.4 .
. .5:
.. .
...;
......

.

.,. _
I ”%

.... .1.
5...“; .

I.- 353. .l
.03.. a

In 0 . y
g. ... .... ...”...
n). .I .
......ﬁﬁ. ..
1‘. - 1 ex
, , .3103 «rt: . .
km...“ 2... .f...%m&gl .
....wdwe . «gag» m¢.%...ww%§
M . _.
.....n. .

b
[1»

$
it! In 1A....
. 3....

h b.

D it I

. t .r. .
.. EH;
ﬂamb...»

“PM

1.:
5 3...»...

iii. 3“. s
‘2... an: ...!

...mu 1.? ...;
a ......53. .

 

1.. ....
....lt? 2:.

 

 

 

1"... I’d”.
. ..ﬂuNhEPgi. .
. . 11.. ....u ..., . . , .
...... .... .. :22}:
:22}... .539 army"...

 

mcezs‘

3 LIBRARY
we Michigan State
University

This is to certify that the
dissertation entitled

Assessing Jurisdiction-Level Crime Trends During the 19905:

An Analysis of the Impact of Policing Changes

presented by

David R. Lilley

has been accepted towards fulﬁllment
of the requirements for the

degree In Criminal Justice

 

 

Qﬁf%

Major Professor’ 3 Signature

April 5, 2006

 

Date

MSU is an Afﬁrmative Action/Equal Opportunity Institution

. ----—----O-

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2105 pIICIRC/DaleDue.indd-p.1

 

ASSESSING JURISDICTION-LEVEL CRIME TRENDS DURING THE 19908:
AN ANALYSIS OF THE IMPACT OF POLICING CHANGES

by

David R. Lilley

A DISSERTATION
Submitted to
Michigan State University
In partial fulﬁllment of the requirements for the degree of
DOCTOR OF PHILOSOPHY

School of Criminal Justice

2006

ABSTRACT
ASSESSING JURISDICTION-LEVEL CRIME TRENDS DURING THE 19903:
AN ANALYSIS OF THE IMPACT OF POLICING CHANGES
By David R. Lilley
Although a number of researchers have suggested that changes in policing may

have contributed to the “crime drop” of the 19905, little is known regarding the relative
impact that changes in policing tactics, force strength and arrests may have had in
jurisdictions across the nation (Blumstein, Steinberg, Bell & Berger, 2002; Sherman,
Gottfredson, MacKenzie, Eek et al., 1998). Criminal justice literature asserts that
implementation of crime analysis, regulatory code enforcement, community policing and
problem solving became more widespread during this decade. Additionally, data suggest
that the number of ofﬁcers per capita and arrests for minor drug and disorder offenses
also changed substantially. Using panel data from two national samples of police
jurisdictions during the 19905, this study assessed the relative effects of these policing
changes on each of seven primary UCR offenses within jurisdictions across the nation.
Results indicate that implementation of crime analysis, civil and regulatory code
enforcement, as well as increased problem solving and citizen involvement in police
activities were associated with substantial reductions in rates of robbery, burglary,
larceny and vehicle theﬁ and smaller reductions in murder, rape and aggravated assault.
Additional crime declines were also associated with increases in ofﬁcers and minor drug
and disorder arrests. Patterns among the types of policing changes that were most closely

associated with reductions in each type of offense were also identiﬁed.

 

Copyright by
DAVID ROBERT LILLEY
2006

To my wife, Tara.

There you are in the early light of day
Every time I turn around
When I’m lost and when I’m found
Like an angel standing guard
There you are

Every time I take a breath
And when I forget to breathe
You’re watching over me
There you are

When I look for the light
In the darkness of the night
And when I move afar
There you are

ACKNOWLEDGEMENTS

I would like to express my appreciation to Professor Bruce Benson, who stepped
forward to offer assistance after an unexpected committee loss and guided me through the
process as I worked in Washington DC. I would also like to thank Professor David
Carter for his efforts to cut through the bureaucracy with regard to data access and for
supporting my efforts to undertake this challenging analysis. I am grateful for the
counsel of Professor Kevin Ford, who was with me throughout the entirety of my
graduate career. Finally, I would like to thank my family for their continuous
encouragement and to offer a special thank you to my sister, Krista, who facilitated my

return to Michigan to complete the remaining portion of this study.

TABLE OF CONTENTS

 

 

 

 

LIST OF TABLES ...... - - -- -- - - -- - ........ xi
LIST OF FIGURES - - - ...... - - - -- _ -- _- xiv
CHAPTER 1
INTRODUCTION - - - - - ................ - - - -- -- - 1
CHANGES IN POLICING AND THE CRIME DECLINE .................................... 2
Technology and Community Policing ......................................................... 2
Systematic Problem Solving Efforts ............................................................ 3
Policing Research - Paradigm Shift ............................................................. 3
Incarceration and Police Force Size ............................................................. 4
CHARACTERISTICS OF THE CRIME DROP ..................................................... 5
STATEMENT OF THE RESEARCH PROBLEM ................................................. 7
THE NEED FOR THE CURRENT STUDY ........................................................... 8
PURPOSE AND OBJECTIVES OF THIS STUDY ............................................... 9
CHAPTER 2
REVIEW OF THE LITERATURE - POLICING CHANGES - - 11
EVENTS LEADING TO THE PARADIGM SHIFT OF THE 19903 .................. ll
POLICING CHANGES DURING THE 1990s ..................................................... 13
Efforts to Improve Community Appearance and Orderliness ................... l3
Improving Citizen Collaboration as a Tactic ............................................. 17
Combating Fear as a Tactic ........................................................................ 18
The Importance of Social Cohesion, Collective
Efﬁcacy and Networks ............................................................................... 20
Directed Patrol and Renewed Crime Focus ............................................... 23
Increasing Police Force Concentration — Hot Spots .................................. 25
Aggressive Policing and Crackdowns ....................................................... 26
Area Denial of Criminal Activities ............................................................ 28

V1

Problem Solving ......................................................................................... 29

 

Controlling Crime Facilitators ................................................................... 31
Focusing on Repeat Victims ...................................................................... 31
Focusing on Repeat Offenders ................................................................... 32
Increasing Police Force Size ...................................................................... 34
Mechanisms through which Policing Effects Occur .................................. 36
Increased Incarceration and “Zero Tolerance” Strategies ......................... 41
Advances in Police Technology ................................................................. 42
Rapid Expansion of Police Technology ..................................................... 43
Crime Analysis and Compstat ................................................................... 45
Increased Federal Funding for Local Law Enforcement ........................... 49
CHAPTER 3
CHANGES IN CRIME TRENDS DURING THE 1990s - ....................... 53
METHODOLOGIES OF ANALYZING CRIME TRENDS ................................ 53
“CRIME TREND” DEFINED ............................................................................. 57
PERCENTAGES OF JURISDICTIONS BELOW 1990 CRIME
RATES (BY SIZE CATEGORY) ......................................................................... 58
MEAN PERCENTAGES OF CRIME CHANGE IN
JURISDICTIONS (BY SIZE CATEGORY) ......................................................... 67
OVERALL WEIGHTED MEAN PERCENTAGES NATIONWIDE VERSUS
NATIONAL SUMMED RATES ........................................................................... 72
NATIONAL WEIGHTED MEAN PERCENTAGE CHANGES
BY CRIME TYPE ................................................................................................. 75
DESCRIPTIVE CRIME TRENDS - SUMMARY OF OBSERVATIONS .......... 80
DESCRIPTIVE ANALYSES OF CONTROL VARIABLES - SYSTEMIC
INFLUENCES ON CRIME ................................................................................... 81
SYSTEMIC CHANGES AND CRIME TRENDS - A REVIEW OF
LITERATURE ....................................................................................................... 85
Youth Population and Crime Trends ......................................................... 85
Unemployment and Crime Trends ............................................................. 88
Interrelationships between Race, Youth, Wages and Unemployment ....... 89
Systemic Changes and Crime Trends - A Look at the Data ..........................
Youth Demographics Data (1990-2001) .................................................... 91
Unemployment, Income and Racial Demographics Data (1990-2001) ..... 91

VII

Officers Per Capita - Data (1990-2001) ..................................................... 92

 

CHAPTER 4

METHODOLOGY _- .......... - - - . ........ ..... 93
STUDY OBJECTIVES AND HYPOTHESES ..................................................... 93
METHODOLOGICAL OVERVIEW .................................................................... 94
POPULATION OF INFERENCE AND DATABASE OVERVIEW ................... 96
PRIMARY DATA SOURCES .............................................................................. 96
Changes in Crime ....................................................................................... 96
Changes in Arrests and Ofﬁcer Rates ....................................................... 97
Changes in Policing Practices .................................................................... 98
Control Variables ....................................................................................... 99
Linking Strategy ......................................................................................... 99
Increases in Ofﬁcers Per Capita - Hypothesis 1 ...................................... 101
Increases in Enforcement of Minor Offenses - Hypothesis 2 .................. 101
Policing Practices - Hypotheses 3 and 4 .................................................. 103
Creation and Analyses of Policing Practice Indexes ............................... 105
Creation and Analyses of Individual Policing Practice Variables ........... 108
PANEL DATA REGRESSION PROCEDURES ................................................ 110
Analysis of Officers and Arrests .............................................................. 110

Initial Models of Officers and Minor Arrests .......................................... 112

Final Model (Ofﬁcers and Minor Arrests) ............................................... 114
Regression of Policing Practices Indexes vs. Individual Items ............... 114
Regression Models for Policing Practices ............................................... l 15

Data Reduction and Final Model (Policing Indexes) ............................... 118
Analyses of Individual Policing Practice Items ....................................... 119

Data Reduction and Final Model (Individual Policing Items) ................. 120
METHODOLOGICAL CHALLENGES ............................................................. 120
Over-Reporting of Innovative Policing Practices .................................... 120

Arrest Data Standardization Issues .......................................................... 121
Omitted Variables and Threats to Validity .............................................. 122
ROBUSTNESS CHECKS ................................................................................... 123

VARIABLE DESCRIPTIVES PERTAINING TO THE OFFICER AND
ARREST DATABASE ........................................................................................ 123

viii

VARIABLE DESCRIPTIVES PERTAINING TO POLICING THE

CHAPTER 5

 

CHAPTER 6

 

PRACTICES DATABASE .................................................................................. 125
RESULTS - - - -- -_ - 131

SECTION I - THE INFLUENCE OF OFFICERS, DRUG AND

DISORDER ARRESTS ....................................................................................... 13]
Preliminary Granger Causality Assessments ........................................... 131
Full Models Using Granger Regression ................................................... 140
Final Model Effect Sizes .......................................................................... 147
Summary of Results - Ofﬁcers, Drug and Disorder
Arrests - Hypotheses l and 2 ................................................................... 150

RESULTS SECTION II - POLICING PRACTICES .......................................... 151
Initial Models - Policing Practice Indexes ............................................... 152
Full Model - Policing Practice Indexes .................................................... 154
Summary of Results - Policing Indexes - Hypothesis 3 .......................... 158
Initial Models - Individual Policing Practices ......................................... 159
Final Models - Individual Policing Practices ........................................... 162
Summary of Results - Individual Policing Tactics - Hypothesis 4 .......... 172
Robustness Checks ................................................................................... 174

DISCUSSION - ..................... - ..... - - - - ----178

OFFICERS AND CRIME - CONNECTIONS TO PRIOR EMPIRICAL

FINDINGS ........................................................................................................... 178

DRUG ARRESTS - CONNECTIONS TO PRIOR EMPIRICAL

FINDINGS ........................................................................................................... 181

DISORDER ARRESTS - CONNECTIONS TO PRIOR EMPIRICAL

FINDINGS ........................................................................................................... I83

THEORY AND MECHANISMS OF THE EFFECTS OF OFFICERS AND

MINOR ARRESTS ON CRIME ......................................................................... 186
Enforcement Based Perspectives (Deterrence and Incapacitation) ......... 186
Community Crime Theory and Ecological Perspectives ......................... 189

CHANGES IN OFFICERS AND MINOR ARRESTS IN THE TWO WAVE
PANEL ................................................................................................................. 191

ix

POLICING PRACTICE INDEXES - CONNECTIONS TO PRIOR

 

EMPRICAL FINDINGS ...................................................................................... 192
Broadening Community Policing Policies ............................................... 193
Broadening Community Collaboration Practices .................................... I93
Broadening Citizen Involvement Practices .............................................. 194
Broadening Problem Solving Practices ................................................... 195
Broadening Crime Analysis Techniques ................................................. 197
Broadening Place Oriented Practices ....................................................... 197
Broadening Civil Code Enforcement Practices ....................................... I98
Broadening Efﬁciency Oriented Practices ............................................... 200

INDIVIDUAL POLICING PRACTICES - CONNECTIONS TO CRIME

THEORY ............................................................................................................. 202
Larceny and Individual Policing Practices .............................................. 203
Burglary and Individual Policing Practices ............................................. 205
Vehicle Theﬂ and Individual Policing Practices ..................................... 206
Robbery and Individual Policing Practices .............................................. 208
Rape and Individual Policing Practices ................................................... 208
Aggravated Assault and Individual Policing Practices ............................ 209
Murder and Individual Policing Practices ................................................ 210

DISCUSSION SUMMARY AND CONCLUSIONS .......................................... 211
Policing Practices ..................................................................................... 21 1
Changes in Ofﬁcer and Minor Arrest Rates ............................................ 212
The Relationship Between Policing Changes and Crime
During the I990S ...................................................................................... 213

LIMITATIONS .................................................................................................... 2 l 6

POLICY RECOMMENDATIONS ..................................................................... 218
Practices that are Low Cost and Easy to Implement ................................ 219
Practices that were Associated with Reductions of
Multiple (3 or more) Offense Categories ................................................. 221
Practices that were Strongly Associated with the Reduction
of a Speciﬁc Offense ................................................................................ 223

CONCLUDING REMARKS ............................................................................... 224

APPENDICES 226
REFERENCES 245

 

LIST OF TABLES

Table 1. Violent Crime Nationwide During the 19903 - Comparison of

Three Methodologies ......................................................................................................... 56
Table 1a. Comparison of UCR Population and Arrest Sample ....................................... 124
Table 2. Changes in Ofﬁcer and Arrests Rates from FBI data ....................................... 125
Table 3. Comparison of UCR Population and Police Survey Subsample ...................... 126
Table 4. Changes in Policing Practices from the National Survey of

Community Policing Strategies ....................................................................................... 127
Table 5. Correlation of Officers and Minor Arrests and Policing Indexes ..................... 128
Table 6. Correlations Among the Policing Practice Indexes (1993-1997) ..................... 130
Table 7. Granger Analyses of the Impact of Crime on the Number of Police ................ 132
Table 8. Granger Analyses of the Impact of the Number of Police on Crime ................ 133

Table 9. Granger Analyses of the Impact of Crime on Misdemeanor
Drug Arrests .................................................................................................................... 134

