary, and motor vehicle theft. We found a negative
association between COPS expenditures and larceny, but this effect was not
statistically significant. Finally, we found a positive but statistically
insignificant association between COPS expenditures and rape. (See table
17 in app. VI.)

Additionally, for agencies that served populations of 10,000 or more
persons, we found that the effects of COPS expenditures on the total crime
rate were consistent across agencies that served populations of varying
sizes with the exception of agencies that served populations of between
25,000 and 50,000 persons. The magnitude of the effects tended to increase
with the size of agencies, where agency size refers to the population
served by the agency. In general, as the size of agencies

3

Evans and Owens, "Flypaper COPS," 2005.

4

Levitt, Steven D. "Using Electoral Cycles in Police Hiring to Estimate the
Effect of Police on Crime: Reply" American Economic Review, September
2002, 92(4), pp. 1244-50. Justin McCrary found that Levitt's original
estimates of the effect of officers on crime suffered from a computation
error. Levitt was able to confirm his results after correcting the error
using an alternative instrument. See McCrary, Justin, "Do Electoral Cycles
in Police Hiring Really Help Us Estimate the Effect of Police on Crime:
Comment." American Economic Review. June 2002, 92(4), pp. 1236-43.

Appendix IV: COPS Expenditures Led to Increases in Sworn Officers and
Declines in Crime

increased, we found that the impact of COPS expenditures on the total
crime rate also increased. For agencies serving populations between 25,000
and 50,000, we observed a negative relationship between COPS expenditures
and crime. However, the estimated effect was not statistically
significant. (See table 18 in app. VI.)

As there are uncertainties associated with formulated regression models,
and point estimates derived from a single regression model can give
misleading information, we estimated our regressions under different
assumptions about how COPS expenditures could affect crime. Under the
various models, we introduced lagged effects, nonlinear effects for COPS
hiring grants, and effects for the year of receipt of COPS grants-to test
whether the impact of COPS occurred in the years in which the money was
spent. From the various specifications, we estimated the elasticity of
crime with respect to officers. We found that the elasticity for total
crimes ranged from -0.41 to -0.95. The elasticity that we used to
calculate the impact of COPS on the decline in index crimes was -0.42,
which is at the lower end of the range of elasticities that we estimated.
Therefore, under assumptions different from the preferred specification
about how COPS expenditures are related to officers and crime, we would
arrive at a larger estimated impact of COPS on the decline in crime than
we report above. Also, under the varying assumptions about how COPS
expenditures are related to crime, we estimated elasticities of violent
crimes with respect to officers and elasticities of property crimes with
respect to officers. For violent crimes, the elasticities derived from
these regressions ranged from -0.76 to -1.8. The elasticity that we used
to estimate the impact of COPS on the decline in violent crimes was -0.8.
This elasticity is at the lower end of the range of elasticities that we
estimated, which implies that the impacts of COPS on violent crimes could
be larger than the impacts that we reported. For property crimes, the
range of estimated elasticities was from -0.35 to -0.80. (See table 20 in
app. VI.)

In addition to our findings of the effects of COPS expenditures on crime,
we found that LLEBG expenditures were consistently associated with
declines in total crime rates and declines in the murder, rape, robbery,
aggravated assault, burglary, and larceny crime rates. Only for motor
vehicle theft did we not find a significant effect of LLEBG expenditures.
However, because LLEBG grant funds are related to the levels of violent
crime occurring within a jurisdiction, the relationship between LLEBG

Appendix IV: COPS Expenditures Led to Increases in Sworn Officers and
Declines in Crime

Factors other than COPS Expenditures Contributed Larger Amounts to the
Reduction in Crimes, but COPS Contribution Was in Line with COPS
Expenditures

expenditures and crime may be one of bidirectional causality. 5 By this,
we mean because LLEBG grant amounts were determined in part on the levels
of violent crime, violent crime in a community can be construed as a cause
of LLEBG grants in addition to an effect of having received them. (See
table 17 in app. VI.)

The decline in crimes attributable to COPS expenditures accounted for at
most about 10 percent of the total drop in crime from 1993 to 1998, and
about 5 percent of the drop from 1993 to 2000. Therefore, various factors
other than COPS expenditures were responsible for the majority of the
total decline in crime during the 1990s. While in our regression models of
the effects of COPS funds on crime, we were able to control for the
effects of many factors that could be related to the decline in crime, we
did not attempt to estimate the amount that each of these factors
individually had contributed to the overall drop in crime. 6 Rather, by
isolating the amount by which crime rates declined because of COPS and
comparing that amount with the total decline in crime from our 1993
baseline year, we calculated COPS contribution to the overall decline in
crime. The amount of the total drop in crime not associated with COPS
expenditures reflects the amount due to factors other than COPS.

While COPS' contributions to the decline in crime rates did not account
for the majority of the total drop in crime rates, the amounts of declines
in crime rates attributable to COPS were on the same order of magnitude as
were COPS expenditures' contributions to local law enforcement
expenditures for police. From 1994 through 2001, COPS expenditures
amounted to slightly more than 1 percent of total local expenditures for

5

According to officials at the Bureau of Justice Statistics, the formula
for determining LLEBG grant amounts is based in part upon the level of
violent crime occurring within a jurisdiction. By comparison, there was no
requirement for COPS funding to be related to violent crime. Therefore,
without an instrument to isolate the relationship between LLEBG
expenditures and crime rates, we cannot conclude that the estimated
effects of LLEBG expenditures on crime would hold if we were able to
isolate statistically the causal direction of effects.

6

Some of the factors associated with the crime drop have been discussed in
Blumstein and Wallman (2002). See Blumstein, A., and J. Wallman (eds.),
The Crime Drop in America, Cambridge, United Kingdom: Cambridge University
Press, 2000.

Appendix IV: COPS Expenditures Led to Increases in Sworn Officers and
Declines in Crime

police services nationwide. As we found and reported, COPS expenditures
were responsible for about a 1.4 percent decline in the total crime rate.

Appendix V: COPS Expenditures Associated with Policing Practices That Crime
Literature Indicates Are Effective in Preventing Crime

Comparisons of Pre- and Within-COPS Grant Program Levels of Reported
Policing Practices Show That COPS Grantee Agencies Reported Larger
Increases than Non-COPS Agencies

This appendix addresses our third reporting objective: determining the
extent to which COPS grant expenditures during the 1990s were associated
with police departments adopting policing activities or practices that the
crime literature indicates could contribute to reductions in crime.
Specifically, it describes the results of our analyses of the
relationships between COPS grant expenditures and changes in policing
practices reported in two surveys of local law enforcement agencies, and
it summarizes our assessment of studies that conducted systematic reviews
of research on the effectiveness of various policing practices. Our
analysis of the first of the two surveys of policing practices compares
changes in reported policing practices between 1993 and 1997, that is,
prior to the distribution of COPS grants and after many COPS grants had
been distributed. In our analysis of the second survey, we compare changes
from 1996 to 2000, or during the implementation COPS program. In addition,
we provide a limited summary of our analysis of systematic reviews of
evaluations of policing practices that could contribute to reductions in
crime. (See app. VII for the details related to our methodology for
analyzing policing practices.)

Prior to the implementation of COPS grants, many local law enforcement
agencies had adopted a number of problem-solving, place-oriented, crime
analysis, and community collaboration policing practices. Problem-solving
practices refer to efforts by the police to focus on specific problems and
tailor their strategies to the identified problems. Place-oriented
practices include attempts to identify the locations where crime
repeatedly occurs and to implement procedures to disrupt these recurrences
of crime. Crime analysis includes the use of tools such as geographic
information systems to identify crime patterns. Community collaboration
includes attempts to improve or enhance citizen feedback about crime
problems and the effectiveness of policing efforts to address them.

Our analysis of the Policing Strategies Survey data for 1993-the year
before COPS grants were distributed-indicates that surveyed agencies that
received a COPS grant between 1994 and 1997 reported higher mean levels of
the above policing practices than agencies that did not receive a COPS
grant between 1994 and 1997. For example, in 1993, the mean number of all
practices reported by grantee agencies was about 13 out of a possible 38
practices, while the mean number of all practices reported by nongrantee
agencies was about 11 practices. However, among the agencies that received
a COPS grant between 1994 and 1997, there were larger increases in the
mean level of all reported practices between 1993 and 1997 except for
those related to crime analysis. COPS grantee

Appendix V: COPS Expenditures Associated with Policing Practices That
Crime Literature Indicates Are Effective in Preventing Crime

agencies reported in 1997 an increase of about 3.5 practices overall, as
compared with a mean increase of less than 2 practices by the agencies
that did not receive COPS grants during this period. The largest
differences between COPS grantees and nongrantee agencies in the reported
increase in practices occurred for the problem-solving and place-oriented
practices (table 11).

Table 11: Mean Levels of Policing Practices in 1993 and 1997, by Category
of Policing Practices and whether Agencies Received a COPS Grant between
1994 and 1997

Agencies that did not COPS grantee agencies receive a COPS grant

Category of policing practice 1993 1997 Difference 1993 1997 Difference

Problem solving 4.57 5.80 1.24 4.16 4.68

Place oriented 2.98 4.21 1.23 2.38 2.84

Community collaboration 3.48 4.41 0.93 2.69 3.45

Crime analysis 1.88 1.93 0.05 1.66 1.71

Total 12.90 16.34 3.44 10.89 12.69

Source: GAO Analysis of Policing Strategies Survey, Office of Justice
Programs financial, Bureau of Economic Analysis, National Center for
Health Statistics, U.S. Census Bureau, and Uniform Crime Report data.

From a series of regression models of the effects of COPS grants on
changes in policing practices, we found that both the receipt of a COPS
grant, and the amount of per capita COPS expenditures by agencies were
associated with increases in the levels of reported policing practices
between 1993 and 1997. Our regressions control for the underlying trend in
the reported use of policing practices, for differences in agency
characteristics that could be associated with increases in reported levels
of policing practices-such as the size of the jurisdiction-and changes in
the economic and social characteristics of the county in which the agency
was located. We estimated separate regressions of the effect of the
receipt of a COPS grant and of the cumulative per capita amount of COPS
expenditures on the levels of reported policing practices.

Our regression models for estimating the effects of receipt of a COPS
grant on the change in police practices between 1993 and 1997 show that
agencies that received at least one COPS grant had significantly larger
changes in the overall number of practices than did agencies that did not
receive a COPS grant during this period. Specifically, according to our
analysis of the survey data, the average number of practices increased by

2.9 over this period, and the receipt of a COPS grant accounted for 1.8 of
this reported increase. Further, when we examined our results from

Appendix V: COPS Expenditures Associated with Policing Practices That
Crime Literature Indicates Are Effective in Preventing Crime

The Effects of COPS Grants on Agencies' Reported Increases in Policing
Practices Differed across Agencies Serving Populations of Different Sizes

separate regressions for the different categories of practices, we found
that receipt of a COPS grant was associated with significant increases in
reported levels of problem-solving and place-oriented practices, but was
not related to changes in community collaboration or crime analysis
practices. (See app. VII for details.)

Our regression models further show that changes in practices were also
associated with the cumulative amount of per capita spending on COPS
grants. All other things being equal, a $1 increase in per capita spending
was associated with an increase of 0.23 policing practices. As we found
for the effects of the receipt of a grant on changes in police practices,
these regressions also showed that the level of per capita spending on
COPS grants was significantly associated with increases in problem-solving
and place-oriented practices. However, per capita spending on COPS grants
was also associated with increases in crime analysis practices. (See app.
VII for details.)

Receipt of a COPS grant was associated with increases in the overall
adoption of policing practices among agencies serving populations of
different sizes. Regardless of the size of populations served, agencies
that received COPS grants adopted almost twice as many practices between
1993 and 1997 as agencies that did not receive COPS grants. However, in
both years, agencies serving larger populations also reported higher mean
levels of policing practices (table 12 and fig. 9).