Table 10. Granger Analyses of the Impact of Misdemeanor Drug Arrests
on Crime ........................................................................................................................... 134

Table 1 l. Granger Analyses of the Impact of Crime on Misdemeanor
Disorder Arrests .............................................................................................................. 135

Table 12. Granger Analyses of the Impact of Misdemeanor Disorder
Arrests on Crime .............................................................................................................. 135

Table 13. One and Two-year Beta Coefﬁcients for the Effects of
Ofﬁcer Rates on Crime .................................................................................................... 136

Table 14. One and Two-Year Beta Coefﬁcients for the Effects of

Drug Arrests on Crime ..................................................................................................... 136

Table 15. One and Two-Year Beta Coefﬁcients for the Effects of Disorder

Arrests on Crime .............................................................................................................. 137
xi

Table 16. Adjusted Effect Sizes for Initial Granger Regressions of

Officers and Crime ........................................................................................................... 138
Table 17. Adjusted Effect Sizes for Initial Granger Regressions of Drug

Arrests and Crime ............................................................................................................ 139
Table 18. Adjusted Effect Sizes for Initial Granger Regressions of Disorder

Arrests and Crime ............................................................................................................ 139
Table 19. Final Granger Regression for Murder (Full Model) ....................................... 140
Table 20. F inal Granger Regression for Rape (Full Model) ........................................... 141
Table 21. Final Granger Regression for Robbery (Full Model) ..................................... 143
Table 22. Final Granger Regression for Aggravated Assault (Full Model) ................... 144
Table 23. Final Granger Regression for Burglary (Full Model) ..................................... 145
Table 24. Final Granger Regression for Larceny (Full Model) ...................................... 146
Table 25. Final Granger Regression for Vehicle Theft (Full Model) ............................. 147
Table 26. Effect Size Calculations for Final Models of Ofﬁcers and Crime .................. 148
Table 27. Effect Size Calculations for F inal Models of Drug Arrests and Crime .......... 149
Table 28. Effect Size Calculations for Final Models of Disorder Arrests and Crime 150
Table 29. Regression Coefﬁcients and Standard Errors for Crime and Policing

Tactics Indexes ................................................................................................................. 153
Table 30. Regression Results of Vehicle Theft and Policing Practice Indexes .............. 155
Table 31. Regression Results of Larceny and Policing Practice Indexes ....................... 156
Table 32. Regression Results of Burglary and Policing Practice Indexes ...................... 157
Table 33. Regression Results of Robbery and Policing Practice Indexes ...................... 158
Table 34. Signiﬁcant Individual Policing Practices from Initial Crime

Regression Models ........................................................................................................... 161
Table 35. Signiﬁcant Policing Practices After Data Reduction Procedure - Final
Regression Models ........................................................................................................... 163

xii

Table 36. Regression Results of Vehicle Theft and Individual Policing
Practices - Final Model .................................................................................................... 165

Table 37. Regression Results of Larceny and Individual Policing
Practices - Final Model .................................................................................................... 166

Table 38. Regression Results of Burglary and Individual Policing
Practices - Final Model. ................................................................................................... 167

Table 39. Regression Results of Agg. Assault and Individual
Policing Practices - Final Model ...................................................................................... 168

Table 40. Regression Results of Robbery and Individual Policing
Practices - Final Model .................................................................................................... 169

Table 41. Regression Results of Rape and Individual Policing
Practices - Final Model .................................................................................................... 171

Table 42. Regression Results of Murder and Individual Policing
Practices - Final Model .................................................................................................... 172

Table 43. Robustness Check - Adding Ofﬁcers, Arrests to Policing
Index Regression of Robbery ........................................................................................... 175

Table 44. Robustness Check - Adding Ofﬁcers, Arrests to Individual
Policing Practice Regression of Larceny ......................................................................... 176

Table 45. Robustness Check - Removing Agencies With Over 250,000 and Less
Than 25,000 Population - Regression of Robbery and Problem Solving Index .............. 177

Table 46. Robustness Check - Removing Agencies With Over 250,000 and Less
Than 25,000 Population - Regression of Larceny and Citizen Involvement Index ......... 177

Table 47. Estimated Percentages of National Crime Declines Associated
with Policing Changes ..................................................................................................... 215

Appendix B. Table 1. National Survey of Community Policing Strategies.
Complete List of 80 Policing Tactics from Questionnaires ............................................. 230

Appendix C. Tables 1 - 7. Individual Policing Survey Items - Signiﬁcantly
Related to Crime Reduction ............................................................................................. 233

Appendix D. Table 1. Changes in Each of the 80 Policing Practices ............................ 237

xiii

LIST OF FIGURES

Figure 1. National Violent Crime Index (UCR 1980-2000) ............................................... 5
Figure 2. Trends in Agencies Below their 1990 Levels for Violent Crime

by Jurisdiction Size .............................................................................................................. 6
Figure 3. Partial Model of Social Disorganization and Fear ............................................ 19
Figure 4. Violent Index Trends (% agencies below 1990 levels) ..................................... 59
Figure 5. Property Index Trends (% agencies below 1990 levels) ................................... 59
Figure 6. Murder Trends (% agencies below 1990 levels) ............................................... 61
Figure 7. Manslaughter Trends (% agencies below 1990 levels) ..................................... 62
Figure 8. Aggravated Assault Trends (% agencies below 1990 levels) ........................... 63
Figure 9. Rape Trends (% agencies below 1990 levels) ................................................... 64
Figure 10. Robbery Trends (% agencies below 1990 levels) ........................................... 64
Figure 11. Burglary Trends (% agencies below 1990 levels) ........................................... 65
Figure 12. Larceny Trends (% agencies below 1990 levels) ............................................ 66
Figure 13. Vehicle Theﬁ Trends (% agencies below 1990 levels) ................................... 66
Figure 14. Murder (Mean % change by size group) ......................................................... 67
Figure 15. Manslaughter (Mean % change by size group) ................................................ 68
Figure 16. Rape (Mean % change by size group) ............................................................. 68
Figure 17. Aggravated Assault (Mean % change by size group) ..................................... 69
Figure 18. Robbery (Mean % change by size group) ....................................................... 70
Figure 19. Burglary (Mean % change by size group) ....................................................... 70
Figure 20. Larceny (Mean % change by size group) ........................................................ 71

xiv

Figure 21. Vehicle Theﬁ (Mean % change by size group) ............................................... 71
Figure 22. Violent Change Nationwide (Weighted mean % change) ............................... 73
Figure 24. National Violent Rate (Summed offenses/Population) ................................... 73
Figure 23. Property Change Nationwide (Weighted mean % change) ............................. 74
Figure 25. National Property Rate (Surnmed offenses/Population) .................................. 74
Figure 26. Murder Change Nationwide (Weighted mean % change) ................................ 76
Figure 27. Manslaughter (Weighted mean % change) ...................................................... 76
Figure 28. Rape Change Nationwide (Weighted mean % change) .................................. 77
Figure 29. Aggravated Assault Change (Weighted mean % change) ............................... 77
Figure 30. Robbery Change (Weighted mean % change) ................................................. 78
Figure 31. Burglary (Weighted mean % change) ............................................................ 78
Figure 32. Larceny (Weighted mean % change) .............................................................. 79
Figure 33. Auto Theﬁ (Weighted mean % change) ......................................................... 79
Figure 33. Population Aged 15-24 (Mean % Change) ..................................................... 82
Figure 34. Percent Employed (Mean % Change) ............................................................. 83
Figure 35. Proportion Nonwhite (Mean % Change) ......................................................... 83
Figure 36. Per Capita Income (Mean % Change) .............................................................. 84
Figure 37. Per Capita Officers (Mean % Change) - Jurisdiction Level ............................ 85
Figure 38. Illustration of the “Canceling Effects” of Fixed Effects Regression ............... 95
Figure 39. Composition of Summative Arrest Indexes .................................................. 103
Figure 40. Composition of Summative Indexes Pertaining to Policing Practices .......... 107
Figure 41. Initial Models of Crime as Explained by Ofﬁcers, Drug and

Disorder Arrests ............................................................................................................... 1 12
Figure 42. Formula for Calculation of Effect Sizes in Granger Analyses ...................... 113

xv

 

Figure 43. Final Model of Crime as Explained by Ofﬁcers, Drug and
Disorder Arrests ............................................................................................................... 1 14

Figure 44. Initial Model of Crime as Explained by Policing
Practice Indexes ............................................................................................................... 1 17

Figure 45. Final Model of Crime as Explained by Policing
Practice Indexes ............................................................................................................... l 19

Figure 46. Initial Model of Crime as Explained by Individual Policing
Practice Items ................................................................................................................... 1 19

Figure 47. Final Model of Crime as Explained by Individual Policing
Practice Items ................................................................................................................... 120

Appendix A. Figures 1-6. National Crime Trends (Summed Rates) ............................. 226

xvi

CHAPTER 1
INTRODUCTION

What impact did changes in policing practices have on crime during the 1990s?
During this decade a steep decline in crime rates occurred within nearly all major crime
index categories (murder, robbery, aggravated assault, burglary, larceny, auto theﬁ) in
jurisdictions across the United States. Subsequently, a number of researchers reviewed
broad systemic changes that were thought to be key determinants yielding inconclusive
results (Blumstein et al., 2002; Lafree, 1998a; Travis & Waul, 2002). The list of causal
factors that have been hypothesized pertaining to this rapid and substantial decrease in
crime (often referred to as the "crime drop") includes changes in population age
demographics and drug trafﬁcking, improvements in labor market conditions, perceptions
of the legitimacy of the criminal justice system, and increases in incarceration (Blumstein
& Wallman, 2000; Lafree, 1998b; Ralston, 1999). However, a number of criminal justice
researchers have also asserted that changes in policing might have contributed to the
nationwide decline (Eck & Maguire, 2000; Lafree, 1998b; Ouimet, 2002; Sherman,
Gottfredson, MacKenzie, Eck, et al., 1998). These assertions regarding the possible role
of new policing efforts have not yet been rigorously tested using a methodology that
allows generalization of ﬁndings to jurisdictions across the nation.

At present, little is known about which speciﬁc changes in policing tactics,
organizational strategies and technologies were most closely associated with reductions
in crime during the 1990s. For example, no study has yet been published that assesses
the relative effect sizes or amounts of possible contribution that these policing practices

may have demonstrated with regard to declines in various types of offenses such as

robbery, burglary or larceny. This study is intended to provide a starting point for
researchers in their efforts to identify mechanisms through which policing efforts may
reduce crime by examining relationships between speciﬁc types of police practices and
crime rates. Additionally, through the assessment of changes within jurisdictions of all
sizes and hem all regions of the nation, policing practices that demonstrated robust
relationships with crime among a variety of environments and organizational contexts
will be identiﬁed, providing crime control information for police administrators.
Changes in Policing and the Crime Decline

Technology and Community Policing

During the 19903, the institution of American policing began to incorporate
different tactics, strategies and technologies resulting in what could be described as the
most substantial transformation in its history. The emergence of personal computers,
vehicle-mounted mobile data terminals, cellular phones and investigative soﬁware
provided real-time information that allowed ofﬁcers in the ﬁeld to take action as events
were occurring. Additionally, these technology changes were accompanied by a
paradigm shiﬁ from a "professional model " that was designed to detach individual
ofﬁcers ﬁ'om the community to an approach where many police organizations began to
embed routine citizen contact and feedback into decision-making processes pertaining to
neighborhood problems and potential solutions (Braga, 2002; Skogan & Frydl, 2004;
Trojanowicz, Kappeler, Gaines, & Bucqueroux, 1998). As a result, hundreds of
thousands of police ofﬁcers were encouraged to address neighborhood disorder and
citizen concerns that previously may have been considered to be beyond the scope of

traditional law enforcement (Braga, 2002; Roth, Ryan, Gaffigan, Koper et al., 2000).

 

Systematic Problem Solving Efforts

This era also emphasized systematic problem solving such that police began to
actively and routinely intervene to limit or remove recurring situation-speciﬁc factors that
were conducive to criminal activity, such as poorly managed taverns that allow
intoxicated patrons to congregate outside the building and dimly lit parking lots that
provide a location for the distribution of contraband (Braga, 2002). The systematic
search for solutions to recurring problems led to increased police awareness of crime
“hot-spots” which are geographic locations or buildings where criminal activity is
frequent (Eck & Weisburd, 1995; Sherman, 1995). Subsequent strategies that emerged
with regard to hot spots involved police efforts to simply block or deny criminal
opportunities in speciﬁc places by concentrating patrol presence (Fritsch, Caeti & Taylor,
2003) and address problems indirectly through the use of civil code enforcement against
absentee landlords or inattentive property managers (Mazerolle & Roehl, 1998).

Policing Research - Paradigm Shift

Criminal justice research during the 19903 also began to shiﬂ from a debate about
whether the police could signiﬁcantly impact crime toward analysis of the conditions
under which police crime-prevention activities could be most effective (Sherman,
Gottﬁ'edson, MacKenzie, Eck, et al., 1998). In years prior, some prominent researchers
had suggested that police should abandon their crime prevention efforts. One such
researcher declared that “Suppression of crime is a will-of-the-wisp which the police
should no longer pursue.”(Walker, 1984, p. 88). However, within 10 years after this
declaration, a number of crime-prevention studies indicated that new policing tactics

could substantially reduce crime in places where criminal activity was most prevalent.

One recent review that assessed the impact of concentrating police presence in high-
crime areas indicated that 12 of 13 studies demonstrated local crime reduction effects
(Scott, 2004). Additionally, case studies have found that problem solving efforts were
associated with reported crime reduction in New Jersey, Oakland, and Santa Anna
(Braga, 1997; Green, 1995; Jesilow, Meyer, Parsons, et al., 1998).
Incarceration and Police Force Size

Although these signiﬁcant changes in policing practices occurred at
approximately the same time as one of the largest and longest lasting crime decreases in
American history, some researchers have suggested that increases in incarceration or the
number of ofﬁcers per capita may be more important driving forces behind the crime
drop of the 19903 (Kleiman, 2000; Levitt, 2004; Marvell & Moody, 1996; Rosenfeld,
2000). However, it is important to note that many places that experienced decreases in
crime during the 19903 did not increase incarceration rates or overall size of police forces
relative to population. Additionally, researchers have suggested that among the largest
cities that experienced a crime reduction from 1990 to 1996, some did so without a
substantial increase in ofﬁcers per capita (Eck & Maguire, 2000). In a recent review of
this issue, Eck and Maguire pointed out that San Diego had one of the largest crime
reductions during this period but their ofﬁcer to citizen ratio increased by only 1%.
Dallas and Seattle also reported substantial crime decreases while ofﬁcer rates declined
approximately 3%t and 6% respectively (Eck & Maguire, 2000). In addition, the nation
of Canada has historically experienced very similar violent crime patterns to those in the

US. including a substantial decrease in crime during the 19903. However, the ratio of

officers to citizens did not increase in that nation during this period and incarceration
rates actually declined as a result of budget shortfalls (Ouimet, 2002).
Characteristics of the Crime Drop

Another useful mechanism in analyzing the impact of recent policing changes in
the US. is to examine the trends of the dependent variables (i.e. crime rates). At an
aggregate, national level, the rate of violent felony index offenses across the nation
peaked around 1991 and then began a sharp decline around 1993 (Figure 1). This decline
continued throughout the 19903 (Blumstein, Steinberg, Bell, & Berger, March, 2002).

Figure 1. National Violent Crime Index
(UCR 1980-2001)

 

800
700 4
600
500 ~
400
300 .. A , , _ z

200 I . T : T T T T v r T
1980 1985 1990

 

 

 

 

 

 

 

 

I T T T

1 995 2000

 

 

 

However, national trends in crime are driven by large cities and when subgroups
of cities with populations below 100,000 are examined, the decline of violent index
offenses appears substantially later as city population decreases. Using the 1990 violent
index rate as a baseline, for example, UCR data indicate that 60% of cities with

population greater than 100,000 reported a decline in violent index offenses in 1996 1.

 

‘ Analysis of 8,940 agencies that reported a full 12 months of data from 1990-2001 by population
subgroups indicates that the violent crime trend occurred in larger places several years prior to this trend in
smaller cities. A “trend” was deﬁned as at least 60% of agencies within the subgroup reporting violent
crime at below 1990 levels. See pp. 51-52 for further details.

However, as shown in Figure 2, this trend did not occur in cities with populations from
50-75,000 until 1998, followed by cities of 25-50,000 population in 1999. Further, data
indicate it was nearly 5 additional years (post 1996) before 60% of cities with population
10,000-25,000 would report violent crime rates below 1990 levels. Thus, although crime
trends across jurisdictions in the US. tended to decline during the 19903, they did not
take place uniformly with regard to temporal occurrence.

Figure 2. Trends in Agencies Below their 1990 Levels
for Violent Crime by Jurisdiction Size

 

 

 

>150,000
100to150k
75to100k
50to75k
25t050k - .....
10t025k -_ . . ~ ”Vi
<10,000 j

 

 

 

 

 

 

 

 

l

N '1. '5 <9 '\ e e e x
6% ’99 '99 \gei‘NQQ 94°99 '99 '90. (190 “96

 

 

 

Recent research has also suggested that the spatial distribution of reduction in
crime may have been uneven or disproportionate within city boundaries. In Seattle,
Washington for example, one study indicated that the vast majority of the overall city-
wide reduction in crime during the 19903 took place within a few narrow high-crime
neighborhoods (Weisburd, Bushway, Lum & Yang, 2004). Approximately 86% of city
sectors with low and moderate historical levels of crime did not experience a signiﬁcant
decrease and several of these sectors actually experienced a crime increase.

Consequently, although the overall city crime rate declined by nearly 24%, the physical

locations where substantive crime reductions actually took place represented only about
14% of the land mass of the city.
Statement of the Research Problem

Given that principles of causal inference generally require that cause precede
effect and that effects occur whenever causal conditions are met, the types of variables
that should be examined most closely as contributors to the crime drop during the 19903
should include factors that emerged (or changed) similarly in all places where declines
occurred, effected large cities ﬁrst followed by places with smaller population, and
impacted speciﬁc, high-crime segments of cities to a substantially greater extent than
places with low or moderate historical levels of criminal activity. It is noteworthy,
however, that very few of the commonly asserted causal variables with regard to the
recent crime drop actually meet these criteria. Broad changes in labor market conditions,
for example, tend to impact large and small cities at approximately the same time and
would likely effect low and moderate crime neighborhoods as well as those relatively few
places where crime has been previously most problematic. Changes in age demographics
might be expected to impact large and small cities as well as high and low crime
neighborhoods in similar manner with regard to timing and neighborhood impact.
Additionally, as previously mentioned, increases in incarceration and police ofﬁcers per
capita were not universal among places where crime declined.

Recent changes in policing tactics and technologies, however, do meet these
causal criteria and are consistent with patterns of crime reduction that occurred during the
19903. Technological advances in policing began in large agencies and diffused to

smaller jurisdictions as cost decreased (Eastern Kentucky University, 2003; Mullen,

 

1996). Additionally, recent studies indicate that the implementation of problem solving
and community policing approaches expanded throughout the US. and Canada during
the 19903 (McKenna, 2000). However, as with technological innovations, data
pertaining to policing practices indicate that large agencies began to implement
systematic problem solving and community policing approaches several years prior to
places with smaller population (Rosenthal & Fridell, 2002; Roth et al., 2000). There is
evidence, for example, suggesting that agencies below 50,000 population required several
additional years to implement policing innovations (Roth et al., 2000). Moreover,
problem and place oriented tactics were often directed toward hi gh-crime places and
would likely result in the crime reductions that were concentrated in speciﬁc high crime
areas, consistent with patterns that were observed in Seattle.
The Need for the Current Study

Despite the evidence that recent changes in policing tactics and technologies are
consistent with patterns of crime reduction throughout the US. during the 19903, the
relationship between these policing changes and crime rates has not yet been closely
scrutinized across jurisdictions of various sizes nationwide. Although a number of
researchers have acknowledged that such a relationship likely exists (Eck & Maguire,
2000; Lafree, 1998b; Ouimet, 2002; Sherman et al., 1998), there are at least four factors
that may have contributed to this lack of speciﬁc attention in recent criminal justice
literature. First, consistent evidence that speciﬁc types of policing actions such as
problem solving may impact crime in places has emerged only recently (since the early
19903). As a result, there has not been a great deal of available data pertaining to the

implementation of speciﬁc policing tactics that would allow inference to jurisdictions

throughout the nation. Second, the relationship between police tactics and crime
reduction is likely to be complicated by differences in jurisdiction size, region and
composition. For example, certain tactics such as foot and bike patrol that are commonly
utilized in urban areas may not be feasible in rural locations where the population is more
widely dispersed. Additionally, seasonal differences may decrease the effectiveness of
these tactics in regions of the nation where cold and inclement weather persists
throughout much of the year. Third, the effects of speciﬁc tactics may vary based on
differences in police organizational structure, personnel and extent of implementation.
For example, strategies that focus on crime hot spots may require that the number of
police ofﬁcers per capita be at a certain level to avoid displacement of criminal activities
to other areas of the city. Finally, there is a possibility that changes in policing tactics
may be reciprocally or simultaneously related to crime patterns such that increases in
crime result in changes in tactics and these new tactics also result in changes in crime. If
crime and tactics are endogenously related, the effects of speciﬁc policing efforts may be
difﬁcult to measure unless a sufﬁcient instrumental variable can be found to assist in the
isolation of these relationships.

Purpose and Objectives of this Study

Despite these research challenges there is emerging evidence that policing tactics
and strategies can result in crime reduction. Furthermore, recent changes in the
implementation of problem solving, use of civil and regulatory codes, crime analysis and
other policing practices have been consistent with recent patterns of crime reduction
across the United States. Consequently, although the purpose of this study is not to
explicitly model the “crime drop,” it is intended to provide a more detailed examination

of these relationships in an effort to provide useful information for researchers and

 

practitioners. At present, no nationally representative study has examined the effects of
speciﬁc policing tactics on crime. In this regard, it has been noted, for example, that the
vast majority of case studies concerning problem and place oriented strategies have been
conducted in urban environments and have involved large police agencies. Thus, it is
currently unknown whether relationships from case studies of problem solving, use of
civil and regulatory codes to reduce disorder and other recent policing innovations will be
found in environments that substantially differ in terms of population demographics, level
of urbanicity, and police agency size.

Additionally, at present, it is not known to what extent speciﬁc types of tactics
might be effective against different types of offenses. Foot patrol, for example may be
effective in reducing vehicle theft as a result of a increased police presence (Di Tella &
Schargrodsky, 2004). However, this approach may not demonstrate any effect with
regard to the offense of murder, which often occurs in locations that are not easily
surveilled. The purpose of this study is to examine these relationships in jurisdictions
across the nation in a manner that will identify recent policing changes that were
associated with crime reductions to enable further research pertaining to the mechanisms
by which community security is effected. Additionally, by identifying the speciﬁc
policing tactics and strategies that were associated with reductions in murder, rape,
robbery, aggravated assault, burglary, larceny and vehicle theft, policy recommendations
regarding tactics that should be emphasized or selected when only a few are possible

given limited agency resources may become apparent.

10

CHAPT ER 2
REVIEW OF THE LITERATURE - POLICING CHANGES

Events Leading to the Paradigm Shift of the 19903

Prior to the 19903, a number of criminal justice researchers asserted that police
actions could not substantially affect jurisdiction-wide crime rates (Bayley, 1994;
Walker, 1984). There were both theoretical and empirical reasons supporting this
assertion. From a theoretical standpoint, criminologists had argued that a substantial
portion of crimes were committed as consequences of extreme poverty (Cullen & Agnew,
2003). Consequently, since police could do little to address this causal factor, it was
assumed that little could be done to reduce crime rates. Additionally, studies regarding
incarceration suggested that "nothing worked" with regard to recidivism (Martinson,
1974). Finally, in the policing arena studies did not demonstrate any crime reduction
effects from randomized patrol and shorter response time to criminal incidents (Kelling et
al., 1974; Sparrow, Moore & Kennedy, 1990). Consequently, a primary mechanism of
crime control under the professional model of American policing was to seek the arrest of
more serious offenders after their crimes were committed (Skogan & Frydl, 2004;
Trojanowicz et al., 1998; Visher, 2000).

During the 19803 and early 19903, however, research ﬁndings increasingly
suggested that speciﬁc types of police activities could impact community security and
perceptions of well being. For example, a Neighborhood Foot Patrol Program in Flint,
Michigan was associated with crime rate reductions of 8.7 percent (Trojanowicz, 1982).
Additionally, the Newark Foot Patrol Experiment was associated with improved
perceptions of neighborhood safety among citizens (Police Foundation, 1981). In

Houston, implementation of community police stations, foot patrol, and community

11

organizing teams also resulted in decreased fear as reported by area residents (Pate,
Skogan, Wycoff & Sherman, 1986). These modest successes highlighted the connections
between police-community relations, citizen fear, and crime, encouraging a paradigm
shift where police researchers began to seek out conditions under which crime and fear
could be reduced, rather than focusing on police responses after events unfolded.

During the late 19803, concepts relating to community oriented policing and
problem solving approaches were increasingly adopted by police agencies (Braga, 1997).
Although there is no universal agreement as to the precise deﬁnition of community
oriented policing, this approach emphasizes efforts to improve the quality of police
services by reducing disorder and fear while improving the ability of citizens to engage
community self-regulation. In contrast to traditional policing, which emphasized
efﬁciency and limiting ofﬁcer mistakes by reducing discretion, community policing
advocates argued for a broader police role (Bayley, 1994; Trojanowicz et al., 1998). The
concept of systematic problem solving also emerged as a tool to assist police agencies in
addressing aggravating factors that result in recurring crime problems (Goldsetin, l979)2.
Frequently, problem solving was employed as a technique to achieve broader community
policing objectives (Braga, 1997; Cordner & Biebel, 2005).

By 1985, the Police Executive Research Forum had evaluated several problem
solving efforts by the Newport News Police Department in Virginia (Eck & Spelman,
1987). One of these efforts included the demolition and rebuilding of an apartment
complex, resulting in a 35 percent reduction in burglary. Another effort in Baltimore
County, Maryland involved the hiring of an additional 45 ofﬁcers and creation of a

special unit with the mission of reducing fear and developing stronger community ties

12

(Cordner, 1988). Subsequent surveys of residents indicated that fear decreased
signiﬁcantly after the special unit was operational.
Policing Changes During the 19903

The number of police agencies that reported transitioning from traditional law
enforcement to a community oriented policing approach rose from around 50% in 1994
to nearly 80% in 1998 (Breci & Erickson, 1998; Johnson & Roth, 2003; Oliver, 1998).
Frequently, these transitions to community policing were accompanied by the permanent
assignment of officers to speciﬁc areas so that they could become familiar with residents
and the unique characteristics of each neighborhood (Kelling & Moore, 1991).
Additionally, data from a study of two large agencies indicated that community police
ofﬁcers tended to spend less time on general patrol than traditional ofﬁcers and more
time in community meetings and problem solving activities (Parks, Mastrofski, Delong,
& Gray, 1999). Whereas traditional policing only required officers to react to crisis,
community ofﬁcers were expected to have the ability to address problems before a crisis
occurred (Reiter, 1999).
Efforts to Improve Communig Apmarance and Orderliness

One method that was utilized by police organizations to achieve this new
proactive agenda was to attempt to improve community appearance and orderliness.
These efforts were based, in part, on criminal justice theories that asserted that patterns
or cycles of interrelated behavior within communities and neighborhoods could create an
environment that made criminal behavior more likely (Eck & Weisburd, 1995; Green,

1995; Scott, 2000). Thus, if these cycles of behavior could be altered, it was argued that

 

2 Strategic problem solving will be discussed in detail in a subsequent section.

 

police might be able to decrease or disrupt undesirable pattems of behavior, rather than
simply responding to crime after it occurred.

Under the ecological perspective, for example, Burgess’ concentric zone theory
(Burgess, 1925) asserted that rapid population turnover and urban decay led businesses to
move out of the centers of cities, causing resources, routine maintenance, and property
upkeep to decline. However, from a theoretical perspective, the subsequent decline in
property values within these central urban zones could also create an incentive for new
development if fear and social incivilities could be reduced. Toward that end, some
police agencies began to coordinate with business leaders, schools and local community
groups to support the removal of grafﬁti and reduction of juvenile loitering, and public
drug dealing. One such project in Oakland, California involved the creation of a special
team (known as SMART or Specialized Multi-Agency Response Team) which
coordinated with other city agencies to inspect and identify drug/problem properties that
failed to comply with civil ordinances. This team utilized regulatory codes and drug
abatement laws to control drug and disorder problems. The study examined 321 places
using time series analysis and ﬁndings indicated that calls regarding disorder and drug
offenses were reduced while the overall appearance of the neighborhood improved
(Green, 1995).