Table 12: Mean Levels of Policing Practices in 1993 and 1997, by Size of
Agency and whether Agencies Received a COPS Grant between 1994 and 1997

Agencies that did not COPS grantee agencies receive a COPS grant

Jurisdiction population (number of persons) 1993 1997 Difference 1993 1997
Difference

10,000 to fewer than 50,000 11.87 15.14 3.27 10.12 11.80

50,000 to 150,000 14.58 18.70 4.12 14.40 16.81

More than 150,000 19.30 22.82 3.52 19.00 20.91

    Source: GAO Analysis of Policing Strategies Survey and Office of Justice
                            Programs financial data.

Appendix V: COPS Expenditures Associated with Policing Practices That
Crime Literature Indicates Are Effective in Preventing Crime

Reported Levels of Policing Practices among COPS Grantees Did Not Increase
Overall from 1996 to 2000

Figure 9: Reported Levels of Policing Practices in 1993 and 1997 in
Agencies That Received and Did Not Receive COPS Grants, by Size of
Population Served

  Mean number of policing practices

7

6

Problem-solving: COPS grantee

5

Problem-solving: Non-COPS agency

Place-oriented: COPS grantee

4

3

Place-oriented: Non-COPS agency

2

1

0

1993 1997

Calendar year

Source: GAO analysis of Policing Strategies Survey and Office of Justice
Programs financial data.

Our regressions of the effect of COPS expenditures on changes in reported
levels of policing practices between 1993 and 1997, indicate, however,
that the effects of receiving a COPS grant were larger in agencies in
jurisdictions serving fewer than 50,000 persons and in jurisdictions
serving more than 150,000 persons, than in agencies in jurisdictions
serving populations of between 50,000 and 150,000 persons.

Our analysis of the National Evaluation of COPS Survey data on policing
practices in 1996 and in 2000 also showed that agencies that received COPS
grants reported larger increases in the mean level of policing practices
than did non-COPS grantee agencies, but that the effects were not
statistically significant. The findings suggest that there was no
continued overall increase in reported policing practices in the period
from 1996 to 2000.

Regardless of when agencies received COPS grants and made COPS
expenditures, we found that COPS grantee agencies reported larger
increases in policing practices between 1996 and 2000 than did the
agencies that did not have COPS grants in these years. For example, for
the agencies that received their first COPS grant in 1996 or before, the
average increase in reported use of policing practices from 1996 to 2000
was about 21 percent, and for the agencies that made COPS grant

  Appendix V: COPS Expenditures Associated with Policing Practices That Crime
             Literature Indicates Are Effective in Preventing Crime

expenditures after 1996, the average increase in reported use of policing
practices was about 17 percent. By contrast, for the agencies that had not
made any COPS grant expenditures by 2000, there was about a 0.2 percent
decrease in the reported use of policing practices from 1996 to 2000, and
for the agencies that did not make any COPS grant expenditures after 1996,
there was about a 3 percent increase in the reported use of policing
practices from 1996 to 2000 (table 13).

Table 13: Difference in Mean Levels of Reported Policing Practices in 1996
and 2000, by Category of Policing Practices and Timing of COPS Grant
Expenditures

Made COPS Made COPS expenditures expenditures in 1996 after 1996 or before

Did not make COPS Did not make COPS expenditures between expenditures
after 1996 1994 and 2000

Category of policing 1996 Change 1996 Change 1996 Change 1996 Change
practice

Problem solving and place oriented 6.09 1.00 6.08 1.45 6.91 0.11 7.08
-0.13

Community 3.36 0.53 3.42 0.56 3.28 0.47 3.33 0.38 collaboration

            Crime analysis 1.67 0.32 1.70 0.39 1.87 -0.20 1.88 -0.26

               Total 11.12 1.86 11.21 2.38 12.06 0.38 12.30 -0.02

Crime Literature Provides Evidence for Effectiveness of Some Policing
Practices

Source: GAO analysis of National Evaluation of COPS Survey and Office of
Justice Programs financial data.

Although we observed larger average increases in reported policing
practices among agencies that spent COPS grant funds than among agencies
that did not spend COPS grant funds, when we controlled for underlying
trends in the reported adoption of policing practices and agency
characteristics, we found that changes in per capita COPS expenditures
made between the period preceding wave 1 of the survey (1994 through 1996)
and the period following wave 1 of the survey (1997 through 2000) were not
associated with changes in reported overall policing practices between
1996 and 2000 (app. VII). This suggests that there was no continued
overall increase in reported policing practices in the period from 1996 to
2000, as a function of COPS grant expenditures.

Our analysis of six systematic reviews of evaluations of the effectiveness
of various policing practices in preventing crime indicates that the
current evidence ranges from moderate to strong that problem-oriented
policing practices and place-oriented practices are either effective or
promising as strategies for addressing crime problems. For example,
problem-oriented approaches that focus on criminogenic substances such as
guns and drugs

Page 70 GAO-06-104 Community Policing Grants Appendix V: COPS Expenditures
Associated with Policing Practices That Crime Literature Indicates Are
Effective in Preventing Crime

appear to be effective in reducing both violent and property crimes. And
hot spots approaches-place-oriented approaches that temporarily apply
police resources to discrete locations where crime is concentrated at much
higher rates than occur jurisdictionwide-have also been found to be
effective in reducing crime. However, the magnitudes of the effects of
these interventions are difficult to estimate, especially on citywide
crime rates, as the interventions that were reviewed as effective
generally were concentrated in comparatively small places. Further, the
enduring nature of these interventions is not fully understood. It is not
known, for example, how long the effects of a problem- or place-oriented
intervention persist. In addition, some of the reviews point out that
research designs undertaken to date make it difficult to disentangle the
effects of problem-oriented policing from hot spots policing. There is
suggestive, but limited, evidence that the combination of these practices
may be more effective in preventing or reducing crime than any one
strategy alone.

In contrast to the findings on problem-oriented and place-oriented
policing practices, there is little evidence in the literature for the
effectiveness of community collaboration practices-such as increasing foot
patrol, establishing community partnerships, and encouraging citizen
involvement-in reducing or preventing crime.

Appendix VI: Methods Used to Estimate the Effects of COPS Funds on Officers and
Crime

Prior Literature on the Relationship between Officers and Crime Addresses
Issues Relating to Estimating the Effects of COPS Funds on Crime

In this appendix, we describe the methods we used to address our reporting
objective regarding the impacts of the COPS funds on officers and crime:
determining (1) the extent to which COPS grant expenditures contributed to
increases in the number of sworn officers in police agencies in the 1990s
and (2) the extent to which COPS expenditures contributed to declines in
crime in the 1990s through their effects, if any, on officers.

In examining the effect of COPS funds on crime, we estimate the impacts of
the funds on crime through their impacts on officers. The effect of police
on crime has a theoretical basis in the economics literature. Economic
models posit that criminals weigh the gains from criminal activity against
its costs-the possibility of arrest and incarceration. Anything that
increases the probability of arrest, such as additional police, will thus
deter criminal activity; we might call this the deterrence effect. A
second effect stems from arrests directly. If criminals are arrested and
incarcerated, they will not be able to commit street crimes; we might call
this the incapacitation effect.

The relationship between police and crime has been studied empirically,
with mixed results. Several reviews of research that investigated this
relationship have reported that a minority of papers find a significant
negative relationship between increases in the number of officers and
crime. 1 However, these reviews also point out that many of the studies
have methodological flaws. In a report to Congress on what works in crime
prevention, Lawrence Sherman and others drew upon a limited body of
research that addressed the methodological concerns and concluded that
increases in the number of police officers work to prevent crime. 2

One of the major methodological issues associated with estimating the
relationship between police officers and crime is the issue of reverse
causality. This issue revolves around determining how to disentangle the
relationship between the number of police officers and crime, as

1

For example, see: Marvell, Thomas, and Carlisle Moody. "Specification
Problems, Police Levels, and Crime Rates," Criminology 1996, 34. pp.
609-46. Eck, John, and Edward Maguire, 2000. "Have Changes in Policing
Reduced Violent Crime? An Assessment of the Evidence." in A. Blumstein and
J. Wallman, eds., The Crime Drop in America. New York: Cambridge
University Press, 2000. pp 207-65.

2

Sherman, Lawrence, 1998. "Policing for Crime Prevention," in Sherman, L.,
et al. (eds.) Preventing Crime: What Works, What Doesn't, What's
Promising: A Report to the United States Congress. Washington, D.C.:
National Institute of Justice, Chapter 8.

municipalities having higher crime rates generally also have more
officers. For example, Detroit has twice as many police per capita as
Omaha and four times the violent crime rate, but it would be incorrect to
conclude that the additional officers in Detroit were the cause of its
higher crime rate than Omaha's. 3 By simply comparing a municipality's
police force and crime rate to those in other municipalities, one would
incorrectly infer that Detroit's higher crime rate was caused by its
additional police officers.

Repeated observations on crime and police in a locality lead to a more
robust research design by controlling for the time-invariant differences
in rates of crime and police between areas. This is done by introducing
fixed effects into regression models. Using this approach, the question
that the analysis attempts to address becomes: Do we see the crime rate
fall as the number of police rises? By controlling for the "baseline"
crime rates in different areas, some researchers have estimated a negative
relationship between police and crime. 4

However, if the rise in the number of police in a locality is a response
to increasing crime rates, including fixed effects does not resolve the
issue of reverse causality raised by the Detroit example. A next step is
to introduce an instrument-for example, a variable that affects the size
of the police force but that, given this size, does not affect crime. In
one study, the researcher made use of the fact that the size of a police
force increases before an election. If the only way that crime is affected
by the election is through the number of police, then this approach can be
used to estimate the relationship between crime and police. In this study,
the researcher found that crime fell in several index categories before an
election. 5

A series of more recent papers that used instruments found a negative
relationship between police and crime. Two studies used an increase in

3

Levitt, Steven D. "Using Electoral Cycles in Police Hiring to Help
Estimate the Effect of Police on Crime." American Economic Review, 87
(1997): 270-290.

4

See, for example, Levitt (1997) and Marvel and Moody (1996).

5

McCrary (2002) found that Levitt's estimation of standard errors suffered
from a computational error. Levitt (2002) was able to confirm his results
when the error was corrected by using an alternative instrument-the number
of municipal workers and firemen. McCrary, Justin. "Using Electoral Cycles
in Police Hiring to Estimate the Effect of Police on Crime: Comment."
American Economic Review. June 2002, 92(4), pp. 1236-43. Levitt, Steven D.
"Using Electoral Cycles in Police Hiring to Estimate the Effects of Police
on Crime: Reply" American Economic Review, September 2002, 92(4), pp.
1244-50.

Our Approach to Estimating the Effects of COPS Expenditures on Officers
and Crime

police presence because of a terrorist alert and showed declines in
nonterrorist-related crimes within a single city. In a study of Buenos
Aires, the researchers found that police stationed in response to a
terrorist threat on Jewish centers caused a decline in automobile theft. 6
In another paper, the researchers showed that crime fell in Washington,
D.C., on days when the Department of Homeland Security increased the
terror alert level. 7 At the national level, researchers at the University
of Maryland used the number of police officers granted through the COPS
program as an instrument for the actual number of police and estimated
negative relationships between increases in police officers and crime. 8

We adopted a two-stage approach to estimating the effects of COPS
expenditures on crime. Much as the University of Maryland researchers did,
we used COPS funds as a source of variation to explain officers. However,
while the University of Maryland researchers used officers granted by COPS
funds, we used COPS expenditure amounts-the actual COPS dollars spent by
agencies in given years-as the source of variation. We began with an
analysis of the "first stage" and tested whether COPS funds had an effect
on the number of officers. To the extent that hiring funds affected the
number of police but did not affect crime in any other way, these funds
would be a valid instrument for estimating the effect of officers on
crime. We then estimated the "reduced form," or the relationship between
COPS expenditures and crime. Using parameters estimated from these
regressions, we are able to calculate the relationship between police and
crime.