During the 19903, a central tenet of the community policing movement involved
the expansion of the functional role of police to include efforts to address “quality of life”
issues in communities (Trojanowicz et al., 1998). Much of the motivation behind this
broadened policing role emanated from the “broken windows” thesis, which asserted that

signs of disorder and decay in the visible environment (e. g. litter, grafﬁti, abandoned

automobiles, boarded-up buildings, and other signs of physical disorder) serve as cues to
residents and passers by that residents in the neighborhood do not care about disorder,
cleanliness and upkeep (Wilson & Kelling, 1982). An extension of this thesis further
suggests that “behavioral incivilities” such as aggressive panhandling, public urination
and defecation, prostitution, loitering, public intoxication and other signs of social
disorder increase perceptions that personal victimization may be likely, causing law-
abiding citizens to fear for their safety and thus, attempt to avoid the area (Field, 2002;
Kelling & Coles 1996; Skogan, 1990). Consequently, one police strategy that was
developed during the 19903, attempted to utilize civil and regulatory codes to persuade or
coerce law abiding individuals and non-offending third parties to take responsibility and
action to prevent or end criminal or nuisance behavior (Mazerolle & Roehl, 1998).
Tactics that were utilized in conjunction with this strategy included providing
training for landlords and property managers, ﬁnes for non-compliance, and civil actions
to achieve property forfeiture. In San Diego, California, an experiment was conducted
among 121 rental properties that had been recent targets of drug enforcement (Eck &
Wartell, 1998). Landlords in one treatment group received a follow-up letter indicating
that the city could take legal action if the drug problem was not addressed. Landlords in
a second treatment group were contacted by a narcotics detective and city code
enforcement official who inspected the property and worked with the property owner to
devise a plan to remove or reduce the drug problem. A 30 month review indicated that
crime had declined by approximately 60% in places where landlords were met by city
ofﬁcials while crime in places that had received a letter of warning declined by 13%

when compared with a control group.

l5

 

These types of non-traditional approaches became more popular as law
enforcement ofﬁcials recognized that criminal remedies were often ineffective for dealing
with recurring behaviors or problems that increase the likelihood of crime but are not
criminal in and of themselves. For example, the purchase of spray paint by minors was
legal in many jurisdictions but often led to problems with grafﬁti. As a result, another
strategy that was employed was to seek legal prohibitions of spray paint sales to minors,
drug paraphernalia, cigarette vending machines and other items that may support or
encourage criminal activity (Mazerolle & Roehl, 1998).

Civil remedies were also expanded to speciﬁcally target offending parties such as
gangs, batterers, and delinquent youth. In Southern California, police and legal ofﬁcials
gathered evidence that gang activities created a “public nuisance” and sought gang
injunctions that would allow civil ﬁnes or arrest of certain individuals that were identiﬁed
as gang members (Maxson, Hennigan & Sloan, 2003). Subsequently, Ventura and
Redondo Beach ofﬁcials reported a near 90% decrease in gang crimes and a 70%
decrease in police contact with gang members (Maxson, Hennigan & Sloan, 2003).
Additionally, a time series analysis of violent felony offenses in 14 Los Angeles counties
between 1993 and 1998 indicated that these offenses declined by 5-10% overall when
compared with the pre-injunction period (Grogger, 2002).

By 1998, civil remedies were described as “the norm,” rather than the exception
by some researchers (Mazerolle & Roehl, 1998). However, although the use of
regulations, civil gang injunctions, and asset forfeiture may have provided strong
leverage for law enforcement ofﬁcials, some researchers suggested that extensive use or

misapplication of these tactics could result in new negative consequences in some

16

communities. In the legal arena, constitutional challenges concerning the breadth of civil
gang injunctions were made and some California courts required that ofﬁcials provide
evidence of conduct that posed a substantial threat to the health or property rights of
citizens before action was allowed (Cheh, 1998). Additionally, in order to be effective in
the long term, these invasive non-criminal measures needed to be supported by the
community.
Improving Citizen Collaboration as a Tactic

It is noteworthy that this substantial increase in law enforcement intervention in
the routine affairs of citizens might not have been tolerated under the traditional or
professional model of policing. During the professional era, concerns over the corruption
of ofﬁcers who developed close contacts with community ofﬁcials and individual citizens
resulted in policies that required the rotation of ofﬁcers to different neighborhoods or
areas of the jurisdiction, often on a daily or weekly basis (Sparrow, Moore & Kennedy,
1990). Unfortunately, this procedure, combined with decreased usage of foot patrols
(due to the emergence of patrol cars) and urban geographic expansion, led to a decline in
routine police-citizen interactions (Bratton & Knobler, 1998; Sparrow, Moore &
Kennedy, 1990). Consequently, some researchers argued that this relative lack of
interaction resulted in distrust between citizens and police ofﬁcers (Trojanowicz et al.,
1998). Additionally, some research suggests that crime increases and decreases oﬁen
correlate with, or are causally related to public perceptions regarding the legitimacy of
the government and criminal justice system (Lafree, 1998a, 1998b). There is evidence
that citizen trust and perceptions of legitimacy increased signiﬁcantly during the 1990s

(Skogan & Hartnett, 1997). Nevertheless, a substantial portion of the aforementioned

l7

civil remedies and “quality of life” tactics may not have been possible if police-
community relations had not improved.

A related theme in recent criminal justice literature asserts that policies of the
professional policing model increased isolation of police ofﬁcers from the needs and
concerns of the communities that they served (Sparrow, Moore & Kennedy, 1990). As a
result, a substantial portion of policing literature has asserted that the types of
neighborhood problems and environmental inﬂuences that were distressing to citizens
often differed from those to which ofﬁcers invested much of their time and effort (Carter,
1995; Kelling & Coles, 1996; Trojanowicz etal., 1998; Wilson & Kelling, 1982).
Consistent with the traditional and more narrow law enforcement role, police agencies
had invested resources in the apprehension of felons after they had committed serious
offenses through rapid response to calls, use of tracking dogs, and vehicular pursuits
(Sparrow, Moore & Kennedy, 1990). However, citizens were often more interested in
safety, physical decay and minor but routine disorderly conduct (Carter, 1995;
Trojanowicz et al., 1998). Consequently, agencies that improved their collection and
processing of citizen feedback may have been more effective in achieving crime
reductions than those that did not.

Combating Fear as a Tactic

Another aspect of the relationship between communities and crime that received
increased attention during the 1990s was citizen fear of victimization (Field, 2002;
Skogan & Frydl, 2004). Fear is important not only as a quality of life or police service
issue, but because a number of theoretical perspectives assert that it may actually effect

crime rates indirectly. For example, the standard model of social disorganization

suggests that fear leads individuals to physically and psychologically withdraw from
community life (Shaw & McKay, 1942). This withdrawal may lead to a weakening of
the informal social control processes that inhibit crime and disorder resulting in further
deterioration of the neighborhood.

However, in recent years, a number of researchers have also suggested that the
relationship between fear and disorder is reciprocal or dynamic as illustrated in Figure 3
(Bursik & Grasmick, 1993; Markowitz, Bellair, Liska, & Liu, 2001; McGarrell,
Giacomazzi & Thurman, 1999). A recent study that employed structural equation
modeling to analyze three waves of interviews from over 10,000 British residents found
evidence for the existence of a reciprocal relationship between fear and disorder as well
(Markowitz, Bellair, Liska, & Liu, 2001). Additionally, researchers that assessed crime
and disorder in 196 neighborhoods found that the inclusion of a reciprocal effect between
disorder and violence improved the overall ﬁt and explanatory power of their models

(Sampson & Raudenbush, 1999).

Figure 3. Partial Model of Social Disorganization and Fear

 

4

 

Neighborhood cohesion + Disorder/Crime ‘5 Fear

 

The existence of a dynamic or reciprocal effect involving fear, disorder and crime
may have important implications for policing tactics. If, for example, police are able to
disrupt this feedback loop by signiﬁcantly decreasing citizen fear, the resulting decreases

in disorder and crime could be substantial. Interestingly, a number of studies have

19

suggested that community policing as a general strategy or speciﬁc components of
community policing have resulted in measurable reductions in citizen fear. Assessments
of the impact of foot patrols in both New Jersey and Michigan suggested that citizen fear
was reduced (Braga, 2002). Additionally, a review of community policing studies by the
Committee on Law and Justice, sponsored by the National Research Council found
evidence that door to door contacts by patrol ofﬁcers, storefront ofﬁces and community
newsletters regarding crime often reduced perceptions of disorder (Skogan & Frydl,
2004)

Another noteworthy ﬁnding with regard to the relationship between fear and
disorder is that community fear can arise from observations of physical signs or cues (e. g.
abandoned cars, trash or grafﬁti) that disorder in the area may be problematic (Perkins &
Taylor, 1996). In this regard, Perkins and Taylor (1996) obtained data from 412 residents
across 50 block segments within 50 different neighborhoods in Baltimore, Maryland and
found that the relationship between disorder and fear was generally supported at the
individual, block segment and community levels. Further, their analyses of surveys and
on-site observations suggested that citizens rely both on physical and social cues in
assessing the danger of victimization. This ﬁnding may suggest that clean-up, ﬁx-up
projects may have crime control implications if residents fear is signiﬁcantly reduced.
The Immrtance of Social Cohesion, Collective Efficacy and Networks

Ecological and social disorganization perspectives suggest that social cohesion
may play an important role in regulating community crime in addition to that of citizen
fear of victimization. In this regard Hunter (1985) argued that informal and formal social

mechanisms can be classiﬁed into three levels of control (private, parochial, public).

20

Private control is the most basic level and involves interactions among intimate groups
(e. g. family and close ﬁ‘iends). The primary mechanisms of inﬂuence under this category
involve allocation or withdrawal of social support which can result in direct criticism,
ridicule or ostracism from the group. Parochial inﬂuence comprises the second level of
social control via broader interpersonal networks and interconnections with local
institutions such as churches, stores and schools. The ﬁnal level of social control takes
place in the public arena. Inﬂuences in the public sphere emanate primarily from formal
process and allocation of resources via institutions such as police agencies, juvenile
justice ofﬁces and social welfare organizations.

A similar line of reasoning with regard to mechanisms of community control can
be found in the work of Greenberg et al. (1982), who identiﬁed three additional forums
that are dedicated to community self-regulation (Bursik & Grasmick, 1993). First,
informal surveillance was described as typically casual but inﬂuential in terms of the
observation of neighborhood streets. Additionally, it was asserted that movement-
goveming rules may result from the avoidance of areas in the city that are viewed as
unsafe. Lastly, direct intervention or questioning of strangers and residents in the
neighborhood about suspicious activities may be utilized to exert inﬂuence on individual
behavior.

Consistent with these concepts regarding levels of control and community self
regulation, researchers have also advanced the idea of collective efﬁcacy, or shared
community expectations of intervention to maintain community stability and order
(Sampson, & Raudenbush, 1999). For example, Sampson (1986) argued that rapid rates

of population turnover and increases in structural density can lead to a greater proportion

21

of strangers and neighborhoods who are unlikely to intercede on behalf of local residents.
Thus, these systematic social processes may decrease the collective efﬁcacy in some
communities.

Research has been generally supportive of levels of control, collective efﬁcacy
and self regulation as contributors to community social control. One study implemented
a systematic social observation of urban neighborhoods via a video taping method to
assess crime and disorder in 196 neighborhoods (Sampson & Raudenbush, 1999).
Analyses of these data led authors to assert that collective efﬁcacy was linked to lower
rates of violent crime after controlling for disorder and the reciprocal effects of violence.
Additionally, data from a representative survey of 2,482 Illinois residents suggested an
interaction-effect between fear and social ties (Ross & J ang, 2000). The authors of this
study concluded that interpersonal social connections decreased fear of victimization and
that participation in neighborhood organizations provided a “buffering effect” with regard
to fear and perceptions of disorder (Ross & J ang, 2000).

However, Carr (2003) has cautioned that the social disorganization perspective
does not necessarily imply that social cohesion and collective efﬁcacy are the sole factors
that inﬂuence local crime rates. Findings from his ﬁve year ethnographic study of a
Chicago neighborhood indicated that residents have changed what they do in response to
controlling youths whose families are unwilling or unable to exert sufﬁcient private
control. Rather than contacting parents of problem youth directly, many residents in this
study contacted community police ofﬁcers who intervened on their behalf. Carr asserted

that this approach has resulted in a new form of parochialism that involves increasing

22

private reliance on public ofﬁcials at key points of intervention to achieve community
self-regulation (Carr, 2003).

Carr’s ethnographic analysis is also consistent with recent work by Reisig and
Parks (2004), who analyzed survey data from over 6,000 citizens in 62 neighborhoods in
Indiana and Florida, ﬁnding that individuals that view the police as partners with the
community perceive fewer problems with neighborhood incivilities and report higher
perceptions of safety. Taken together, these studies suggest that if police agencies are
able to establish routine, ongoing and supportive relationships with neighborhood
residents, new public-private partnerships may result and mechanisms of social control
that were not previously employed by many communities may develop.
Directed Patrol and Renewed Crime FM

Despite assertions in criminal justice literature regarding the importance of a
broadened police role in American society, another theme that was reﬂected in the
literature during the 19903 is more consistent with the professional model in that it
maintains that some of the most signiﬁcant changes with regard to crime involved
continued efforts toward police efﬁciency. During this era, researchers became
increasingly aware that most crime was concentrated in a few speciﬁc places within
jurisdictions. For example, an analysis of 323,000 calls for police service in Minneapolis
during 1986 indicated that 3% of the geographic locations generated 50% of all calls for
police service (Sherman, 1995). Further, 5% of reporting addresses produced 100% of
calls for predatory offenses such as robbery, rape and auto theft. Consequently, although
randomized police patrol, a key component of the professional-era had been

demonstrated to be ineffective (Kelling, Pate, Dickman & Brown. 1974), a new approach

23

known as “directed patrol” was developed which required police administrators to
organize patrols more efﬁciently to target speciﬁc high-crime areas (Eck & Weisburd,
1995; Koper, 1995).

In 1992, a study of directed patrol was conducted in Kansas City with ﬁnding via
a federal program entitled “Weed and Seed” (Sherman et al., 1995). This initiative
concentrated police presence in areas with very high levels of homicide and ﬁrearm-
related violent crime. For 6 months, a group of ofﬁcers patrolled these areas without the
responsibility of responding to calls for service, providing more than 1,200 hours of
police presence in these locations. Records from this study indicate that ofﬁcers issued
nearly 1,100 trafﬁc warrants, and made approximately 600 arrests, primarily through
vehicle stops, resulting in a 65% increase in gun seizures and a 49% decrease in gun
crimes during the study period.

A number of similar studies also supported the notion that speciﬁc geographic
locations, known as “hot spots” were often the source of a disproportionate number of
offenses. Spelman (1995) assessed the variability of reported crime in places over time,
ﬁnding that around 10% of locations accounted for 50% of calls for service.
Additionally, 20% of disorder offenses and 14% of crimes against persons occurred in 56
drug crime hot spots in Jersey City, New Jersey (Weisburd & Green, 1995). Similar
patterns were also reported in the Bronx and Baltimore (Eck et al., 2000).

Evidence from these studies is consistent with a number of theoretical
perspectives and hypothesized relationships in criminal justice literature. For example,
Newman (1972) asserted a “defensible space” perspective that crime is more likely in

areas that are familiar to the perpetrator and that the probability of success is increased if

24

 

perpetrators can take advantage of buildings and locations that provide concealment.
Additionally, routine activities theory, described by Cohen and Felson in 1979 suggests
that predatory crime involves more than individuals in that it requires the simultaneous
intersection of a target, motivated offender and a suitable location mlace) for the offense
to occur. Eck (1994) added another element of inﬂuence known as place managers,
asserting that the absence of these observers is one element that may increase the
likelihood of criminal activity. Sherman (1995) also advanced a “patron hypothesis”
which suggests that certain places such as bars and nightclubs oﬂen tend to attract a
larger than normal proportion of the criminal element.