This approach has limitations, however. For example, we learn very little
about how agencies operate. If agencies were to use the additional
officers to employ different police tactics, and were able to reduce
crime, we would be unable to say whether it was the increase in officer
numbers or tactics that was the true cause of the decrease. Thus, we would
be unable

6

Di Tella, Rafael, and Ernesto Schargrodsky. "Do Police Reduce Crime?
Estimates Using the Allocation of Police Forces after a Terrorist Attack."
American Economic Review. March 2004, 94(1). pp. 115-133.

7

Klick, Jonathan, and Alexander Tabarrok. "Using Terror Alert Levels to
Estimate the Effect of Police on Crime." Journal of Law and Economics,
April 2005, vol. XLVIII.

8

Evans, William N., and Emily Owens. "Flypaper COPS," College Park,
Maryland: University of Maryland. Available online
www.bsos.umd.edu/econ/evans/wpapers/Flypaper%20COPS.pdf, 2005.

to contribute to the question of whether increases in officer strength are
either necessary or sufficient to reduce crime, without a change in police
tactics.

A second concern is that agencies that were more likely to take initiative
in applying for and receiving COPS grants might be those that were also
more effective in preventing crime. These agencies might also be those
that achieved larger or more rapid declines in crime. If this were the
case, we might incorrectly associate declines in crime with COPS grant
expenditures because of other possible factors. To assess this potential,
we estimated a regression that predicted whether an agency spent COPS
funds in a given year from 1994 through 2001 based on demographic
characteristics, economic conditions, and lagged property and violent
crime rates. From the regressions, we predicted the probability of
spending COPS grant funds-or the propensity of agencies to spend COPS
funds. Whether or not an agency actually spent COPS funds, it received a
propensity score, based upon the values of its characteristics in the
model that predicated the probability of spending COPS funds. Agencies
that actually spent COPS funds can then be compared to similar agencies-
those with similar propensity scores-that did not spend COPS funds. We
grouped agencies into five categories based on their propensity scores.
Within each of these five categories, we compared the patterns of violent
crime rates and property crime rates between the agencies that spent COPS
funds and those that did not spend them. Our analysis showed that within
these groupings of agencies having similar propensity scores, the agencies
that actually spent COPS funds generally had larger declines in crime
rates than did those that did not spend COPS funds.

Another question is whether a drop in a specific crime type, such as
automobile theft, in a certain locality is a net gain for society as a
whole. For example, the rationality of criminals may lead them to respond
to an increase in the number of police by moving to an area with fewer
police or switching to a different type of crime. 9 In addition, there is
the possibility that an increase in the number of police increases the
reporting rate of crimes, and not the crimes themselves. 10 This
possibility, however, would

9

Cook, Philip, "The Clearance Rate as a Measure of Criminal Justice System
Effectiveness," Journal of Public Economics, Vol. 11, 1979, pp. 135-142.

10

Swimmer, Eugene, "The Relationship of Police and Crime: Some
Methodological and Empirical Results," Criminology, Vol. 12, 1974: pp.
293-314.

Page 75 GAO-06-104 Community Policing Grants

Model of the Effect of COPS Expenditures on the Number of Police Officers

lead us to underestimate the effects of COPS funds on crime, as discussed
in appendix I.

Our main specification estimated the effect of COPS funds on officers,
using the following control variables:

(1) POLICEit= bHIRE+ b2MORE+ bINNOVit+ bMISC + b5BYRNEDIS +

1itit34it it

b6LLEBG+ b7NONCOPS+ gXit+ a+ a + a + (quartile of prior growth

it it i tst

rates) * (population stratification) * year

Where

     o POLICEit is the dependent variable, the sworn officers per 10,000 in
       population in agency i in year t;
     o HIREit is the amount of money paid in Hiring grants; and
     o MOREit are COPS MORE grants; INNOVit are COPS grants for innovative
       policing, and MISCit refers to the remaining types of COPS grants; all
       are expressed as expenditure in per capita amounts.
     o BYRNEDISit are Byrne discretionary grant expenditures, 11 LLEBG are

it

LLEBG grant amounts, and NONCOPSit are all other federal non-COPS law
enforcement grants; all are expressed in per capita amounts. We introduce
these variables to control for other federal funds.

o  Xit contains a number of demographic and economic control variables,
including local employment rates, per capita income, and population
composition variables that measured the percentage of population 15 to 24
years old and the percentage of the population that was nonwhite. The
economic and demographic controls were measured at the level of

11

Because Byrne formula grants are passed through states to local agencies
and the methods to track the amount of Byrne dollars going to local
agencies are unreliable, we were unable to include Byrne formula grant
amounts in our models. Moreover, according to an Abt Associates evaluation
of Byrne formula grants, about 40 percent of the amounts passed through
the states to local law enforcement agencies went to multijurisdictional
task forces, thereby further complicating the task of tracking Byrne
discretionary grant expenditures to local law enforcement agencies. See
Dunworth, Terence, and Aaron J. Saiger, National Assessment of the Byrne
Formula Grant Program: Where the Money Went-An Analysis of State Subgrant
Funding Decisions Under the Byrne Formula Grant Program, Report 1,
Washington, D.C.: National Institute of Justice Research Report, December
1996.

the county within which a particular agency was located. The

parameters for these variables are represented by g.

We included state-by-year fixed effects-represented by ast-to correct for
changes in crime policy at the state level, such as changes in the number
incarcerated and changes in sentencing policy. We included agency fixed
effects-represented by ai-to capture time invariant differences across
agencies, and time fixed effects-represented by at-to capture changes
affecting the entire nation.

Because of how the money was distributed, there may be some concern that
our estimate of the effect of the COPS money on officers is biased. For
example, it might be that agencies that received a disproportionate share
of the money relative to their populations had the benefit of preexisting
positive growth of numbers of officers, in addition to possible declines
in crime. If the trends continued, we might be incorrectly associating
increases in officers or decreases in crime with the amount of COPS money
received, rather than these preexisting trends.

To address this concern, we separated the agencies into four groups, based
on the growth rate in both officers and crime during 1990-1993, when the
COPS program was introduced. We constructed each combination of these
groups, producing 16 cells. These cells were then "interacted" with each
year and four population categories, for a total of 768 effects. In
essence, each agency was compared with another agency that had a similar
"trajectory" of crime and officers in the pre-COPS period. 12 These growth
trends are represented by the (quartile of prior growth rates) expression
in equation (1).

Finally, to obtain estimates of the effects of COPS expenditures on
officers relative to the average person in the United States, we estimated
weighted regressions where the weights were the population served by an
agency.

Because of these effects, the parameters of interest, b1 though b4, are
the effect of the COPS funds once other federal funds, demographic and
economic conditions, time and agency fixed effects, and these "growth
rate" effects are controlled for.

This approach was proposed by Evans and Owens (2005).

Model of the Effect of COPS Expenditures on Crime

As with our methodology in estimating the effect of COPS funds on
officers, we estimate the effect of COPS funds on crime. Our main
specification used the following controls in the following equation:

(2) CRIMEit = u1HIREit + u2MOREit + u3INNOVit+ u4MISCit + u5BRYNEDIS+

it

u6LLEBGit + u7NONCOPS+ pXit + d + d + d + (quartile of prior growth

it itst

rates) * (population stratification) * year

The independent variables are identical to those defined for equation (1).
The dependent variable (CRIMEit) is the UCR total-or index-crime rate. We
also estimate separate equations for the crime rates of components of the
crime index: murder and non-negligent manslaughter, forcible rape,
robbery, aggravated assault, burglary, larceny theft, and motor vehicle
theft. 13 Again, the parameters of interest are u1 through u4.

As in equation (1), the economic and demographic covariates in equation

(2) are represented by Xit; di, d, and d represent the agency, year, and

tst

state-times-year fixed effects; and we also include the pre-1993 growth
rate variables.

The Implied Relationship between Police Officers and Crime

Unlike the other COPS grant types, COPS hiring grants were to be used
specifically for hiring officers. Consequently, variation in the number of
officers coming from COPS hiring grants should be unrelated to other
changes in police expenditures. In this sense, it may be a valid
instrument for officers.

Using the coefficients of officers in equations (1) and (2), we calculated
an estimate of the change in crime with respect to change in officers:
(u1/b1), u1 and b1 are the coefficients from equations (1) and (2).

The elasticity is a measure of the percentage change in crime derived from
a percentage change in police. We used coefficients of officers in
equations (1) and (2) to calculate an estimate of the elasticity of crime
with respect to officers in 1993:

(3) ELASTICITY = (u1 / b1)*(POLICE1993/CRIME

1993)

13

We excluded arson from our analysis, because of limited reporting of this
crime to the UCR, as indicated by the FBI. (See app. I.)

Page 78 GAO-06-104 Community Policing Grants

where

     o u1 and b1 are the coefficients from equations (1) and (2) and
     o POLICE1993 and CRIME1993 are average police strength and crime rates
       for 1993.

To test the robustness of our estimates under different assumptions about
how COPS grant expenditures are related to officers and crime, we
estimated the elasticity of crime with respect to officers under a number
of different specifications, as described in table 14.

Table 14: Alternate Specifications of the Relationship between COPS
Expenditures and Crime

                      Variable in specification 1 2 3 4 5

MORE, Innovative, and Miscellaneous COPS expenditures x x

LLEBG, Byrne discretionary, and other federal non-COPS x x expenditures

"Got grant" specification

Lagged values of MORE, Innovative, and Miscellaneous COPS x x expenditures

Lagged values of LLEBG, Byrne discretionary, and other federal x x
non-COPS expenditures

                  Demographic and economic controls x x x x x

Growth rate cells x x

                  Lagged value of Hiring grant expenditures x

                 Quadratic term for Hiring grant expenditures x

                      State by year fixed effects x x x x

Source: GAO analysis.

Note: An X indicates that a variable was included in a specification.

Other than the "got grant" specification, all variables are as defined
above. Including the "got grant" variable provides a test for whether the
effects of COPS grants occurred in the year in which the money was
actually spent-as we specified in equations (1) and (2)-or whether the
announcement of a grant award led to changes in officers and,
subsequently, crime. If the announcement of the award were more important
than the actual expenditures, it would imply that estimates of the effect
of changes in expenditures on officers or crime in equations (1) and (2)
would overstate the effects. To address this, we added indicator variables
for the year in which a grant was received. Additionally, the

Data Used in Our Analysis

quadratic term for COPS hiring grant expenditures provides a test for
nonlinear effects of COPS hiring grants on crime. This specification
examines whether the effects of officers on crime diminish as the number
of officers rises above certain levels.

We use data on 4,247 police agencies that reported complete crime (12
months of crime) in any year and that served populations of 10,000 or more
persons. These agencies represented about 23 percent of the agencies that
appeared in the UCR data that we received from the FBI. However, they also
covered more than 86 percent of the crimes and they represented about 77
percent of the population in the UCR data that we received. Because of
concerns about data quality, we restricted our sample to agencies that met
these criteria of complete crime reporters and serving populations larger
than 10,000 persons. Across years, the number of agencies that met these
conditions varies, so our panel of data is unbalanced. We used grant
expenditure data from the OJP financial data, which we linked to the crime
and officer records of agencies. We included county level demographic and
economic data from the Census Bureau, the National Center for Health
Statistics, and the Bureau of Economic Analysis. (See app. I for more
information regarding the construction of the dataset.)

Table 15 provides the means and standard deviations of the variables
included in the regression models. As shown in the table, the per capita
expenditures derived from COPS hiring grants exceeded the per capita
amounts from other federal grants.

 Table 15: Means and Standard Deviations of Variables Used in Regression Models

Standard Variables Mean deviation

                    Officers per 10,000 persons 20.31 12.37

                     Federal grant expenditures per capita

                             COPS hiring 0.978 2.18

                              COPS MORE 0.292 1.35

                          COPS innovative 0.082 0.496

                         COPS miscellaneous 0.003 0.043

                        Byrne discretionary 0.045 0.471

                                LLEBG 0.770 1.93

                   Crime rate variables (per 100,000 persons)

                         Total index crime 5,349 3,170

                                Murder 8.7 10.9

                              Forcible rape 38 31

                                Robbery 247 317

                           Aggravated assault 424 391

                               Burglary 1,034 647

                           Larceny theft 2,990 1,752

                          Motor vehicle theft 608 593

                            Other control variables

                        Log per capita income 10.12 0.33

                   Employment-to-population ratio 0.631 0.453

             Fraction of population aged 15 through 24 0.141 0.027

                  Fraction of population nonwhite 0.186 0.136

Source: GAO analysis of Uniform Crime Report, Office of Justice Programs
financial, Bureau of Economic Analysis, National Center for Health
Statistics, and U.S. Census Bureau data.

In this section, we discuss our regression analyses and describe how we
arrived at the results that are discussed in this report. Results of Our
Analysis

The Effect of COPS To arrive at the effects of COPS expenditures on
officers, we estimated Expenditures on the specifications for equation
(1), as shown in table 16. With only the fixed Number of Police Officers
effects, the models explain more than 90 percent of the variation in
officer

strength. In specification 1, we added only the COPS hiring grant

Page 81 GAO-06-104 Community Policing Grants

expenditures per capita to the model that contained only the fixed
effects. The effects of hiring grants are significant at the 1 percent
level, and the coefficient indicates that an additional dollar of hiring
grant expenditures per capita changes the officer rate (measured per
10,000 persons) by 0.317. In specifications 2 through 5, we introduce
various combinations of the growth rate cells, demographic and economic
conditions, and the other grant types. Across specifications 2 through 5,
the estimated coefficient on the hiring grant variable remains fairly
consistent, ranging from 0.227 in specification 5 to 0.261 in
specification 3, where the interpretation of the coefficient is the effect
of a $1 increase in per capita COPS hiring grant on the per 10,000 person
rate of officers. Specification 5 presents our preferred specification, in
that it includes all of the relevant controls. Using the coefficient on
COPS hiring grant expenditures from specification 5, we calculate the
effect of $25,000 in COPS hiring grant expenditures in a given year to
produce roughly 0.6 additional officers in a given year. 14 Finally, in
addition to the COPS hiring grant expenditures, COPS MORE and LLEBG grant
expenditures also consistently predict officer strength, as indicated by
the MORE and LLEBG parameter estimates in specifications 2 through 5.

Bearing in mind that the officer strength is per 10,000 in the population,
we arrive at this result by the following calculation:
(25,000)*(.227/10,000) = 0.57 officers.

Page 82 GAO-06-104 Community Policing Grants

Effect of COPS Expenditures on Crime

Table 16: Parameter Estimates from Regressions of Officers Per Capita on
COPS Hiring Grant Expenditures and Other Outside Funds (Standard Errors in
Parentheses)

                               Variable 1 2 3 4 5

Hiring 0.317 0.231 0.261 0.247 0.227 (0.055) (0.025) (0.047) (0.028)
(0.025)

MORE 0.124 0.238 0.159 0.121 (0.043) (0.090) (0.054) (0.043)

Innovative 0.0477 -0.029 0.042 0.047 (0.050) (0.075) (0.054) (0.050)

                       Miscellaneous 1.46 0.906 1.13 1.43

(1.20) (1.30) (1.28) (1.19)

                         Byrne 0.001 0.169 0.148 0.0003

(0.06) (0.129) (0.102) (0.06)

                         LLEBG 0.172 0.259 0.201 0.168

(0.05) (0.065) (0.049) (0.049)

Federal non-COPS 0.056 0.022 0.033 0.053 (0.045) (0.066) (0.047) (0.045)

Demographic and economic No No Yes Yes Yes covariatesa

                     Population weights Yes Yes Yes Yes Yes

                      Growth rate cells No Yes No Yes Yes

                  State-by-year fixed effects No Yes No No Yes

Source: GAO analysis of Uniform Crime Report, Office of Justice Programs
financial, Bureau of Economic Analysis, National Center for Health
Statistics, and U.S. Census Bureau data.

Notes: Officers per capita is measured in terms of officers per 10,000
persons; all expenditure variables are in per capita amounts. All
regression specifications include agency and year fixed effects. Bold-face
parameter estimates and standard errors indicate that a parameter estimate
is statistically significant at the 5 percent level using robust standard
errors.

a

Demographic and economic covariates include log per capita income,
employment to population ratio, percentage of population between 15 and 24
years of age, and percentage of population that is nonwhite.

Our reduced-form estimates of the effects of COPS expenditures on crime,
the result of our estimating equation (2) appear in table 17. This first
column (labeled "Officers") repeats the results from specification 5 of
table 16. The other columns of table 17 show the parameter estimates for
the effects of hiring grants and outside funds on the crime rate for index
crimes and separately for type of index crime (except for arson). With the
exception of rape, COPS hiring grant expenditures per capita have a
negative effect on index crime rates and the crime rate for each type of
index crime. Further, while the direction of the effect of the hiring
grant variable on the larceny rate is negative, the effect is not
significant at the 5 percent level. LLEBG expenditures have a negative and
significant effect on all crime types. The other grant fund types have a
negative effect on some crime types.

We estimated the effect of COPS hiring grant expenditures on index crimes
to be -29.19. In other words, $1 in COPS hiring grant expenditures per
capita translates into a reduction of almost 30 index crimes per 100,000
people.

Table 17: Parameter Estimates from Regressions of Crime Rates on COPS
Hiring Grant Expenditures and Other Outside Funds (Standard Errors in
Parentheses)

Motor Officers Index Murder Rape Robbery Assault Burglary Larceny vehicle

                                 1 2 34 56 7 89

Variable

Hiring 0.227 -29.19 -0.1330.128-4.94 -2.77-8.01 -4.18 -9.26 (0.025) (6.67)
(0.028) (0.075) (1.07) (1.08) (1.33) (3.05) (3.44)

MORE 0.121 -17.14-0.083 0.008-2.80 -1.72-2.04 -6.91 -3.58 (0.043) (6.55)
(0.031) (0.063) (0.919) (0.86) (1.14) (3.43) (1.51)

Innovative 0.047 -88.25 -0.219 -0.102 -8.45 -9.71 -17.62 -23.30 -28.8
(0.050) (17.80) (.081) (.255) (2.13) (3.80) (4.81) (11.5) (6.77)

      Miscellaneous 1.43 -123.7 1.13 2.37 41.2 -13.56 -90.61 -121.8 57.51

(1.19) (18.79) (.887) (2.31) (29.74) (33.27) (36.15) (101.4) (46.98)

          Byrne 0.0003 11.72 -.099 -0.388 .270 7.01 -0.172 10.16 -5.25

(0.06) (16.03) (.069) (.280) (1.61) (1.33) (3.87) (10.14) (3.39)

LLEBG 0.168 -73.13 -0.365 -0.784 -13.07 -16.00 -16.06 -15.2 -11.59 (0.049)
(10.60)

(.051) (.132) (1.68) (2.09) (2.31) (3.87) (2.20)

Federal non-0.053 22.96 .027 .082 2.25 1.57 1.34 10.40 7.28 COPS (0.045) (9.14)
                (.038) (.090) (1.07) (1.34) (1.60) (5.07) (1.90)

Source: GAO analysis of Uniform Crime Report, Office of Justice Programs
financial, Bureau of Economic Analysis, National Center for Health
Statistics, and U.S. Census Bureau data.

Notes: All regressions include agency and year fixed effects,
state-by-year fixed effects, and growth rate cells. Additionally,
regressions include log per capita income, employment over population
ratio; percentage of county population aged 15 to 24; and percentage
nonwhite. Officers are per 10,000 persons; all grant expenditures are per
capita amounts. Observations are weighted by the population of the agency
to obtain the national effect. Bold-face parameter estimates and standard
errors indicate that a parameter estimate is statistically significant at
the 5 percent level using robust standard errors.

The Effects of Different Population Sizes across Agencies

Given the variation in per capita COPS expenditures that occurred across
agencies serving populations of different sizes, we explored whether COPS
hiring grants had different effects on crime rates based on the size of
the population served by agencies. We stratified agencies into four
population size groups: those serving populations of between 10,000 and
25,000 persons; between 25,000 and 50,000 persons; between 50,000 and
150,000 persons; and more than 150,000 persons. We found that the effect
of the hiring grant was consistent across all population categories less
than 150,000, but insignificant in the population category of more than
150,000 persons. We found that negative effect of COPS hiring grants on
index crime rates ran across all population size categories. However, the
effects of hiring grants were largest in the 50,000 to 150,000 population
category, and insignificant in the 25,000 to 50,000 population category
(table 18).

Table 18: Parameter Estimates from Regressions of Index Crime Rates and
Officers Per Capita on COPS Hiring Grant Expenditures and Other Outside
Funds, by Population Size Category (Standard Errors in Parentheses)

                                   Population

10,000 to fewer than 25,000 to fewer 50,000 to fewer than 25,000 than
50,000 150,000 More than 150,000

          Officers Index Officers Index Officers Index Officers Index

                                  1 2 3456 78

Grant

Hiring .180 -10.11 .288 -8.79 .245 -39.1 .095 -31.5 (.019) (4.74) (.032)
(10.00) (.034) (10.1) (.074) (15.2)

MORE .043 2.86 .027 -14.67 .102 3.79 .053 -35.2 (.021) (2.32) (.027)
(13.25) (.069) (14.16) (.148) (24.0)

Innovative -.007 -19.94 -.058 -28.8 .036 -87.7 -.043 -108 (.043) (20.67)
(.137) (36.9) (.052) (23.3) (.130) (48.32)

Miscellaneous -.282 -379 -.338 -473 -.996 -145 4.79 -161 (.360) (179)
(.692) (234) (.574) (231) (1.97) (368)

Byrne -.010 -7.87 .440 -4.37 -.084 24.88 .173 40.2 (.074) (12.74) (.516)
(55.52) (.084) (29.91) (.153) (26.5)

LLEBG .010 -23.08 .032 -141.3 -.012 -109 .492 -90.4 (.013) (6.87) (.046)
(18.28) (.069) (15.6) (.176) (22.6)

  Federal non-.031 -1.177 -.087 20.06 -.016 36.06 .045 11.4 COPS (.016) (4.74)
                  (.113) (16.79) (.096) (13.03) (.127) (32.6)

Source: GAO analysis of Uniform Crime Report, Office of Justice Programs
financial, Bureau of Economic Analysis, National Center for Health
Statistics, and U.S. Census Bureau data.

Page 85 GAO-06-104 Community Policing Grants

Calculations of the Elasticity of Crime with Respect to Officers

Notes: All regressions include agency and year fixed effects,
state-by-year fixed effects, and growth rate cells. Additionally,
regressions include log per capita income, employment over population
ratio; percentage of county population aged 15 to 24; and percentage
nonwhite. Officers are per 10,000 persons; all grant expenditures are per
capita amounts. Observations are weighted by the population of the agency
to obtain the national effect. Bold-face parameter estimates and standard
errors indicate that a parameter estimate is statistically significant at
the 5 percent level using robust standard errors.

As COPS hiring grants were to be used only to hire officers, we explored
their use as an instrument to predict the effect of officers on crime.
Assuming that COPS grants were used in that way, our preferred
specification from our regressions crime on COPS hiring grants and other
outside funds would produce estimates of the elasticity of crime with
respect to officers that are shown in table 19.

To assess the degree to which the elasticities that we calculated were in
line with those appearing in the economics of crime literature, we
compared our elasticities with those estimated by Evans and Owens (2004),
Levitt (1997), Levitt (2002), and Klick and Tabarrok (2005). Our estimates
are in line with those in the literature (table 19).