Findings from the previously mentioned studies regarding crime hot spots during
the 19903 were generally supportive of the assertions of routine activities, defensible
space and related perspectives. Over time, these concepts and empirical ﬁndings
developed into a theory of place. Thus, the physical locations where criminal activities
occur became an additional subject of analysis (Eck & Weisburd, 1995). Consequently,
one police response to these ﬁndings regarding hi gh-crime locations has been to increase
patrol and enforcement activity in these areas.

Increasing Police Force Concentration — Hot Spots

Empirical ﬁndings relating to the concentration of police in high-crime areas have
supported crime reduction effects on a fairly consistent basis. Scott (2004), for example,
reviewed 13 studies pertaining to targeted police presence and found that all but one
resulted in reductions of criminal offenses or crime related calls for police service.
Furthermore, unlike most other police tactics, targeted policing has been subjected to

rigorous evaluation. Weisburd and Green (1995) conducted a randomized experiment

25

 

among 56 drug hot spots that were identiﬁed via computer crime analysis and found
consistent strong reduction effects pertaining to disorder-related emergency calls for
service. This study also resulted in a diffusion of beneﬁts to locations outside the
experimental areas.

Another randomized controlled trial attempted to provide 3 hours of daily police
patrol of hot spots during a year long study (Sherman & Weisburd, 1995). Despite the
fact that the 3 hour goal was not consistently achieved, total crime was reduced by 6 to 13
percent in treatment locations (Sherman & Weisburd, 1995). Interestingly, lengthy
duration of police presence may not be required in order to substantially reduce disorder.
Using observational data from a one-year study in Minneapolis, Koper (1995) found that
patrol presence of about 10 minutes was sufﬁcient to generate signiﬁcantly longer time
without disorder and that uniformed police presence of longer than 15 minutes resulted in
diminishing returns.

Agggessive Policing and Crackdowns

Although there is substantial evidence that targeted police presence in hot spots
may result in crime reduction, many of these policing efforts have been coupled with
increases in enforcement of disorder or drug offenses. Thus, an empirical question arises
as to whether the cause of sustained reduction emanates from the heightened enforcement
aspect (i.e. police crackdown) or the presence of ofﬁcers in these locations. In one study,
court authorized raids of drug houses were randomly conducted at 98 of 207 locations
where at least ﬁve calls for police service had been made in the preceding 30 days
(Sherman, et al., 1995). Experimental blocks showed reductions in both calls for service

and offense reports but effects were small and decayed within approximately two weeks.

26

 

Additionally, in this study raids where arrests were made (23 out of 98) did not
consistently result in differing impacts from raids in which no arrests were made.

In another study, a crackdown in a Richmond, Virginia neighborhood resulted in a
92% reduction in reported crime during the month-long trial period (Smith, 2001 ).
However, these results also decayed after the raid-month and the cost of this policing
effort was very high. A similar decay in crime reduction effects was also reported in the
Kansas City drug raid experiment that was discussed previously. Nevertheless, some
researchers have noted that if crackdowns are followed up with further police contacts,
crime prevention beneﬁts could be reinforced and sustained for long periods of time
(Skogan & Frydl, 2004). Recall, for example, that the Eck and Wartell (1998) study
demonstrated that immediate follow up contact with landlords led to long term reductions
in offenses in speciﬁc locations. As a result, the evidence with regard to raids and
crackdowns suggests that sustainable crime reduction typically requires more than short-
term enforcement efforts.

The precise range or duration with regard to the length of the effects of targeted
policing presence has not yet been established. However, it is noteworthy that crime
patterns in at least one extensive study indicated that long term reductions in high crime
areas are possible. Analyses of over 29,000 street segments in Seattle during a 14 year
time period indicated that although this city experienced a substantial overall crime drop
in the aggregate during the 1990s, crime rates did not actually decline in the vast majority
of the places within the city (Weisburd et al., 2004). Rather, nearly all of the decrease in
criminal activities occurred within narrowly deﬁned high-crime areas that comprised no

more than 16% of geographic locations. This ﬁnding is noteworthy because it provides

27

 

evidence that crime decreases that occurred in high-crime places may have resulted from
focused or directed patrols and other place oriented tactics.
Area Denial of Criminal Activities

During the 1990S aggressive policing and hot spot policing were sometimes
combined into a “saturation patrol” that formed the basis of an area denial strategy
(Fritsch, Caeti & Taylor, 2003). In 1995, John Eck described the reasoning behind this
approach by stating that “Sellers and buyers of illicit goods and services (e. g. drugs, sex,
stolen merchandise, stolen and illegal ﬁrearms) must ﬁnd ways of meeting each other and
making exchanges in order to get the rewards they seek.” Thus, area denial approaches
seek to remove certain problematic locations from public use by offenders. Eck (1995)
also laid out a “general model” of illicit retail marketplaces, noting that routine activities
theory suggests that sellers of illegal goods and services will strive to limit the distance
that they move in order to maximize proﬁts. Eck suggested that the most desirable
marketplaces would frequently be located along major arteries where place management
was weak or non-existent. A similar argument was also made by Matthews (1993) who
noted that shoppers for illicit good and services often circle through an area, looking for
sellers. Consequently, these researchers have suggested that limiting or controlling
circular trafﬁc patterns via one-way streets and other techniques might inhibit these
routine criminal patterns (Eck, 1995; Matthews, 1993).

Although Eck’s general model has not yet been explicitly tested, a number of
police agencies during the 19908 employed area denial or saturation patrol strategies.
With funding from the COPS (Community Oriented Policing Services) ofﬁce, Dallas

police department dramatically increased patrols in areas with high gang violence in

28

1996. Results from this one-year quasi-experiment indicated 64 to 73 percent reductions
in gang-related offenses in targeted areas. These treatment areas signiﬁcantly differed
from control areas where gang offenses decreased by only 22 percent during the same
time period (Fritsch et al., 2003). A similar approach was also implemented by Detroit
police department between 1995 and 1998 yielding promising but less spectacular results
(Bynum & Varano, 2003).

Problem Solving

Place-oriented policing tactics, such as those discussed in the preceding section
fall into a broader stratagem, commonly known as problem solving, that was also
increasingly recognized and implemented by agencies across the nation throughout the
19905. Although police in America have a longstanding tradition of seeking solutions to
a variety of community problems, the implementation of a formal problem solving
strategy during the 19903 differed from most previous efforts in two important respects.
First, policing efforts focused on limiting or removing events that serve to mediate or
moderate criminal activity, rather than attempting to reduce crime directly. In other
words, police began to systematically focus on intermediate variables that repeatedly
occur between broad societal causal factors, such as poverty, and subsequent outcome of
a speciﬁc type of criminal behavior.

It is noteworthy that the amount or contribution of these intermediate-level factors
or variables to criminal activity may be unique to a speciﬁc place or situation. For
example, the contribution of poor lighting to criminal activity in a public parking lot that
is visible from a busy roadway may be less problematic than that of dim lighting in a

more obscure location. As a result, assessment of the impacts these contributing factors

29

is also needed. Criminal justice researchers encouraged police ofﬁcials to begin the
process of identifying these intermediate context-speciﬁc aggravating problems by
looking for patterns that occur with some frequency. More speciﬁcally, three main types
of “repeats” (i.e. repeat places, offenders, and victims) have been advanced as a starting
point of assessment with regard to problem solving (Eck & Weisburd, 1995).

A second difference in the implementation of this strategy during the 19903 was
that formal models of the problem solving process that facilitated systematic policing
efforts and analysis became widely known (Braga, 2002). One such design, commonly
referenced as the “SARA model,” has become virtually synonymous with the problem
solving approach. This model begins with a problem identiﬁcation phase (scanning),
followed by an assessment of recurring factors that aggravate the situation (analysis),
then action to limit, reduce or remove the aggravating factors (response) and
examination of data to determine whether the response was effective (assessment)
(Braga, 2002). This systematic approach to problem solving differs from traditional
police responses to crime problems that have often been somewhat ad hoc. For example,
enforcement of trafﬁc and criminal laws has often been left to individual ofﬁcers without
coordination from higher levels of the organization or emphasis on a speciﬁc location
(Braga, 1997). Additionally, the location of narcotics enforcement has often been
inﬂuenced by convenience or case of apprehension more than community concern.
Moreover, both of these approaches focus on crime (the behavioral outcome) rather than

recurring intermediate events as the problem.

30

Controlling Crime Facilitators

In addition to a focus on hot spots, another problem solving strategy involves
efforts to limit or control the crime facilitators such automobiles, guns, and cash. These
approaches have been more prominently implemented in Europe than in America
(Sherman et al., 1998). However, simple techniques such as requiring passengers on
busses to have exact fare and removing large quantities of coins from vending machines
have reduced thefts in some US. jurisdictions (Clarke, 2001). In a related approach in
Britain, electronic monitoring of phone booths resulted in a 49% reduction in attacks on
coin compartments in telephones. Additionally, in Australia, researchers reported
substantial reductions in vandalism incidents following a combined target hardening and
rapid repair program (Challinger, 1991 ).

Focusing on Repeat Victims

Removing substances that entice or encourage criminal behavior has also been
utilized as a victim-oriented problem solving approach. For example, the removal of
coin-vending machines reduced burglary in a public housing complex in England by 75%
and these reductions also spilled over to other houses in the area (Braga, 1997; Pease,
1991). Another approach that has been utilized in an effort to reduce repeat victimization
involved placing multiple clerks in convenience stores that have been robbed. The
evidence of the effectiveness of this technique, however, has been very mixed.

In Gainesville (Florida) the police evaluated a city ordinance requiring two clerks
to be on duty and found that convenience store robberies declined immediately after the
ordinance took place (Clifton, 1987). Nevertheless, these declines may also have

occurred because the offenders were arrested just before the ordinance became active

31

(Sherman et al., 1998; Wilson (1990). Another study, compared 230 convenience stores
with two or more clerks on duty at night, to 346 stores with only one clerk on duty and
found no impact across all stores. However, stores that had recently been robbed did
demonstrate a reduction in subsequent robberies during the study (National Association
of Convenience Stores, 1991; Sherman et al., 1998).

In terms of repeated victimization with regard to domestic violence, one study
indicated that 70% of police were not ﬁling any charges at all in residences with frequent
domestic violence calls (Braga, 2002). Once ofﬁcers and supervisors received training
and were teamed with social workers, 47% of repeated victims left abusive relationships,
compared to a control group (26%).

Another recent study indicated that the risk of burglary is predictable, ﬁnding that
properties within 400 meters of a burgled household are at a signiﬁcantly elevated risk of
victimization for up to two months aﬁer an initial event (Bowers, Johnson, & Pease,
2004). This ﬁnding suggests that police could proactively distribute ofﬁcers in to protect
neighborhoods ﬁ'om ever becoming repeat victims, rather than simply reacting after
repeated victimization occurs.

Focusing on Repeat Offenders

Focusing on repeat offenders is another problem solving approach. One such
project was conducted in Boston during the mid-19905, resulting in the arrest of more
than 20 members of a single street gang in one day after a 9 month investigation
(Kennedy, 1998). Many of these gang members were arrested on gun charges after
coordination with federal ofﬁcials. Another project was conducted in Lowell,

Massachusetts in 1997 where, one by one, 20 of the worst troublemakers were brought

32

 

into a meeting with 14 local and state ofﬁcials. These ofﬁcials informed the youths that
they were being targeted and that their unlawful conduct would no longer be tolerated
(Braga, 2002). The youths also received counseling and offers of assistance in getting
back into school or ﬁnding a job. These approaches have been noted as “promising” in
criminal justice literature and utilize a tactic known as “pulling levers” where community
ofﬁcials and citizens from a variety of criminal justice and social service agencies work
together in an effort to exert a combined inﬂuence on the youths that is stronger than that
of any single individual or organization (Kennedy, 1998; Sherman et al., 1998).

With regard to the research concerning repeat offenders, the literature has noted
that the majority of criminal offenders specialize in a type of offense such as burglary or
larceny (Chaiken & Chaiken, 1982; Spelman, 1990). However, high-rate offenders tend
to be generalists and commit four times the crimes as others (i.e. twice as many violent
and property offenses) (Chaiken & Chaiken, 1982; Spelman, 1990). As a result,
approximately 50% serious offenses are committed by a small group of frequent
offenders that comprise only about 10% of the offending population (Chaiken &
Chaiken, 1982). Consequently, it has been suggested that if police were to focus on these
relatively few high rate offenders, crime could be reduced without burdening the criminal
justice system (Braga, 2002).

A repeat offender program (ROP) in Washington, DC. was assessed by the Police
Foundation and results indicated that ROP targeted individuals were signiﬁcantly more
likely to be re-arrested, prosecuted and convicted (Martin & Sherman, 1986).
Additionally, analysis from a repeat offender research study known as “ADAM”

(Arrestee drug abuse monitoring) indicated that heavy drug users are typically frequent

33

offenders (Kleiman, 1997). Further, those with expensive habits were often both high
frequency and persistent long term offenders. Criminal justice literature indicates that
most frequent offenders also deal drugs in high volume (Chaiken & Chaiken, 1982).
These ﬁndings suggest that tactics that target drug dealing and usage may also reduce
crime if these hi gh-rate offenders can be located and arrested. However, Spelman
(1990) has noted that targeting repeat offenders is often problematic due to lack of timely
information among police agencies as well as the mobility of frequent offenders, and
surveillance coordination issues.
Increasing Police Force Size

Many of the problem solving and place oriented approaches described in this
chapter require increases in the number of police ofﬁcers that are available within the
organization. Assigning a team of ofﬁcers to a speciﬁc hot spot or task implies that calls
for service throughout the jurisdiction that were previously handled by these ofﬁcers
must be addressed by replacements. At least one analysis of policing effects, for
example, found that the increased focus on drug offenses during the 1980’s resulted in
fewer resources being devoted to other types of offenses and a subsequent increase in
property offending (Benson, Sebastian, & Rasmussen, 2001). As a result, agencies that
attempted to implement innovative policing approaches during the 19908 often sought to
increase the overall size of their forces. In 1994, federal legislation created the
Community Oriented Policing Service (COPS), which had primary objectives to
encourage the hiring of additional ofﬁcers and encouraging innovations in policing
practices in agencies across the nation (Johnson & Roth, 2003; Roth et al., 2000). In the

years directly preceding the formulation of the COPS ofﬁce, the nationwide average of

34

ofﬁcers per capita was approximately 21 per 10,000. However, after 1994, the rate
increased to over 23 per 10,000. This overall change in ofﬁcers suggests that, a typical
city of 100,000 would have added approximately 20 ofﬁcers to a 210 ofﬁcer department
(a change of approximately 10%) if the change was dispersed equally.