Table 19: Elasticities of the Impact of Police Officers on the Crime Rate
                                                                        Motor
Elasticity           Murder Rape  Robbery Assault Burglary Larceny vehicle 
                             1     2       3       4        5       6       7 
Estimate                                                           
Average crime rate      11     40     311     484     1183    3173     703 
1993                                                               
Levitt 1997a         -1.98 -0.27    -0.79  -1.09     -0.05   -0.43   -0.50 
Levitt 2002          -0.91 -0.03    -0.45    0.40    -0.20   -0.14   -1.70 
Evans and Owens      -0.84 -0.42    -1.34  -0.96     -0.59   -0.08   -0.85 
2005b                                                              
GAO (this report)    -1.04   0.28  --1.36  -0.49     -0.58   -0.11   -1.12 
Klick and Tabarrok                                                   -0.30 
2005                                                               
GAO aggregate elasticity, by                                       
crime category                                                     
Index                -0.42                                         
Violent              -0.78                                         
Property             -0.36                                         

Source: GAO analysis of Uniform Crime Report, Office of Justice Programs
financial, Bureau of Economic Analysis, National Center for Health
Statistics, and U.S. Census Bureau data.

Notes: Estimates are derived from the parameter estimates in tables 16 and
17. The average police count per 10,000 in 1993 is 19.38. Crime is per
100,000.

aLevitt's (1997) elasticities are taken directly from his regression
specification. Levitt calculates elasticities for a range of alternate
specifications that are not reported here.

Page 86 GAO-06-104 Community Policing Grants

b

Evans and Owens' (2005) elasticities were evaluated at the same mean level
of crime as were GAO's.

In addition, Evans and Owens report aggregate point elasticities for
violent and property crimes of -0.99 and -0.26, respectively, and Levitt
reports aggregate point elasticities for violent and property crimes of
-0.44 and -0.50, respectively. Our aggregate elasticities for violent and
property crimes fall between these two sets of estimated point
elasticities.

Equations (1) and (2) depend on certain assumptions about the way that
COPS hiring grant expenditures and other outside funds affect officers and
crime. For example, the specifications reported previously only allow the
effect of the federal funds to affect crime contemporaneously. However, it
may take a certain amount of time for the expenditures to have an effect
on either officers or crime, as it may take a certain amount of time for
new officers to become fully acclimated to a department, or to become
proficient in their duties. To explore the robustness of our findings
under varying assumptions about how COPS hiring grant expenditures could
affect officers and crime, we recalculated our elasticities after
estimating our regressions under the specifications outlined previously in
table 20. We report the elasticities that we calculated from these various
regression models (in the last three rows of the table). The elasticities
for index crimes range from -0.41 to -0.95; those for violent crimes range
from -0.76 to -1.8; and those for property crimes range from -0.35 to
-0.8. The elasticities that we report in our results all fall at the lower
end of the range of elasticities that we estimated.

Estimating the Net Number of Officers Paid for by COPS Expenditures

Table 20: Elasticity of Violent and Property Crime with Respect to
Officers under Alternate Specifications of the Relationship between COPS
Expenditures and Crime

                      Variables in specification 1 2 3 4 5

MORE, Innovative, and Miscellaneous COPS x x expenditures

LLEBG, Byrne discretionary, and other federal non-x x COPS expenditures

"Got grant" specification

Lagged values of MORE, Innovative, and x x Miscellaneous COPS expenditures

Lagged values of LLEBG, Byrne discretionary, and x x other federal
non-COPS expenditures

                  Demographic and economic controls x x x x x

                           Growth rate cells x x x x

                  Lagged value of Hiring grant expenditures x

                 Quadratic term for Hiring grant expenditures x

                      State by year fixed effects x x x x

                                   Elasticity

                    Violent crimes -1.17 -.76 -1.8 -.81 -.76

                    Property crimes -.51 -.35 -.80 -.37 -.35

                     Index crimes -.61 -.41 -.95 -.44 -.41

Source: GAO analysis of Uniform Crime Report, Office of Justice Programs
financial, Bureau of Economic Analysis, National Center for Health
Statistics, and U.S. Census Bureau data.

Note: An X indicates that a variable was included in a specification.

We used our regression results to derive estimates of the net number of
officers paid for by COPS grant expenditures separately for each year. By
net number of officers, we refer to the increase in the number of officers
on the street attributable to COPS net of attrition. For example, if at
the beginning of a year, there were 100 officers on the street, while
during a year COPS grants were responsible for hiring 10 officers and 5
officers left the force, the net number of officers due to COPS would be
5.

To obtain the total number of officer-years due to COPS expenditures, we
summed the number of officers across years. Table 21 presents the
estimated number of officers that COPS expenditures funds paid for in each
year. In column 1 we present the actual number of per capita officers used
in our regressions that generated the results in table 21. Not shown in
the table, but used in the calculation of the number of officers due to
COPS expenditures are the per capita amounts of COPS expenditures,
including COPS hiring, MORE, innovative, and miscellaneous grant
expenditures. Column 2 presents our estimate of what the per capita number
of officers would have been absent the COPS expenditures. Columns 3 and 4
show the number of officers per capita and the percentage of officers per
capita explained by COPS expenditures. Column 5 presents our estimates of
the number of officers in each year in the sample of agencies that we
analyzed that were explained by COPS expenditures. To arrive at the number
of officers in the United States due to COPS expenditures, we weighted the
numbers in column 5 up to the

U.S. population total (in column 6). On the basis of this analysis, in
year 2000, for example, when they peaked, the COPS expenditures per capita
were responsible for about 2.9 percent of the net increase in officers in
the United States, or more than 17,000 officers. Across all years, we
estimate that COPS was responsible for an increase of about 88,000
officer-years during the years from 1994 through 2001.

Table 21: Estimated Per Capita Effect of COPS Expenditures on the Number
of Officers
          Police per        Police explained by
      10,000 population              COPS funds          Number of police in
                                              Percentage 
               Number Minus COPS Per capita  difference   Sample       United 
                      funds                                            States 
Year             1          2           3           4       5            6 
1991         19.32      19.32           0           0       0            0 
1992         19.32      19.32           0           0       0            0 
1993         19.38      19.38           0           0       0            0 
1994         19.65      19.65    0.003           0.02      64           84 
1995         20.55      20.47        0.07        0.35   1,407        1,916 
1996         20.71      20.39        0.32        1.55   6,210        8,639 
1997         21.05      20.54        0.51        2.42  10,085       13,897 
1998         21.18      20.54        0.64        3.02  12,900       17,630 
1999         21.61      21.02        0.59        2.72  12,153       16,415 
2000         21.15      20.53        0.62        2.91  13,335       17,387 
2001         20.89      20.46        0.43        2.05   9.535       12,226 
Total                                                  65,688       88,195 

Source: GAO analysis of Uniform Crime Report, Office of Justice Programs
financial, Bureau of Economic Analysis, National Center for Health
Statistics, and U.S. Census Bureau data.

Estimating the Number of Crimes Reduced by COPS Expenditures

On the basis of our analysis of the increase in officers attributable to
COPS expenditures, we estimated the amount of crime that could be
attributable to COPS, given the estimated effect of COPS expenditures on
officers. On the basis of our analysis of the number of officers due to
COPS expenditures and our estimated elasticities of crime with respect to
officers, we can estimate the number of crimes associated with COPS
expenditures through the increase in officers attributable to these
expenditures. In table 22, we show our calculations of the decline in
crime attributable to COPS for each year, compared with the 1993 levels of
crime, the pre-COPS baseline year.

Columns 1 through 3 of table 22 give the average crime rates per 100,000
persons in the agencies in our sample. Columns 4 through 6 give the
percentage change from 1993 in crime rates for each category of crime.
Columns 7 though 9 report data on officers. Column 7 reports the growth in
the officer rate from 1993 due to the change in COPS expenditures. Column
8 presents the growth (from column 7) as a percentage change from 1993.
Columns 9 through 11 provide estimates of the percentage change in crime
rates from 1993 using the elasticities shown in table 22. Finally, columns
12 through 14 provide the estimated amount of change in crime rates from
1993 that arise from COPS expenditures.

Table 22: Estimated Per Capita Growth of COPS Expenditures on Police
Officers and Crime from 1993

Average number Percentage change of crimesa in crimeb

                 Violent Property Total Violent Property Total

                                  Year 123456

                              1991 868 5519 6,387

                               1992 854 5235 6090

                               1993 846 5058 5904

                      1994 816 4973 5789 -3.55 -1.68 -1.95

                      1995 784 4919 5703 -7.33 -2.75 -3.42

                     1996 723 4718 5440 -14.54 -6.72 -7.86

                     1997 697 4593 5290 -17.61 -9.19 -10.40

                    1998 649 4313 4962 -23.29 -14.73 -15.96

                    1999 588 3947 4535 -30.50 -21.97 -23.19

                    2000 568 3799 4367 -32.86 -24.91 -26.03

                    2001 561 3845 4406 -33.69 -23.98 -25.37

                         Expected percentage        Change in crime per
Number of officers c  change in crime due        100,000 persons
                         to COPS funds              
              Percentage                                                      
Predictedd change      Violent   Property  Total Violent  Property   Total
            7          8          9    10      11         12    13         14 

19.38     
19.38       .02     -.01     -.01     -.01     -.09      -.25         -.34 
19.45       .37     -.29     -.13     -.16     -2.08     -5.90       -8.01 
19.70      1.65    -1.29     -.60     -.70     -9.56    -27.11      -36.75 
19.89      2.63    -2.05     -.95    -1.11    -15.48    -43.86      -59.47 
20.02      3.30    -2.57    -1.19    -1.39    -19.67    -55.74      -75.58 
19.97      3.04    -2.36    -1.10    -1.28    -19.26    -54.57      -74.00 
20.00      3.18    -2.48    -1.15    -1.34    -20.19    -57.22      -77.59 
19.81      2.21    -1.72     -.80     -.93    -14.08    -39.90      -54.10 

Source: GAO analysis of Uniform Crime Report, Office of Justice Programs
financial, Bureau of Economic Analysis, National Center for Health
Statistics, and U.S. Census Bureau data.

a

Average number of crimes per 100,000 for the agency; means are weighted by
population.

b

Percentage change in crime from 1993.

cColumn 7 is the predicted level in the number of officers from only a
change in COPS funds from 1993; column 8 is the percentage change from
1993.

dPredicted number of officers due to growth in COPS funds, from base 1993
level of officers of 19.38 per 10,000 persons.

Appendix VII: Methods Used to Assess Policing Practices

Methods to Address Changes in Policing Practices

Our objective in assessing policing practices was to determine the extent
to which COPS grant expenditures were associated with police departments'
adoption of policing activities or practices that may have contributed to
reduction in crime during the 1990s. To determine whether COPS grants were
associated with changes in policing practices, we analyzed data from two
national surveys of local law enforcement agencies on the policing
practices that they reportedly implemented in various years from 1993 to
2000. In addition, we analyzed systematic reviews of research on the
effectiveness of policing practices in preventing crime.

To address whether COPS grants were associated with changes in policing
practices that may be associated with preventing crime, we analyzed data
from the two administrations of the Policing Strategies Survey (in 1993
and 1997) and two of the four administrations of the National Evaluation
of COPS Program Survey (in 1996 and 2000). Because the purposes of the
surveys differed, each used different samples of agencies (with some
agencies appearing in both surveys). The Policing Strategies Survey drew a
sample representative of all municipal police, county police, and county
sheriff agencies in the United States with patrol functions and with more
than five sworn officers in 1992, and the National Evaluation of COPS
Program Survey drew a sample that was representative of all law
enforcement agencies believed to be in existence in the United States that
had received, or were eligible to receive a COPS grant. Each survey
provided respondents in police agencies with lists of items that
identified specific types of policing practices, and respondents were
asked whether they had implemented each of the practices on the list.
Survey responses were obtained from knowledgeable officials within each
agency, such as the police chief or the chief's designee. The number of
items related to policing practices differed between the two surveys.