There is substantial disagreement among researchers as to whether marginal
increases (e. g. increases ﬁom 5% to 10%) in the number of ofﬁcers, without any changes
in policing practices can produce signiﬁcant reductions in crime. After reviewing 27
studies pertaining to increases in ofﬁcer strength, Belt and Maguire concluded that “We
are unaware of any study showing reductions in recent crime rates...” (Eck & Maguire,
2000). Furthermore, research on the police-crime relationship that was conducted prior
to 1995 generally shows police levels have little impact on crime rates (Marvel &
Moody, 1996). However, two recent analyses presented evidence that most of these prior
police-crime studies were methodologically ﬂawed and found that increased police levels
are associated with reductions in crime (Levitt, 1997; Marvel & Moody, 1996). The
primary ﬂaw in prior work emanated from failure to address the simultaneous or
reciprocal relationship between levels of ofﬁcers and crime. Increases in crime may
cause ofﬁcials to seek additional ofﬁcers. Nevertheless, these additional ofﬁcers may
result in crime reduction which may lead to layoffs or force reductions (Marvel &
Moody, 1996).

To address this problem of endogeneity, Marvel and Moody (1996) utilized
multiple time series analysis with Granger testing for direction of causality. After
adjusting for endogeneity via the lag structure in the time series, they found that marginal

increases in police per capita at the state level did result in small crime reductions during

35

the 1968-1993 period. Another study by Levitt (1997) utilized an instrumental variable
and two-stage least squares (ZSLS) regression to isolate the effects of police on crime.
An instrumental variable is one that is correlated with only one of the two primary
variables (i.e. crime or ofﬁcers) that are reciprocally related. In Levitt’s (1997) analysis,
the number of ﬁreﬁghters per capita in the same jurisdiction were utilized as the
independent variable to predict the number of ofﬁcers in the ﬁrst stage of the regression.
Next, the resulting predicted values from the ﬁrst stage regression were used as predictors
of crime. Since the number of police and ﬁreﬁghters per capita are correlated but the
number of ﬁreﬁghters is not correlated with changes in crime, it was argued that the
effects of police on crime were isolated. Levitt estimated elasticities (i.e. the percent
change in the dependent variable that results from a percentage change in the independent
variable) of -1 for the effect of additional ofﬁcers on violent crime (and -.2 for property
crime) in large cities during the time period 1970-1993. A more recent study of counties
in Florida from 1980-1998 utilized the Marvel and Moody approach and also found
officer effects. (Kovandzic & Sloan, 2002). The resulting elasticities that were reported
were -.21 for robbery and -.19 for burglary, which indicates that a 10 percent increase in
ofﬁcers would result in a 2.1 percent decrease for robbery and 1.9 percent decrease for
burglary.
Mechanisms through which Policing Effects Occur

A noteworthy critique of these studies associating marginal changes in ofﬁcer
rates with changes in crime is that none of them has included or addressed the primary
rival explanation concerning what it is that the police actually do or the tactics that are

employed (Skogan & Frydl, 2004 p. 225). Any agency that makes a decision to invest

36

additional dollars to increase ofﬁcers at a faster rate than the grth of population is, by
deﬁnition, making a conscious effort to change what it is doing in the community. As a
result, it seems likely that these changes in ofﬁcer rates might also be accompanied by
changes in policing tactics. For example, New York City’s change in policing approach
during the 19908 was accompanied by a redeployment and grth in ofﬁcers (Skogan &
Frydl, 2004 p. 225). This pattern might also occur during a change in leadership, such as
the appointment of a new chief, who then brings in different ideas to the organization
while simultaneously lobbying for additional ofﬁcers.

Nevertheless, parsing out the effects that are due solely to changes in ofﬁcer
levels and not changes in the police organization or tactics may be difﬁcult for both
empirical and practical reasons. First, because these changes may often occur
simultaneously, the timing or lag structure can not be relied upon to disentangle the
effects. Additionally, as a practical matter, data pertaining to officer strength are more
readily available on an annual basis than data pertaining to policing tactics. Finally, it is
possible that both changes in ofﬁcers and changes in policing tactics may be reciprocally
or endogenously related to crime. Consequently, the effects of ofﬁcers and tactics have
not yet been fully tested in existing literature.

Another ongoing debate within justice research literature concerns the extent to
which the effects of police enforcement (e. g. increases in police force strength) occur
through deterrence. The Economics of Crime literature is founded on the assumption that
criminals are rational and make conscious decisions about the potential costs and rewards
of criminal actions (Becker, 1968). Consequently, much analysis by economists

emphasizes the importance of a deterrent effect (Benson, Kim-Iljoong & Rasmussen,

37

 

1994; Levitt, 1997; Levitt, 1998). If the deterrence hypothesis is correct, increases in
arrests or punishment signiﬁcantly impact not only offenders that are caught by police
but also others who are contemplating similar actions. However, criminal justice
literature is much more mixed with regard the precise role and importance of deterrence
(Nagin, 1998). In a review of literature on this topic, Nagin (1998) reported that studies
of police crackdowns often show a temporary deterrent effect. However, with regard to
increasing penalties, he noted that there is little evidence that criminals are aware of most
policy changes.

A substantial point of contention with regard the deterrence issue involves
differences in levels of analysis. Most studies that have supported a signiﬁcant deterrent
effect involve analysis at an aggregate (e. g. neighborhood, city, state) level of analysis.
However, studies at the individual level have not consistently found evidence of a
deterrent effect (DeJong, 1997; Piliavin, Thornton, Gartner, & Matsueda, 1986). One
recent study of approximately 5,000 offenders provided evidence that “naive” or ﬁrst
time offenders that were incarcerated were more likely to recidivate than those who were
not incarcerated (DeJong, 1997). Additionally, incarceration did not demonstrate any
deterrent effect for either naive or most experienced offenders. In another study of 3,3 00
offenders, addicts, and at-risk youth, interviews and analyses of ofﬁcial records were
conducted to determine whether perception of risk or other sanctions would deter future
offenses (Piliavin et al., 1986). Results did not support deterrence due to perceptions of
risk or severity of punishment. However, perceptions regarding potential rewards or

gains due to criminal actions were inﬂuential in the decision making process.

38

Two other studies of the effects of aggressive policing also attempted to identify,
not only whether there was a resulting impact on crime, but the process or mechanism by
which the policing effect occurred. In a study of 35 American cities with population
greater than 250,000, Wilson and Boland (1978), found that cities with a high ratio of
trafﬁc citations per ofﬁcer (a proxy measure of a legalistic or aggressive policing
approach) had lower levels of robbery. A8 a result, they asserted that aggressive policing
may effect crime both through incapacitation of would be offenders and through a
deterrent effect that results from altering perceptions of the likelihood of being caught. In
a study of 171 cities with population greater than 100,000, Sampson and Cohen (1988)
replicated and extended this work by utilizing the number of arrests per ofﬁcer for
disorderly conduct and drunk driving as the proxy for aggressive policing. Their measure
of aggressive policing was employed as an instrumental variable to identify the crime
function via two-stage least squares regression and results supported their assertion of an
aggregate-level deterrent effect with regard to robbery. However the effects were very
weak with regard to burglary.

In sum, the literature does not generally support the deterrent effect of arrest,
incarceration or increased punishment at the individual level of analysis for most serious
or high—risk offenders when substantial rewards can be obtained by offending. However,
deterrent effects have been identiﬁed at aggregate levels of analysis. Consequently, it is
possible that the effects are too small to record at the individual level. For example,
increases in sanctions may decrease offending by some fraction of an offense at the
individual level (on average), which then can be totaled into a signiﬁcant difference in

the aggregate. Nevertheless, given that high-risk or serious offenders are not being

39

deterred at the individual level, and that these offenders commit most offenses, the
efﬁcacy of a general deterrence strategy is in question. Some researchers have noted that
broadened laws (e. g. seatbelts, new gun laws etc.) and increases in ofﬁcers that perform
jurisdiction wide preventive patrol result in “casting a wider net, ” which may have the
greatest impact on the population of generally law-abiding citizens and ﬁrst-time
offenders, rather than the relatively small population of serious offenders (Roach
Anleu,1998; Way, 2002).

Given the evidence that crime declines occurred disproportionately within high-
crime areas within some cities (Weisburd, Bushway, Lum, & Yang, 2004), if these areas
contain a greater concentration of high risk and serious offenders that are not easily
deterred by police activities, a localized deterrence effect would seem an unlikely
explanation for the crime drop. Nevertheless, it is possible that localized crime
reductions within high crime places may have been caused by incapacitation that resulted
from higher concentrations of ofﬁcers and more arrests. It is noteworthy that both prison
and jail populations across the United States increased substantially during the 19908.
For example, Bureau of Justice Statistics (BJ S) data indicate that from 1990-2000, the
proportion of the US. population that was serving time in state prisons rose
approximately 24% (Bonczar, 2003; BJ S, 2005). Data indicate that over one-half of this
total increase was due to increases in incarceration for violent offenses. Nevertheless, it
is unclear how much of this increased incarceration was due to increases in felony arrests
or other police activities versus increases in incarceration risk due to changes in court
outcomes. Additionally, it was noted that the proportion of the US. population that was

serving time in jail also increased by a slightly higher margin (26.1%) that that of state

40

prisoners (81 S, 2005). As a result, another mechanism by which police may have
impacted crime could have occurred through increased targeting of disorder and
interpersonal incivilities.

Increased Incarceration and “Zero Tolerance” Strateg’eg

 

During the 19908 a number of large police agencies implemented order
maintenance or “zero-tolerance” strategies (Green, 1999; Sherman, et al., 1998). New
York Police, for example targeted grafﬁti artists, public drunkenness and other disorderly
activities such as panhandling and public urination (Green, 1999; Kelling & Sousa,
2001). After an in depth analysis of New York’s order maintenance tactics during the
19908, some researchers concluded that this policing approach had a direct effect on
crime reduction, net of other factors, preventing over 60,000 violent crimes from 1989 to
1998 (Kelling & Sousa, 2001; Sousa, 2003). However, the success of order maintenance
tactics coincided with the implementation of a problem solving process driven by
Compstat, a crime analysis and police management program (Compstat will be discussed
in the technology section).

Nevertheless, Green (1999) noted that the number of citizen complaints against
police ofﬁcers in New York increased substantially during the 19908 and that overall
arrest rates increased by 23% in New York (40% for misdemeanor arrests) while arrest
rates fell by 15% overall in San Diego and were accompanied by similar decreases in
crime rates. Green argued that the problem oriented approach, utilized by both police
departments, was better for citizens and more efficient than aggressive policing.
However, data also indicate that the increase in citizen complaints in New York occurred

primarily in a single year (1994-1995) and was followed by a steady decline in

41

complaints through 1999 (Kelling & Sousa, 2001). Additionally, analyses of case studies
pertaining to order maintenance policing in New York suggest that police ofﬁcers
typically varied their approach to dealing with quality-of-life issues, based on the
circumstances, such that non-punitive warnings and reminders were frequent alternatives
to citation or arrest. Consequently, Kelling and Sousa (2001) argued that “broken
windows policing” as implemented by the NYPD is not the rote and “mindless zero
tolerance approach that critics often contend it is.”
Advances in Police Technology

One of the major changes in policing that was common to crime declines that
occurred throughout the US. and Canada during the 19908 relates to rapid advances in
technology (Ouimet, 2002). In fact, it is difﬁcult to conceive of a policing tactic or
strategy that was not substantially effected by the implementation of cellular phones,
computers, mobile data terminals, portable roadside breathalyzers, car-mounted cameras,
laser speed detection, investigative databases and information sharing computer
networks. The majority of these technological advances have not been scientiﬁcally
evaluated (Sherman etal., 1998). Nevertheless, one researcher asserted in 1992 that
research and development itself had become the “core technology of policing" (Reiss,
1992). Although this comment may overstate the importance of research and
development during the 19908, it highlights a key difference in more recent policing
when compared with previous decades where law enforcement was focused on
establishing minimum standards of conduct in communities.

During the 19908, the US. Congress began providing additional funding to the

National Institute of Justice (N IJ ) to assess technology needs among police agencies via

42

the Ofﬁce of Science and Technology (Seaskate, Inc., 1998) Additionally, in 1994 the
National Law Enforcement and Corrections Technology Center (N LECTC) was created
to help assess emerging technologies (N LECTC, 2005). The Ofﬁce of Science and
Technology regularly surveys police agencies and works with (N LECTC) to develop
voluntary product standards, compliance and testing processes (Seaskate Inc., 1998).
Rapid Expansion of ﬂee Teclpro—logy

In addition to the inﬂuence of federal government programs and consumer-based
electronic innovations, the expansion of technologies among police agencies during the
19908 was consistent with two long held values among many police organizations that
may have increased their acceptance and rapid implementation. The ﬁrst is the
professional era goal of efﬁciency. Throughout the three prior decades, for example, law
enforcement agencies made efforts to shorten response time to criminal incidents via
police radio dispatch, emergency (911) telephone protocols, and implementation of
single-ofﬁcer vehicular patrols (Sparrow, Moore & Kennedy, 1990). Thus, many
advances that occurred among electronic devices during the 19908 supported this
objective. For example, mobile data terminals (MDT’s), located in patrol cars allowed
ofﬁcers to directly investigate persons, vehicle license numbers and ﬁrearm registration
numbers without returning to the ofﬁce or waiting for a radio response from
overburdened dispatchers (Stockton, 1999). Additionally, handheld breathalyzers and
vehicle-mounted cameras enhanced the efﬁciency of evidence collection by allowing on-
scene (roadside) testing and documentation.

The other longstanding value among police agencies that was consistent with the

implementation of technology emanates, in part, from the fact that most agencies are

43

small, have limited resources and infrequent encounters with “high-proﬁle” criminal
cases. According to LEMAS (Law Enforcement Management and Statistics), 90% of the
estimated 18,758 police agencies nationwide have fewer than 50 sworn ofﬁcers
(Hickman & Reaves, 2001). However, the few (apx. 975) police departments with over
100 sworn ofﬁcers often have greater resources than the average agency and are more
likely to have routine involvement with serious criminal offenses that have traditionally
been viewed as “real police wor ” by both ofﬁcers and civilians. Many of these agencies
have full-time public relations ofﬁcers that actively promote and publicize departmental
achievements and some also employ research personnel who seek out and promote
innovative programs. As a result, a number of these large and mid-sized agencies
achieve a “legitimacy” or status such that others attempt to emulate them (Crank, 2003;
Mastrofski & Uchida, 1996).

The struggle for institutional legitimacy is a core value that is central to the
decision-making process in police agencies and sometimes encourages agencies to
purchase new technologies, even when more basic equipment needs have not fully been
met. For example, one agency of approximately 85 ofﬁcers purchased a new mobile
command center vehicle for its part-time tactical team for over $250,000, even though
patrol ofﬁcers lacked functional portable radios and other more basic equipment (WYPD
Report, 2000; WYTT Website, 2006). Nevertheless, the desire to have similar equipment
to what other respected agencies are utilizing has pressured even very small agencies to
embrace technologies (Eastern Kentucky University, 2003; Stockton, 1999). However,
due to lack of resources, these new technologies may require substantial time (e. g. years)

for smaller agencies to purchase and implement.

44

It is also noteworthy that crime decreases began in cities with large populations,
followed by jurisdictions of medium size and eventually effecting smaller places because
this trend is consistent with patterns of technological implementation. Data from a
national study of the computerization of law enforcement indicated that large agencies
were more likely to have incorporated computing technology into organizational
functions by 1996 (Mullen, 1996). In addition to agency size, the author of this survey of
over 300 agencies also found that having a large quantity of innovative programs (e. g.
DARE and full time community policing ofﬁcers), and a high number of civilian
employees was a predictor of computerization (Mullen, 1996). Another researcher that
analyzed 1993 LEMAS data reported that agencies with more extensive computerization
had fewer ofﬁcers per capita when compared with other agencies (Nunn, 2001),
providing some evidence that computer technology may be associated with increased
efﬁciency. Nevertheless, by the year 2000, data from a national survey of agencies with
fewer than 20 ofﬁcers indicated that approximately 76% of these agencies were also
using personal computers to conduct routine police activities (Eastern Kentucky
University, 2003).

Crime Analysis and Compstat

William Bratton, the former New York City police commissioner, has stated that
computers were an important element in allowing his agency to pinpoint emerging crime
locations and develop a response before new hot spots became highly problematic
(Seaskate, 1998; Sousa, 2003). In 1994, New York Police Department implemented a
computerized data gathering, problem solving and management program known as

Compstat (Sousa, 2003). Under Compstat, police supervisors were held accountable on a

45

weekly basis for crime that occurred in their areas of responsibility. The ﬁrst step in
achieving this accountability and the development of a local problem solving response
was the gathering of timely and actionable data (e. g. crime statistics) within the
neighborhood or even street-block level of the precinct (Sousa, 2003). These timely and
localized data would then be utilized to implement rapid deployment of ofﬁcers to
emerging hotspots. Additionally, mid-level supervisors were encouraged to formulate
innovative problem solving tactical responses to crime and disorder. Nevertheless, it is
not clear how extensively these innovative strategies other than rapid and concentrated
deployment of ofﬁcers were implemented throughout the city.

Recent analyses of Compstat have noted that the diffusion of this approach to
other agencies ranked among the quickest observed for any policing innovation
(Weisburd, Mastrofski, Mcnally, Greenspan & Willis, 2003). Among agencies with over
100 sworn ofﬁcers, 42% of those in the South, 32% in the West and 27% in the Northeast
reported having adopted Compstat or a similar method by 1999, just ﬁve years after the
program was ﬁrst implemented in New York City. Respondents to a survey of these
moderate to large sized agencies indicated that reduction of serious crime was among the
most important reasons for adoption (Weisburd et al., 2003). However, the extent of
crime reduction impact of Compstat remains a matter of some controversy among
researchers (Green, 1999).

Although the implementation of Compstat in New York was accompanied by
substantial crime decreases, critics have noted that similar decreases also occurred in
other cities that did not implement this approach and argued that this program has not yet

been rigorously evaluated (Green, 1999). Moreover, a number of criminal justice

46

researchers have suggested that the implementation of Compstat and similar programs
may actually run counter to reforms that are often cited as beneﬁcial that occurred during
the 19908. For example, Willis et al. (2004) noted that Compstat is typically
accompanied by rigid and formalized hierarchical management approaches as well
functional specialization and increased bureaucracy. It is also noteworthy that survey
respondents from agencies that have implemented Compstat reported a signiﬁcantly
higher value of management “control” of ﬁeld operations when compared with
respondents from non-Compstat agencies (Weisburd et al., 2004). These rigid
bureaucratic structures could prove counterproductive if assertions by community
policing advocates are correct regarding the importance of ﬂatter (i.e. less hierarchical)
organizational structures and increased discretion and ﬂexibility of line-level personnel
which are argued to lead to greater innovation (Weisburd et al., 2003). Nevertheless,
Compstat, like any other technology, can be viewed as an additional instrument that can
be added to the policing toolbox with regard to crime control (Carter, 2004).
Technologies may also be utilized to support community policing and problem
solving by improving communication between ofﬁcers, citizens and other government
agencies and improving information access to those at the line-level. For example, an
often overlooked technological innovation that rapidly expanded during the 19908 was
that of cellular phones. At a cost of around $35 per month, these devices increased
efﬁciency by allowing direct communication between ofﬁcers and victims and allowing
follow-up calls to be made without returning to the ofﬁce (Schuiteman, 1999).
Additionally, through the use of simple, relatively low cost pagers and cellular phones,

surveillance, undercover and foot patrol ofﬁcers could request backup or additional

47

information without leaving their posts (Schuiteman, 1999). Consequently, by the year
2000, even among smaller agencies, approximately 91% reported using cellular phones to
carry out policing duties (Eastern Kentucky University, 2003).

Another example of the relationship between innovative policing and technology
was documented in the case study of a state police agency, which received an Advancing
Community Policing (ACP) Grant from the COPS ofﬁce to purchase cellular digital
packet data (CDPD) equipment in 1997 (Nunn & Quinet, 2002). CDPD was a
technological innovation that provided less expensive data transfer via existing cellular
phone lines, rather than requiring a special purchase of radio towers for this function.
This state police agency had received previous funding from the COPS Ofﬁce to create
special community problem oriented policing units which were dispersed throughout the
state and ofﬁcers reported that CPDP technology was utilized to improve internal
communication by ofﬁcers (Nunn & Quinet, 2002). Evaluative analysis indicated that
ofﬁcers with access to CPDP technology drove signiﬁcantly fewer miles in their patrol
vehicles when compared with ofﬁcers that did not have access. Nevertheless, the
researchers argued that, although this ﬁnding provides some evidence that efﬁciency was
increased, traditional enforcement measures, such as arrests did not signiﬁcantly increase
as a result (Nunn & Quinet, 2002).

A case study of the effects of implementation of mobile data terminals was also
conducted in Forth Worth, Texas during the early 19908 (Nunn, 1993). This study
employed an interrupted time series design to assess the effect of in-car mobile data
terminals (MDT’s) on motor vehicle theft clearance and recovery rates. Mobile data

terminals allow officers to directly query the status of persons, vehicles and other items.

48

After analyzing monthly vehicle recovery rates from 1980 to 1990, the evaluator
concluded that recovery rates were associated with the use of MDT's. However, in terms
of clearance or arrest rates for vehicle theft, MDT’s had no effect (N unn, 1993).

Although these studies provide evidence that computers, MDT’s and cellular
phones oﬁen improve efﬁciency and may support innovative policing efforts, at present,
there is little indication that these technologies effect crime or arrests rates directly.
Moreover, according to Nunn’s (2001) analysis of 1993 LEMAS data, highly
computerized agencies spend more per capita on law enforcement than other agencies,
even after controlling for population density, size ﬁscal capacity, and crime levels.
However, given that these highly computerized agencies tend to employ fewer sworn
police ofﬁcers to offer police services, it is not clear whether the resulting efﬁciencies
offer sufﬁcient beneﬁts to offset smaller force strength. Nevertheless, the
implementation of in-car portable computers and mobile data terminals increased
substantially during the 19908. Even among the smallest agencies (i.e. those with fewer
than 20 sworn ofﬁcers), approximately 18% reported implementing mobile data terminals
by the year 2000 (Eastern Kentucky University, 2003). Given budget and resource
restrictions among police organizations, this near 1 in 5 implementation ratio for these
very small agencies is striking. Each MDT unit may cost several thousand dollars, such
that many large agencies relied substantially on federal grants to purchase this technology
(Johnson & Roth, 2003; Roth et al., 2000; Stockton, 1999).
Increased Federal Funding for Local Law Enforcement

According to one justice researcher, ﬁnancial relationships between the US.

Department of Justice and local law enforcement agencies were “utterly transformed”

49

(Cunniff, 2002 p.1) during the 19908 with the passage of the Violent Crime Control and
Law Enforcement Act (VCCLEA) of 1994. This act increased funding assistance to
local law enforcement tenfold and created a number of new federal government programs
to deliver funding to agencies that sought to purchase and implement technology
(Cunniff, 2002). One such program, known as MORE (Making Ofﬁcer Redeployment
Effective) was administered by the COPS ofﬁce and provided approximately 1.3 billion
dollars in ﬁrnding to over 4,500 local agencies during the 19908 (COPS, 2005). The
Santa Anna, California police department, for example, began its in-car laptop
implementation program in 1997 via funding from a COPS MORE grant in the amount of
1.8 million dollars (Stockton, 1999).

Another source of federal funding that was designed to assist police agencies was
the Local Law Enforcement Block Grant (LLEBG) program (BJA, 2002). The LLEBG
program, enacted in 1996, is a formula block grant is distributed directly to local law
enforcement agencies based on the proportion of their police expenditures in the state
(BJA, 2002). LLEBG funded more than $1.2 billion during its ﬁrst three years and
approximately 58% of these funds were utilized to purchase and implement technologies
(BJA, 2002).

Prior to the 1994 Violent Crime Act most criminal justice block grants were
awarded to the states, which then distributed the funds to local governments and were
allowed broad discretion in allocating these funds (Cunniff, 2002). Both LLEBG and
COPS grant funds, however, were provided directly to local police agencies, increasing
the speed and efﬁciency of the process (Cunniff, 2002). Additionally, federal funding to

local agencies during the 19908 was provided for a vast array of purposes including the

50

hiring and redeployment of ofﬁcers, crime prevention, and school security enhancements,
and multi-jurisdictional task forces (BJ A, 2002; Johnson & Roth, 2003; Roth et al.,
2000).

The COPS ofﬁce offered more than 30 different grant programs that awarded over
$7.3 billion to local agencies from 1994 to 2001. Approximately 64% of these funds
were provided for additional police personnel (GAO, 2005b). However, a substantial
portion of COPS grants were directed toward technology, ofﬁcer training, and programs
to encourage innovation and problem solving (Johnson & Roth, 2003; Roth et al., 2000).
For example, COPS awarded grants speciﬁcally for youth gang related policing efforts
and to combat local methamphetemine production (Roth et al., 2000). Additionally,
federal grant funding processes were simpliﬁed and streamlined to encourage agencies
that did not have prior experience with the grant process to apply (Johnson & Roth, 2003;
Roth et al., 2000). One such grant program, known as FAST (Funding Accelerated for
Small Towns) provided up to $75,000 for the hiring of an additional ofﬁcer after
completion of a single page application. As a result, a substantial number of agencies
that had never previously applied for a federal grant received ﬁrnding. One estimate
indicated that approximately 84% (GAO, 2005b) of the 13,000 agencies that reported
crime data to the FBI during the 1990-2001 time period received at least one federal
grant.

Many of these grants were awarded with requirements or instructions that local
agencies implement speciﬁc types of problem solving, community collaboration or other
innovative policing approaches. For example, MORE grant applications were awarded,

in part, based upon agency’s proposals for improving efﬁciencies in ways that would

51

allow redeployment of sworn ofﬁcers to street patrol or increased productivity (Roth et
al., 2000). Additionally, applications for the most proliﬁc COPS grant program,
Universal Hiring Program (UHP) contained a number of survey questions inquiring as to
which of a series of tactics or strategies (e. g. problem solving, crime prevention etc.)
would be utilized or increased by the agency after receiving funds to hire additional
ofﬁcers (COPS, 1998; COPS, 2000). The COPS ofﬁce also distributed instructional
guides, technical reports and training courses to local agencies via regional training
centers through partnerships with area universities and local agencies (Roth et al., 2000).
As a result, these increases in federal funding, training and leadership may also have
encouraged the adoption of innovative crime prevention policing methods during the

19908

52

CHAPTER 3
CHANGES IN CRIME TRENDS DURING THE 19908

This study does not attempt to explicitly model or explain the "crime drop" of the
19908. However, given that changes in policing tactics and strategies coincided with an
era of noteworthy downward crime trends, a review of the literature and available data on
this topic are needed to provide some context for subsequent analyses. The precise
meaning of the word "trend" as it relates to crime may differ somewhat depending on the
methodology that is being employed. At a conceptual level, the phrase "crime trend"
refers to consecutive years of increases or decreases in crime. However, the meaning of
this word as utilized in this analysis is more speciﬁc and will be deﬁned after a brief
discussion of three methodological approaches to analyzing aggregated patterns of crime.

Methodologies of Analyzing Crime Trends

Most prior research pertaining to crime trends across the nation has examined
crime at an aggregate national level by summing the total counts of offenses and dividing
them by the US. population3 as shown in the ﬁrst row of Table 1 below (FBI, 2002).
This simple approach provides a crime rate that is very sensitive to relatively small
changes in crime counts. However, since the vast majority of crime occurs within large
cities, such aggregated counts may not always provide an accurate representation of the
nation as a whole. For example, in 1990 the violent crime counts among the 10 cities
with the most crime accounted for nearly 30% of the total number of offenses among the
8,940 agencies that reported data for all 12 months. Consequently, substantial crime
reductions in these 10 cities would signiﬁcantly impact the national rate even if no crime

reduction occurred in 99% of jurisdictions. Similarly, if a single state with a large

53

population changed reporting protocols or had technical reporting difﬁculties, the
resulting effect on the national crime trend could be substantial. Consequently, although
the calculation of a national aggregated crime rate based on summative counts of crimes
can provide a quick assessment of overall trends, when not supplemented with additional
analyses, this statistic can be misleading.

There are, however, a number of other statistical procedures that can provide
additional information regarding the breadth and pervasiveness of crime trends. One
such approach, reﬂected in row two of Table 1 (below) involves an assessment of crime
at the jurisdiction level. The ﬁgures in row two were created by counting the number of
jurisdictions in each year from 1991-2001 where violent crime rates were below the
baseline year (1990) and dividing this count by the total number of agencies4 resulting in
a percentage of jurisdictions where crime declined. Table 1 indicates that in 1995,
although the national crime rate had fallen 6% below 1990 levels, nearly one half of all
jurisdictions (49%) actually experienced violent crime increases. Additionally, in the
following year the national violent crime rate had fallen 23 %. However, 40% of
jurisdictions in this year reported crime increases. This jurisdiction-level percentage
offers some insight as to the breadth of the change in crime during the 19908.
Nevertheless, the use of this statistic also results in a disadvantage in that each locality
receives equal weighting, regardless of population size. Consequently, since most
jurisdictions are small, crime trends among the relatively few large jurisdictions may be

obscured.