We classified items in the surveys into four categories of policing
practices corresponding to general approaches to policing identified in
the criminal justice literature: problem-solving practices, place-oriented
practices, community collaboration activities, and crime analysis
activities. Problem-solving practices call for police to focus on specific
problems and tailor their strategies to the identified problems.
Place-oriented practices include attempts to identify the locations where
crime occurs repeatedly and to implement procedures to disrupt these
recurrences of crime. Community collaboration practices include improving
citizen feedback about crime problems and the effectiveness of policing
efforts to address these problems. Crime analysis includes the use of
tools such as geographic information systems to identify crime patterns.
These tools may help an agency support other practices for preventing
crime, such as problem-solving and place-oriented practices.

Three social science analysts with research experience in criminal justice
independently reviewed the list of policing practice items in each survey
and placed each item in one of the four categories or determined that the
item did not fit in any of the four categories. Following initial
classification, the analysts met to discuss and address any
inconsistencies in their classification of items.

After classifying practices, we created an index to represent the total
number of problem-solving, place-oriented, community collaboration, and
crime analysis practices, and we gave each agency that responded to both
waves of a survey a score equal to the number of these practices that the
agency reportedly implemented in the survey years. We also identified, for
each agency, the number of practices in each of the four categories.

We then analyzed the levels and changes in reported practices within each
survey. Our analysis focused on the differences in levels of practices
reported by agencies that received COPS grants and those that did not
receive them. To assess the influence of COPS grant expenditures on
reported practices, we analyzed changes in reported practices as a
function of the per capita amounts of COPS dollars spent by agencies. For
agencies that did not receive COPS grants, we set their per capita COPS
expenditure amounts to zero.

A limitation of our analysis is that the surveys did not ask explicitly
about the extent to which each listed practice was implemented by law
enforcement agencies. Thus agencies that report that they had implemented
a specific practice may vary considerably, from sporadic use of the
practice among a subset of officers in the agency to more frequent use of
the practice throughout the agency.

Characteristics and Analysis of the Policing Strategies Survey

The Policing Strategies Survey was administered in 1993 and again in 1997.
The Police Foundation administered the 1993 wave of the survey, and ORC
Macro International, Inc. and the Police Executive Research Forum
administered the 1997 wave of the survey. 1 The sampling frame for both
the 1993 and 1997 waves consisted of 11,824 local police and sheriffs'
departments listed in the Law Enforcement Sector portion of the 1992
Justice Agency list developed by the U.S. Bureau of the Census. In
constructing the sampling frame, state police departments, special police
agencies, agencies that did not perform patrol functions, and agencies
with fewer than five sworn personnel were excluded from the larger list of
all law enforcement agencies. A total of 2,337 police and sheriffs'
departments were selected to be in the main sample for the 1993 survey,
and surveys were mailed to 2,314 of them after 23 agencies were found to
be out of scope before the surveys were mailed. 2 Follow-up mailings and
facsimile reminders were sent to nonrespondents. The overall response rate
for the 1993 survey was 71.3 percent. All of the agencies in the first
sample were then selected for participation in the 1997 survey. The survey
employed a multiphased data collection approach, using postal mail for the
first phase, followed by facsimile reminders, a second mailing, and
computer-assisted telephone interviewing for nonrespondents. The response
rate for the 1997 survey was 74.7 percent. A total of 1,269 agencies were
present in both the 1993 and 1997 surveys. The sample was a stratified
random sample with probability of inclusion varying by the number of sworn
personnel (5-9; 10-49; 50-99; and 100 or more sworn personnel). 3

1

The 1993 survey was designed to provide information on what was occurring
and what needed to occur in the development and implementation of
community policing. The 1997 survey was designed to provide information on
the most current practices and trends in community policing. See: A.
Rosenthal et al, Community Policing: 1997 National Survey Update of Police
and Sheriffs' Departments, ORC Macro and Police Executive Research Forum,
Washington D.C.: National Institute of Justice, April 2001.

2

Agencies were considered out of scope if they had fewer than five sworn
officers, no patrol function, or were a state police agency or other
"special" police agency.

3

When ORC Macro and the Police Executive Research Foundation drew the
sample for the 1997 wave of the survey, they discovered that instead of
excluding agencies with fewer than five sworn officers, the Police
Foundation had used information on the agencies' total number of employees
to select the agencies for the sampling frame and had excluded agencies
with fewer than five employees. Thus some agencies were misclassified, and
some were included that should not have been. In addition, the weights
provided with the 1993 data were incorrect for agencies with 10 to 49
employees. ORC Macro and PERF were able to assign the appropriate weights
retroactively to the 1993 sample and were able to exclude agencies with
fewer than five sworn officers.

We identified agencies in the Policing Strategies Survey that responded to
both waves of the survey and had complete data on each of the policing
practices items, and of these, we were able to link the data from 1,188
agencies to our larger database on crime, officers, money, and economic
conditions. 4 For comparability with the analyses of the effects of
funding on officers and crime, we limited our analysis to those agencies
serving jurisdictions with populations of 10,000 or more persons. This
resulted in usable data on 1,003 agencies.

We used the Policing Strategies Survey data to compare reported changes in
the types and levels of policing practices that occurred during the COPS
program with pre-COPS levels of practices. The analyses reported in this
appendix are weighted to adjust for the sample design effects. The
findings are generalizable to all municipal police agencies, county police
agencies, and county sheriff agencies in the United States with patrol
functions and serving jurisdictions with populations of 10,000 or more
persons.

We used 38 items on policing practices from the Policing Strategies
Survey. We combined 12 practices pertaining to increasing officer contact
with citizens and improving citizen feedback into a community
collaboration index. We used 6 items on the crime analysis units within
police departments to create our index of crime analysis. We combined 8
practices pertaining to increasing enforcement activity or place
management in buildings, neighborhoods, or other specific places into an
index of place-oriented practices. And we compiled the data on 12 items
that reflected organizational efforts to reduce or interrupt recurring
mechanisms that may encourage crime into a problem-solving practices
index. The classification of items from the Policing Strategies Survey
into our four indexes of types of policing practices is shown in table 23.

These agencies represented about 94 percent of the agencies that responded
to both waves of the Policing Strategies Survey.

Page 97 GAO-06-104 Community Policing Grants

Table 23: Categories of Policing Practices and Specific Items within Each
Category in the Policing Strategies Survey

                            Community collaboration

                Agency uses foot patrol as a specific assignment

Agency uses foot patrol as a periodic expectation for officers assigned to cars

    Agency uses citizen surveys to determine community needs and priorities

             Agency uses citizen surveys to evaluate police service

             Patrol officers conduct surveys in area of assignment

              Patrol officers meet regularly with community groups

          Supervisors maintain regular contact with community leaders

          Agency has permanent, neighborhood-based offices or stations

           Agency has mobile, neighborhood-based offices or stations

          Patrol officers make door-to-door contacts in neighborhoods

Patrol officers develop familiarity with community leaders in area of assignment

                 Patrol officers assist in organizing community

Crime analysis

            Agency has a decentralized crime analysis unit/function

             Agency has a centralized crime analysis unit/function

    Supervisors manage crime analysis for geographic area of responsibility

Geographically based crime analysis made available to officers at the beat level

         Patrol officers conduct crime analysis for area of assignment

Agency has means of accessing other city or county databases to analyze
community or neighborhood conditions

                            Place-oriented practices

Agency designates some officers as "community" or "neighborhood" officers

    Agency uses building code enforcement as a means of helping remove crime

Agency has landlord/manager training programs for order maintenance and
drug reduction

    Command or decision-making responsibility tied to neighborhoods or beats

           Patrol officers enforce civil and code violations in area

         Fixed assignment of patrol officers to specific beats or areas

          Agency uses other regulatory codes to combat drugs and crime

Agency has beat or patrol boundaries that coincide with neighborhood boundaries

                           Problem-solving practices

Agency prepares agreements specifying work to be done on problems by
citizens and police

Specific training provided to officers for problem identification and resolution

 Training for citizens in problem identification or resolution Patrol officers
               teach residents how to address community problems

        Interagency involvement in problem identification and resolution

Line supervisors elicit input from officers/deputies about solutions to
community problems

Multidisciplinary teams to deal with special problems such as child abuse
and neglect

                        Specialized problem-solving unit

    Patrol officers work with citizens to identify and resolve area problems

Citizens work with police to identify and resolve community or
neighborhood problems

      Organization has been redesigned to support problem solving efforts

  Patrol officers work with other city agencies to solve neighborhood problems

Source: Policing Strategies Survey, 1993 and 1997.

Note: Each individual item is coded dichotomously (yes/no) to indicate
whether an agency implemented the specific practice.

The Policing Strategies Survey provided us with an opportunity to assess
changes in reported policing practices using a pre-COPS grant and
within-COPS grant program framework. The 1993 administration of this
survey occurred several months prior to the distribution of the first COPS
grants, while the 1997 administration occurred after COPS grants had been
made to about 75 percent of the agencies in the sample. To implement the
pre-within examination of the effects of COPS grants on policing
practices, we first compared the levels of practices in 1993 and 1997
between the group of agencies that had received a COPS grant by 1997 and
the group that had not received a COPS grant by 1997.

Second, we estimated separate regressions of the effect of the receipt of
a COPS grant and of the cumulative per capita amount of COPS expenditures
on the levels of reported policing practices. To assess the extent to
which COPS grant expenditures were associated with changes in reported
policing practices, we estimated regressions of the changes in reported
policing practices that occurred within agencies as a function of the
cumulative per capita amount of COPS grant expenditures that they made
during the years from 1994 through 1997. We used two-factor fixed-effects
regression techniques, which allowed us to control for unobserved
characteristics of agencies and underlying trends in the adoption of
policing practices. We also controlled for economic conditions and
population changes in the localities in which the agencies were located.
In addition, we used weighted regressions to address nonresponse patterns
and the probability with which the original sampling units were drawn.

Our regression equations show that both the receipt of a COPS grant and
the amount of per capita COPS expenditures by agencies were associated
with increases in the levels of reported policing practices between 1993
and 1997. Agencies that received at least one COPS grant had significantly
larger changes in the overall number of practices than did agencies that
did not receive a COPS grant during this period. Specifically, of the
roughly 2.9 average increase in the number of practices reported by
agencies over this period, the receipt of a COPS grant accounted for 1.8
of the increase in the reported increase in practices. Further, when we
examined our results from separate regressions for the different
categories of practices, we found that receipt of a COPS grant was
associated with significant increases in reported levels of
problem-solving and place-oriented practices, but was not related to
changes in community collaboration or crime-analysis practices (table 24).

Table 24: Parameter Estimates from Regressions of Changes in Mean Number
of Policing Practices and Category of Practices between 1993 and 1997 on
whether or Not Agencies Received COPS grant between 1994 and 1997 and on
Per Capita COPS Expenditures between 1994 and 1997 (Standard Errors in
Parentheses)

                              Changes in policing practices from 1993 to 1997
Independent         
variable in            All 38 Problem-     Place-   Crime        Community
model               practices  solving  oriented   analysis  collaboration
Regression 1:            1.78       .76       .78        .01           .25
Received COPS                                                
grant                  (.732)    (.284)    (.245)     (.180)        (.273)
Regression 2:       
COPS expenditures        .226        .076       .086      .041        .023 
per                                                               
capita                  (.080)      (.034)     (.034)    (.017)     (.026) 

Source: GAO analysis of Policing Strategies Survey, Office of Justice
Programs financial, Bureau of Economic Analysis, Census, and Uniform Crime
Report data

Notes: All regressions include agency and year fixed effects and changes
in county level demographic variables (percentage of persons aged 15 to
24, percentage nonwhite, and percentage employed) between 1993 and 1997.
Observations are weighted to take into account response rates and the
probability at which the original sampling units were drawn. Bold-face
parameter estimates and standard errors indicate that a parameter estimate
is statistically significant at the 5 percent level.

Our regression models further show that changes in practices were also
associated with the cumulative amount of per capita spending on COPS
grants. All other things being equal, a $1 increase in per capita spending
was associated with an increase of 0.23 policing practices. As we found
for the effects of the receipt of a grant on changes in police practices,
these regressions also showed that the level of per capita spending on
COPS

Characteristics and Analysis of the National Evaluation of COPS Survey

grants was significantly associated with increases in problem-solving and
place-oriented practices. However, per capita spending on COPS grants was
also significantly associated with increases in crime analysis practices.

The National Evaluation of COPS Survey was conducted by the National
Opinion Research Center for the Urban Institute in its national evaluation
of the implementation of the COPS program. 5 The sampling frame for the
survey consisted of 20,894 law enforcement agencies believed to be in
existence between June 1993 and June 1997 who had either received a COPS
grant during 1995 or appeared to be potentially eligible for funding but
remained unfunded through 1995. The list of COPS grantees was obtained
from applicant records from the grants management database from the COPS
Office, and included those agencies that had been funded from the
following programs: FAST, AHEAD, Universal Hiring Program, and MORE. The
list of potentially eligible grantees was derived from the FBI's UCR and
National Crime Information Center data files. The sampling frame was
stratified by COPS grantee status (Not Funded, FAST or AHEAD, Universal
Hiring Program (UHP), MORE), and by population (jurisdictions with
populations of fewer than 50,000 persons and those with populations of
50,000 or more persons), and agencies in each stratum were sampled at a
different rate in order to select a representative sample of law
enforcement agencies. 6 A total of 2,098 agencies were selected to be in
the sample. 7

Telephone interviews with agency representatives were conducted in 1996
(wave 1) and 2000 (wave 4). 8 A total of 1,471 agencies responded to wave
1

5

See Roth, Jeffrey, et al., National Evaluation of the COPS Program-Title I
of the 1994 Crime Act.

6

Roth, et al. note that they lacked population data for 4,208 agencies in
the sampling frame. For sample selection purposes, they treated the
missing agencies as a separate stratum. However, because inspection
indicated that a large majority served jurisdictions of fewer than 50,000
persons, these agencies were analyzed in that population category.

7

Some agencies received more than one type of COPS grant and appeared in
more than one stratum. The analyses were weighted to take into account the
multiple probabilities of selection associated with each grant program.

8

The National Evaluation also conducted two other waves of telephone
interviews, in 1997 and 1998. However, for those surveys, only subsets of
the original sample were contacted.

of the survey in 1996, for a 77 percent response rate. 9 In 2000, all wave
1 respondents were recontacted, and interviews were completed with 1,270,
or 86 percent, of the target agencies.

We were able to link the data from 1,067 of the agencies that responded to
both of these waves of the survey to our larger database on crime,
officers, money, and economic conditions. 10 For comparability with the
analyses of the effects of funding on officers and crime, we excluded from
our analysis state police agencies, and other "special" police agencies,
as well as law enforcement agencies serving jurisdictions with populations
of fewer than 10,000 persons. This resulted in usable data on 724
agencies.

We used the National Evaluation of COPS Survey to compare levels of
practices in 1996 and 2000 between groups of agencies that received COPS
grants and those agencies that were not funded by COPS over this period,
and to assess changes in reported practices in relation to per capita COPS
expenditures. The analyses reported in this appendix are weighted to
adjust for nonresponse and the multiple counting of agencies that received
more than one COPS grant. The findings are generalizable to all law
enforcement agencies in the United States serving jurisdictions with
populations of 10,000 or more persons.

We used 19 items on policing practices from the National Evaluation of
COPS Survey, and we classified these items into the same 4 categories of
practices as we did with the Policing Strategies Survey data (table 25).
However, because of the shortage of items covering place-oriented
practices, for analysis purposes we combined these 3 items with the 7
problem-solving items into one index of problem solving and place oriented
practices.

9

The response rate is not equal to the number of completed interviews of
the number of agencies because of the possibility of agencies appearing in
multiple strata of the sample.

10

These agencies represented about 84 percent of the agencies that responded
to both waves 1 and 4 of the National Evaluation of COPS Survey

Page 102 GAO-06-104 Community Policing Grants

Table 25: Categories of Policing Practices and Specific Items within Each
Category in the National Evaluation of COPS Survey

                            Community collaboration

                  Regular community meetings to discuss crime

Surveys of citizens to determine general community needs and satisfaction
with agency

               Clean-up/fix-up projects with community residents

Considering neighborhood values in creating solutions or planning projects

 Varying styles of preventive patrol (e.g., bike patrol, walk-and-talk patrol)

     Joint projects with local businesses to reduce disorder or petty crime

Crime analysis

Analyzing crime patterns using computerized geographic information systems

Officers analyze community residents' comments to identify recurring
patterns of crime and disorder on their beats

Officers analyze and use crime data to identify recurring patterns of
crime and disorder on their beats

                            Place-oriented practices

Joint projects with community residents to reduce disorder such as
loitering or public drinking

 Beat or patrol boundaries that coincide with neighborhood/community boundaries

    Alcohol, housing, or other code enforcement to combat crime and disorder

                           Problem-solving practices

Designating certain recurring patterns as "problems" or "projects"
requiring nontraditional responses

Analyzing problems with business or property owners, school principals, or
property managers or occupants

    Analyzing problems with probation/parole officers or others who monitor
                                   offenders

         Using agency data to measure effects of responses to problems

       Using citizens' input to measure effects of responses to problems

  Document problems, projects, analyses, responses, failures, and successes in
                                    writing

                        Making sure problems stay solved

Source: National Evaluation of COPS Survey, 1996 and 2000.

Note: Each individual item is coded dichotomously (yes/no) to indicate
whether an agency implemented the specific practice.

Unlike the Policing Strategies Survey, which provided a pre-COPS and a
within-COPS measure of policing practices, both observations (in 1996 and
2000) on policing practices in the National Evaluation of COPS Survey
occurred while the COPS program was making grants. This complicates our
analysis, as in 1996 there were agencies that had already received and
spent COPS funds, and to the extent that COPS expenditures were associated
with the adoption of policing practices, the level of such practices that
they reported in 1996 would reflect their experiences with COPS grants.
Some of these agencies continued to spend COPS funds throughout the years
from 1996 through 2000. However, some of the agencies that spent COPS
funds in 1996 ceased to spend them during the intervening years before
2000. A third group of agencies consists of those that had not received
their first COPS grant in 1996 but had received a grant before 2000. This
third group is analogous to our group of agencies that received COPS
grants in the Policing Strategies survey, with the exception that while
members of this group received their first COPS grant after the first
administration of the National Evaluation survey in 1996, their practices
in 1996 could have been influenced by the COPS program indirectly. A final
group of agencies is those that did not receive a COPS grant before the
1996 administration of the survey or during the years from 1997 through
2000.

Because the effects of experience with COPS grants before and after 1996
could differ, we chose to make two types of comparisons. First, we
examined the mean changes in policing practices from 1996 to 2000 for each
of the following groups of agencies: (1) agencies that made expenditures
on COPS grants in 1994 through 1996, (2) agencies that made expenditures
on a COPS grant in 1997 through 2000, (3) agencies that made no
expenditures on a COPS grant after 1996, and (4) agencies that made no
expenditures on a COPS grant in 1994 through 2000. These mean comparisons
allowed us to see whether changes in practices were associated with
receipt of a grant in either the early period of the program (through
1996) or when the program was more fully implemented (1997 through 2000).

We then examined whether the level of COPS expenditures between the two
administrations of the survey were associated with changes in practices
between 1996 and 2000 by regressing the change in practices on the change
in cumulative per capita COPS expenditures between the period preceding
wave 1 of the survey (1994 through 1996) and the period following wave 1
of the survey (1997 through 2000). As with the Policing Strategies Survey,
we used two-factor fixed-effects regression techniques, which allowed us
to control for unobserved characteristics of agencies and underlying
trends in the adoption of policing practices. We also controlled for
economic conditions and population changes in the localities in which the
agencies were located. In addition, we used weighted regression to address
the complex design of the National Evaluation of COPS Survey. We estimated
separate regressions of the effect of the receipt of a COPS grant and of
the cumulative per capita amount of COPS expenditures on the levels of
reported policing practices.

There were no significant differences in the overall adoption of policing
practices associated with changes in per capita spending on COPS grants
(table 26).

Table 26: Parameter Estimates from Regressions of Changes in Mean Number
of Policing Practices and Category of Practices between 1996 and 2000 on
Whether or Not Agencies Received COPS grant Between 1997 and 2000 and on
Per Capita COPS Expenditures between 1994-1996 and 1997-2000 (Standard
Errors in Parentheses)

                              Changes in policing practices from 1996 to 2000
                                    Problem-
                                  solving and
Independent                    All 19    Place-     Crime        Community
variable in model            practices  oriented   analysis  collaboration
Changes in COPS              
expenditures per             
capita between 1994-1996 and      .056          .030      .011        .016 
1997-2000                        (.032)        (.021)    (.008)     (.008) 

Source: GAO analysis of National Evaluation of COPS Survey, Office of
Justice Programs financial, Bureau of Economic Analysis, Census, and
Uniform Crime Report data

Notes: All regressions include agency and year fixed effects and changes
in county level demographic variables (percentage of persons aged 15 to
24, percentage nonwhite, and percentage employed) between 1996 and 2000.
Observations are weighted to take into account sample design effects.
Bold-face parameter estimates and standard errors indicate that a
parameter estimate is statistically significant at the 5 percent level.

To determine whether the certain types of policing practices may be

Methods to Review

effective in reducing crime, we analyzed systematic reviews of research
Policing Practices studies on the effectiveness of policing practices.

                            How We Selected Studies

We identified six studies that provided summaries of research on the
effectiveness of policing practices on reducing crime. We chose to review
studies that reviewed research, rather than reviewing all of the original
studies themselves, because of the volume of studies that have been
conducted on the effectiveness of policing practices. We reviewed the
following studies:

o  Braga, Anthony. "Effects of Hot Spots Policing on Crime," Annals,
AAPSS, vol. 578 (November 2001), pp. 104-125.

                            How We Reviewed Studies

o  Eck, John. "Preventing Crime at Places" in Sherman, L., et al. (eds.)

Preventing Crime: What Works, What Doesn't, What's Promising: A

Report to the United States Congress. Washington, D.C.: National

Institute of Justice, 1998.

        * Eck, John, and Edward Maguire. "Have Changes in Policing Reduced
          Violent Crime? An Assessment of the Evidence." in Blumstein, A.,
          and
        * J. Wallman, eds., The Crime Drop in America. United Kingdom:
          Cambridge University Press, 2000.
     o Sherman, Lawrence. "Policing for Crime Prevention," in Sherman, L., et
       al. (eds.) Preventing Crime: What Works, What Doesn't, What's
       Promising: A Report to the United States Congress. Washington, D.C.:
       National Institute of Justice, 1998.
     o Skogan, Wesley, and Kathleen Frydl. "The Effectiveness of Police
       Activities in Reducing Crime, Disorder, and Fear," in Skogan, W., and
       K. Frydl, (eds.) Fairness and Effectiveness in Policing: The Evidence,
       Washington, D.C.: National Academies Press, pp. 217-251, 2004.
     o Weisburd, David, and John Eck. "What Can Police Do to Reduce Crime,
       Disorder, and Fear?" Annals, AAPSS, Vol. 593 (November 2004), pp. 42

65.

A limitation of basing our work on reviews is that we did not assess the
original studies, but rather we relied on the descriptions and assessments
as provided by the authors of the reviews. Sometimes the reviews did not
cite specific information about the strength of the methodology of the
underlying studies that were included in reviews.

We developed a data collection instrument to capture systematically
information about the methodologies of the reviews, the types of policing
practices reviewed, findings about each type of practice, and the
reviewers' conclusions about the effectiveness of a particular practice or
group of practices in reducing crime. Each research review was read and
coded by a social science analyst who had training and experience in
reviewing research methodologies. This analyst recorded, for each practice
discussed in the research review, (1) the types of crimes against which
the practices were used (e.g., all crimes, violent crimes, property
crimes, disorder); (2) whether the practice was generally effective in
reducing crime, had no effect in reducing crime, or the impact was
ambiguous; (3) whether there was displacement of crimes away from the

The Research Literature Shows That Some Policing Practices May be
Effective in Reducing Crime

areas where the practices were used; and (4) whether there were negative
effects of the practices (e.g., complaints against the police or the
diversion of resources from other policing activities). A second,
similarly trained analyst then read the reviews and verified the accuracy
of the information recorded by the first analyst. We then summarized the
findings about each practice from the data collection instruments prepared
for each of the six reviews. Some practices were discussed in only one
review, while others were discussed in more than one review.

Our analysis of six systematic reviews of evaluations of the effectiveness
of various policing practices in preventing crime indicates that the
current evidence ranges from moderate to strong that problem-oriented
policing practices and place-oriented practices are either effective or
promising as strategies for addressing crime problems. For example,
problem-oriented approaches that focus on criminogenic substances such as
guns and drugs appear to be effective in reducing both violent and
property crimes. And hot spots approaches-place-oriented approaches that
temporarily apply police resources to discrete locations where crime is
concentrated at much higher rates than occurs jurisdictionwide-have also
been found to be effective in reducing crime. However, the magnitudes of
the effects of these interventions are difficult to estimate, especially
on citywide crime rates, as the interventions that were reviewed as
effective generally were concentrated in comparatively small places.
Further, the enduring nature of these interventions is not fully
understood. It is not known, for example, how long the effects of a
problem- or place-oriented intervention persist. In addition, some of the
reviews point out that research designs undertaken to date make it
difficult to disentangle the effects of problem-oriented policing from hot
spots policing. There is suggestive, but limited, evidence that the
combination of these practices may be more effective in preventing or
reducing crime than any one strategy alone.

In contrast to the findings on problem-oriented and place-oriented
policing practices, there is little evidence in the literature for the
effectiveness of community collaboration practices-such as increasing foot
patrol, establishing community partnerships, and encouraging citizen
involvement-in reducing or preventing crime.

Appendix VIII: Comments from the Department of Justice

Appendix IX: GAO Contacts and Staff Acknowledgments

Laurie E. Ekstrand (202) 512-8777

GAO Contacts

Nancy R. Kingsbury (202) 512-2700

In addition to those named above the following individuals made key

contributions to this report: William J. Sabol, Tom Jessor, David R.
Lilley, Benjamin A. Bolitzer, George H. Quinn, Jr., and Grant M. Mallie.
Others contributing included David P. Alexander, Harold J. Brumm Jr.,
Scott Farrow, Kathryn E. Godfrey, Adam T. Hatton, Ronald La Due Lake,
Terence C. Lam, and Robert Parker.

Bibliography

Attorney General of the United States. Report to Congress: Office of
Community Oriented Policing Services, Washington, D.C.: U.S. Department of
Justice, September 2000.

Blumstein, Alfred, and Joel Wallman (eds.). The Crime Drop in America,
Cambridge, England: Cambridge University Press, 2000.

Braga, Anthony. "Effects of Hot Spots Policing on Crime," Annals, AAPSS,
vol. 578 (November 2001), pp. 104-125.

Cook, Philip. "The Clearance Rate as a Measure of Criminal Justice System
Effectiveness," Journal of Public Economics, vol. 11, 1979, pp. 135-142.

Davis, Gareth, et al. "The Facts about COPS: A Performance Overview of the
Community Oriented Policing Services Program," Washington, D.C.: The
Heritage Foundation, September 25, 2000.

Di Tella, Rafael and Ernesto Schargrodsky. "Do Police Reduce Crime?
Estimates Using the Allocation of Police Forces after a Terrorist Attack."
American Economic Review. March 2004, 94(1). pp. 115-133.

Dunworth, Terence, Peter Haynes, and Aaron J. Saiger. National Assessment
of the Byrne Formula Grant Program, Washington, D.C.: National Institute
of Justice Research Report, June 1997.

Eck, John. "Preventing Crime at Places" in Sherman, L., et al. (eds.)

Preventing Crime: What Works, What Doesn't, What's Promising: A Report to
the United States Congress. Washington, D.C.: National Institute of
Justice, 1998.

Eck, John and Edward Maguire. "Have Changes in Policing Reduced Violent
Crime? An Assessment of the Evidence." in Blumstein, A., and J. Wallman,
eds., The Crime Drop in America. United Kingdom: Cambridge University
Press, 2000.

Ehrlich, Isaac. "Participation in Illegitimate Activities: A Theoretical
and Empirical Investigation." Journal of Political Economy, May June,
81(3), pp. 521-65 (1973).

Evans, William N., and Emily Owens. "Flypaper COPS," College Park,
Maryland.: University of Maryland. Available online
www.bsos.umd.edu/econ/evans/wpapers/Flypaper%20COPS.pdf., 2005.

Executive Office of the President. Performance and Management Assessments:
Budget of the United States, Fiscal Year 2004, Washington, D.C.: The White
House, 2003.

Federal Bureau of Investigation, Crime in the United States, 2002, Uniform
Crime Reports, Washington, D.C.: U.S. Department of Justice. Printed
annually.

GAO, Technical Assessment of Zhao and Thurman's 2001 Evaluation of the
Effects of COPS Grants on Crime, GAO-03-867R (Washington, D.C.: June 13,
2003).

Johnson, Calvin C., and Jeffrey A. Roth, The COPS Program and the Spread
of Community Policing Practices, 1995-2000. Washington, D.C.: The Urban
Institute, June 2003.

Klick, Jonathan, and Alexander Tabarrok. "Using Terror Alert Levels to
Estimate the Effect of Police on Crime." Journal of Law and Economics,
April, Vol. XLVIII (2005).

Koper, Christopher S., et al. Putting 100,000 Officers on the Street: A
Survey-Based Assessment of the Federal COPS Program, Washington, D.C.: The
Urban Institute, October 2002.

Levitt, Steven D. "The Effect of Prison Population Size on Crime Rates:
Evidence from Prison Overcrowding Litigation" Quarterly Journal of
Economics, 111 No. 2. (May, 1996).

Levitt, Steven D. "Using Electoral Cycles in Police Hiring to Estimate the
Effect of Police on Crime." American Economic Review 87 (1997): 270290.

Levitt, Steven D., "The Relationship between Crime Reporting and Police:
Implications for the Use of Uniform Crime Reports," Journal of
Quantitative Criminology, Vol. 14, No. 1998: pp. 61-81.

Levitt, Steven D. "Using Electoral Cycles in Police Hiring to Estimate the
Effects of Police on Crime: Reply" American Economic Review, September
2002, 92(4), pp. 1,244-50.

Lynch, James P., "Exploring the Sources of Non-response in the Uniform
Crime Reports: Things to Do Before Multiple Imputation." Paper presented
at the Annual Meetings of the American Society of Criminology Research
Conference, November 19, 2003.

Marvell, Thomas, and Carlisle Moody. "Specification Problems, Police
Levels, and Crime Rates." Criminology 1996, Vol., No. 4, pp. 609-46.

McCrary, Justin. "Using Electoral Cycles in Police Hiring to Estimate the
Effect of Police on Crime: Comment." American Economic Review. September
2002, 92(4), pp. 1,236-43.

National Center for Health Statistics. Estimates of the July 1, 2000-July
1, 2003, United States resident population from the Vintage 2003
postcensal series by year, county, age, sex, race, and Hispanic origin,
prepared under a collaborative arrangement with the U.S. Census Bureau.
Available on the Internet at
http://www.cdc.gov/nchs/about/major/dvs/popbridge/popbridge.htm. September
14, 2004.

Rosenthal, Arlen M., and Lorie Fridell. National Survey of Community
Policing Strategies Update, 1997, and Modified 1992-1993 Data [Computer
file]. Inter-university Consortium for Political and Social Research
(ICPSR) version. Calverton, Maryland: ORC Macro International, Inc.
[producer], 2002. Ann Arbor, Michigan: ICPSR [distributor], 2002.

Roth, Jeffrey, et al. National Evaluation of the COPS Program-Title I of
the 1994 Crime Act, Washington, D.C.: National Institute of Justice,
August 2000.

Sherman, Lawrence. "Policing for Crime Prevention," in Sherman, L., et al.
(eds.) Preventing Crime: What Works, What Doesn't, What's Promising: A
Report to the United States Congress. Washington, D.C.: National Institute
of Justice, 1998.

Skogan, Wesley, and Kathleen Frydl. "The Effectiveness of Police
Activities in Reducing Crime, Disorder, and Fear," in Skogan, W., and K.
Frydl, (eds.) Fairness and Effectiveness in Policing: The Evidence,
Washington, D.C.: National Academies Press, pp. 217-251, 2004.

Swimmer, Eugene. "The Relationship of Police and Crime: Some
Methodological and Empirical Results," Criminology, vol. 12, 1974: pp.
293-314.

Weisburd, David, and John Eck. "What Can Police Do to Reduce Crime,
Disorder, and Fear?" Annals, AAPSS, vol. 593 (May 2004), pp. 42-65.

Wycoff, Mary Ann, Community Policing Strategies: A Comprehensive Analysis,
Washington, D.C.: The Police Foundation, November 1994.

Zhao, J., and Q. Thurman. A National Evaluation of the Effect of COPS
Gants on Crime from 1994 to 1999. Report submitted to the Office of
Community Oriented Policing Services, Washington, D.C.: U.S. Department of
Justice, December 2001.

Zhao, J., and Q. Thurman. Funding Community Policing to Reduce Crime: Have
COPS Grants Made a Difference from 1994 to 2000? Report submitted to the
Office of Community Oriented Policing Services, Washington, D.C.: U.S.
Department of Justice, July 2004.

(440258)

GAO's Mission

The Government Accountability Office, the audit, evaluation and
investigative arm of Congress, exists to support Congress in meeting its
constitutional responsibilities and to help improve the performance and
accountability of the federal government for the American people. GAO
examines the use of public funds; evaluates federal programs and policies;
and provides analyses, recommendations, and other assistance to help
Congress make informed oversight, policy, and funding decisions. GAO's
commitment to good government is reflected in its core values of
accountability, integrity, and reliability.

The fastest and easiest way to obtain copies of GAO documents at no cost
is through GAO's Web site ( www.gao.gov) . Each weekday, GAO posts GAO
Reports and newly released reports, testimony, and correspondence on its
Web site. To

have GAO e-mail you a list of newly posted products every afternoon, go to
www.gao.gov and select "Subscribe to Updates."

                             Order by Mail or Phone

The first copy of each printed report is free. Additional copies are $2
each. A check or money order should be made out to the Superintendent of
Documents. GAO also accepts VISA and Mastercard. Orders for 100 or more
copies mailed to a single address are discounted 25 percent. Orders should
be sent to:

U.S. Government Accountability Office 441 G Street NW, Room LM Washington,
D.C. 20548

To order by Phone: Voice: (202) 512-6000 TDD: (202) 512-2537 Fax: (202)
512-6061

To Report Fraud, Contact:

Web site: www.gao.gov/fraudnet/fraudnet.htm

E-mail: [email protected]

Federal Programs Automated answering system: (800) 424-5454 or (202)
512-7470

Gloria Jarmon, Managing Director, [email protected] (202) 512-4400 U.S.
Government Accountability Office, 441 G Street NW, Room 7125 Relations
Washington, D.C. 20548

Paul Anderson, Managing Director, [email protected] (202) 512-4800

Public Affairs

U.S. Government Accountability Office, 441 G Street NW, Room 7149
Washington, D.C. 20548

  PRINTED ON

RECYCLED PAPER
*** End of document. ***