 

3 See Appendix A for national trends by crime type.
4 To eliminate agencies that reported inconsistently, only the 8.940 agencies that reported a full 12 months

54

A third approach to the assessment of crime trends nationwide provides a
compromise between analysis at the national level (via aggregated rates) and analysis of
crime within each jurisdiction. This approach, reﬂected in row three of Table 1, involves
the calculation of mean percentage changes within each jurisdiction, weighted in
proportion to the population represented. Thus, for example, the overall mean change in
the violent crime index for a city of size 50,000 would receive twice the emphasis as one
with only 25,000 inhabitants. The weighted mean percentage change statistic represents
the crime trends that were experienced by the average person across thousands of
jurisdictions. Utilizing this approach, Table 1 (row 3) indicates that the average
percentage change in violent crime across thousands of jurisdictions remained slightly
positive throughout most of the 19908. Nevertheless, a similar analysis of property crime
index trends (to be discussed later) conﬁrms that crime decreased substantially for most
property offense types across jurisdictions nationwide, more closely matching the

standard national index (see ﬁgure 23).

 

of data for at least 6 of the 12 years from 1990—2001 were included in analyses of crime trends.

55

.35. com: 05 o. 1.53 39285 82> a;

58.32 Eat Sun .8 E8» 0 .82 an totem»: :2: moﬁcowm bee moUEoE .

. 8.8 25.6 concave ooom Us“ 63— .32 E zocowm 0:0 +

 

... cote—smog 25.0 5>e memoeowm

 

 

 

 

 

 

 

 

 

 

 

 

N- m- T m+ 0+ w+ 2+ 5 _+ 2+ 3+ 2+ -- com Eooeoq omega SEE woﬁmﬁ? .m
8 8 8 8 % S s. a. a. a. S -- ... + .92 82 328 8853 285 .N
:3 :3 83 ASE p E a i Ga A3 at at are --
Yvom mdom mmm Whom ads mémo 0.33 QM :. wdvn mﬁmh <me «AS. 398:0 30233 mead 0635232 A
SON ooom amua— woog hQO— coa— m@o_ VOO— m©©~ N0©_ ~¢a~ ooo~ 30>

 

 

 

 

 

 

62829: as 222 82 328 $3.685

mﬂmoBeoﬁoE 005 he cemwanEeU - Soo— oﬁ wEBQ 31.2282 255 6205 A Bank

56

“Crime Trend” Defined

This review will employ all three of the previously described methodological
approaches to assess changes in crime patterns across the nation during the 19908.
Therefore, for purposes of this analysis, the term "trend" as it relates to standard
aggregated (summed) national rates refers to any period of three or more consecutive
years of changes in crime in the same direction (i.e. upward or downward). Similarly,
when weighted mean percentage changes for each jurisdiction are aggregated to a higher
level, the term "trend" reﬂects consecutive years of crime increases or decreases.

However, when counts of jurisdictions that result in percentages of agencies
above or below 1990 levels are utilized, the term "trend" requires that at least 60% of
agencies changed in the same direction (i.e. increased or decreased) for three consecutive
years. Thus, for example, if 53% of agencies with jurisdiction populations ranging from
size 50,000 to 100,000 reported crime decreases in a given year, this would not be
described as a downward trend (since nearly one-half of agencies in this category did not
report a decline). However, if 60% of agencies a size category reported crime rates that
were below 1990 levels in a given year, this could represent the beginning of a
“downward trend” -- assuming that an equal or greater number of agencies in this size
category (50,000 to 100,000 population) continued to report declines for the following
two years.

In Table 1, trends (as deﬁned above) are highlighted in yellow for each of the
three methodological approaches analyzing crime. These data clearly indicate that some
jurisdictions experienced increasingly large decreases in violent crime after 1994, ranging

from 3 to 31% below 1990 levels. However, in terms of the breadth of crime reduction,

57

over one-third of jurisdictions did not experience any signiﬁcant decline (row 2).
Moreover, with regard to the average change across jurisdictions (row 3), violent crime
rates remained above 1990 levels until the very end of the decade where mean crime
percentages declined by 1% to 2%. In the sections that follow, more detailed analyses of
individual crime types that include speciﬁc subgroups pertaining to jurisdiction size are
conducted to provide additional context with regard to crime trends during the 19908.
Percentages of Jurisdictions Below 1990 Crime Rates (by Size Category)
Figure 4 shows the violent crime index trends for agencies across the nation by
size category where the brown and green lines represent jurisdictions with over 100,000
citizens and the yellow, pink and dark blue lines represent jurisdictions with less than
50,000 citizens. Utilizing the aforementioned deﬁnition of "trend" which in this case
requires that 60% of each subgroup (size category) report violent crime at below 1990
levels for a period of at least three years, it is apparent that large agencies began their
downward trend around 1996, followed by jurisdictions with 50,000 to 100,000 citizens
approximately three years later (1999). Although the smallest three size categories never
ofﬁcially met the criteria for demonstrating a downward crime trend, a majority of
agencies in the smaller sized jurisdictions did report violent crime at below 1990 levels
toward the end of the decade. It also is important to note that the percentages among
each of the size categories are based on an assessment of the crime rate in each individual
agency -- compared to its own previous 1990 crime rate -- and not a group average.

Thus, these percentages represent true jurisdiction-level differences ﬁ'om 1990.

58

Figure 4. Violent Index Trends
(% agencies below 1990 levels)

 

 

 

 

 

 

 

 

 

 

 

 

 

  

 

— <10,000
— 10-25k
25-50k
—— 50-75k
— 75-100k
— 100-150k
>150,000
Figure 5. Property Index Trends
(% agencies below 1990 levels)
1.0 <10,000
0.9
0.8 — 10-25k
0.7 25-50k
0-6 ——50-75k
0.5
0.4 -— 75—100k
0.3 — 100-150k
0'2 i ' V ' " >1soooo
1991 1993 1995 1997 1999 2001 ’

 

 

 

 

 

 

Figure 5 represents the property crime index for the same agencies by size
category. As with the violent crime index, the downward trend began with agencies over
100,000 population, followed by jurisdictions of 50 to 100,000 population and ﬁnally
spreading to 60% of smaller agencies toward the end of the decade. In this regard, two

noteworthy aspects of the trend lines in Figure 5 are illustrated. First, the downward

59

trend in property offenses began approximately three years earlier than the trend for
violent offenses in agencies over 100,000 population. Second, the downward trend
progressed from large to small agencies in an almost categorical, step-by-step fashion
such that 60% of agencies with over 100,000 citizens reported property crime at below
1990 levels in 1992, followed by jurisdictions of size 75,000 approximately three years
later (1995), followed again by jurisdictions with 10 to 50,000 citizens in 1997 and
ﬁnally reaching the smallest jurisdictions around 1999. These step—by-step downward
trends by size category may indicate that theoretical causal mechanisms or disruptive
law-enforcement activities ﬁrst began among larger agencies and later spread to, or were
replicated by, smaller jurisdictions. Further discussion of these trends and their
implications for theoretical explanation of crime rate changes will be discussed in the
subsequent literature review pertaining to crime trends.

The ﬁgures below list the percentages of agencies below 1990 crime levels for
each of the individual crime types contained in the property and violent indexes. As with
Figures 4 and 5 above, these agency trends also reﬂect an assessment of the crime rate in
each individual agency -- as compared with its previously reported baseline rate in 1990.
Interestingly, although the combined violent crime index in Figure 4 suggests that most
small agencies did not experience an overall reduction in violent crime until much later,
analysis of murder rates individually (Figure 6) suggests that over 70% of smaller
jurisdictions (below 50,000 population) were already experiencing homicides at levels
below the baseline year in 1991, approximately four years prior to the beginning of the
downward trend in larger agencies. This may indicate that 1990 was a peak year with

regard to murder rates among smaller agencies and thus, does not serve well as a baseline

60

for this crime type. Similarly, more than 85% of agencies among all size categories
reported immediate declines in manslaughter rates (Figure 7) when compared with the
baseline year (1990). Even though 1990 may not provide an adequate baseline for
murder and manslaughter among smaller jurisdictions, these trend lines clearly indicate
that both murder and manslaughter declined among the vast majority of agencies during
the 19905.

Figure 6. Murder Trends
(% agencies below 1990 levels)

 

 

 

1.0
09,—..-
as
07-7
06/—\
05
04-.
03...,
a2 . . . . . . , .

1991 1996 1996 1997 1999 2001

 

 

 

 

 

 

 

 

 

61

Figure 7. Manslaughter Trends
(% agencies below 1990 levels)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1'0 ____. l '4' r—<10,000‘
0.9 4 ~. _._._.___.,
0.8 .. . _ “—1925"
0.7 2550K
0.6 9....— ‘ -.--.- ﬂ ——50-75k
0.5 ...— --_._-- —75-1(n<
0-4 —100150k
0.3 ~ >150,000
0.2 . . , . . . . . . .

1991 1990 1995 1997 1999 2001

 

 

 

Figure 8 (below) illustrates the trend lines for the offense of aggravated assault
and shows that decreases in aggravated assault began among larger agencies and were
followed by those with smaller populations. Similar trends are illustrated in Figure 10
with regard to robbery rates with one exception. The smallest size category indicated by
the dark blue line suggests that 1990 was the peak year for the majority of these small
agencies throughout the decade with regard to robbery.

It is also noteworthy that the trend lines among this “below 10,000 population”
size group are relatively ﬂat or unchanging for virtually all of the crime types. This may
indicate that causal mechanisms of crime among very small jurisdictions differ from
those with larger populations. Additionally, it may suggest that there are reporting
differences among these very small agencies (Maltz, 1999).

Crime trends pertaining to rape, as illustrated Figure 9, have been identiﬁed as
potentially problematic due to inconsistencies between federal, state, and local deﬁnitions

of this offense and technical reporting problems (Maltz, 1999). Nevertheless, Figure 9

62

suggests that reported rapes decreased throughout the 19908, primarily among
jurisdictions with greater than 100,000 citizens, but also extended to agencies with less
than 25,000 population. Furthermore, comparisons between UCR and nationally
representative victimization data (N CVS) have been conducted by Steffensmeier and
Harer (1999) for the years 1980 through 1998 and very similar trends were reported for
each of the crime categories, including rape (although there is some indication that
reporting percentages for rape have increased since 1992). Finally, the relatively ﬂat or
straight-line for the smaller size groups may indicate that 1990 is an inadequate baseline
year for this crime type among the smallest agencies.

Figure 8. Aggravated Assault Trends
(% agencies below 1990 levels)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1.0 i
0.9 ,. . ,_ . . _ _, —<10,000
08» , ~ — _, ——10-25k
°-7 -...________ 25-50k
0.6 y .. ___5075k
Sji/V' +_ 4 —75-100(
0.3L ,V . . —100-150k
0-2 I . . . . . . r r . . >150,0m
1991 1993 1995 1997 1999 2001

 

 

 

63

Figure 9. Rape Trends
(% agencies below 1990 levels)

 

 

 

  

 

 

 

 

 

 

 

 

1.0 .
0.9 — <‘°-°°°
0.8 — 1045K
0.7 25-5Ck
0.6 _ 50-75k
3:: —75-1(D<
Q3 — 1w150<
0.2 ~ — >150,(XX)
1%1 19% 1%5 1997 1999 2W1
Figure 10. Robbery Trends
(% agencies below 1990 levels)
—- <10.000
— 10-25t
25-50t
— 50-75k
—- 751M
— 1113-151X
— ”$.11”

 

 

 

 

19911Q319$19971999 2001

 

 

Figures 11 through 13 illustrate agency trends with regard to the property offenses
of burglary, larceny, and auto theft. The percentage of agencies reporting declining
levels of burglary begins early in the decade (around 1992), several years prior to this
trend with regard to larceny. Nevertheless, these trend lines clearly indicate that the vast
majority of agencies experienced decreases in burglary and larceny throughout most of
the 19908. In contrast, the trend lines for auto theft indicate little change in most agencies
from all size categories until around 1997, when the majority of agencies began to report
declines below 1990 levels. Consequently, unlike the offenses of burglary and larceny,
vehicle theft did not decline among the majority of agencies until the ﬁnal portion of the
decade. This trend may suggest that causal mechanisms behind auto theft differ

substantially from the offenses of burglary and larceny.

Figure 11. Burglary Trends
(% agencies below 1990 levels)

 

 

 

1.0 ——
0.8 —10-25( x
0.7 25-5Ck ‘
3°: —50-75k
0.4 —76100k
0'3 —100-150<
0.2 —>150,0001
199119901995199719992101—4

 

 

 

65

Figure 12. Larceny Trends
(% agencies below 1990 levels)

 

 

 

 

 

 

 

1991 1993 19% 1Q? 1999

 

Figure 13. Vehicle Theft Trends
(% agencies below 1990 levels)

 

 

 

 

 

 

 

 

1991 1993 19% 1997 1999

 

 

Mean Percentages of Crime Change in Jurisdictions (by Size Category)

Another method to assess the levels of crime reduction that occurred in
jurisdictions across the nation is to calculate the percentage change in each jurisdiction
compared with its reported 1990 crime rate. Figures 14 through 21 below illustrate the
trends in the percentages of change in each jurisdiction by size category. The broader red
line highlights the baseline or “zero mark” which represents the 1990 crime level. Figure
14 shows that the mean percentages of change among agencies across all size categories
with regard to murder rates began to decline substantially during the mid-19903. For
jurisdictions over 25,000 population, the decreases in the year 2000 ranged from
approximately 12% (100,000 to 150,000 population group) to over 35% among the
25,000 to 50,000 population group. As with previous trends, the trend lines among
smallest population categories remained relatively ﬂat but substantially below 1990
levels. Figure 15 indicates that manslaughter rates were substantially below 1990 levels
across the entirety of the decade.

Figure 14. Murder
(Mean % change by size group)

 

 

 

_ <1Q(m

— 1025k
2550<

—- $75k

43.5 \——\ —751ou<
N —1oo1su<

—>15qooo

 

 

 

 

 

 

-1

 

 

1%1 1% 1% 1%7 1%9

 

 

 

67

Figure 15. Manslaughter
(Mean % change by size group)

 

 

 

 

 

 

 

  

 

 

 

 

 

 

 

 

 

 

 

 

0(2) A _ <10,(ID
0.2 \ m / \ _ —1oza<
-O.6 ~ —50-75<
{18 r ‘ " —751ou<
-1 . 4):.
1 2 —1(D-15G<
' . l . l I 1 . l l I —>1m('m
1%1 1% 1% 197 1% ’
Figure 16. Rape
(Mean % change by size group)
— <1o,ooo
—— 10-25(
255m
— 50-75k
— 751W
— 1CD-150<
—- >150,ooo

 

1991 19% 1% 1%7 19%

 

 

 

68

 

Figure 17. Aggravated Assault
(Mean % change by size group)

 

 

_<1o,ooo
—1025k

 

— 50-75k
— 751M
—1w150<
— >15QCID

 

 

 

1%1 1993 1% 1%7 19%

 

 

 

Figure 16 indicates that the mean reported percentage change in rape rates among
jurisdictions between 10,000 and 75,000 population did not decrease during the 19905.
However among jurisdictions over 100,000 population reported decreases ranged from
15% to 20% by the year 2000. It is also noteworthy that aggravated assault rates
increased during the majority of the 1990s with a slight downward trend toward the end
of the decade and among agencies over 150,000 population (Figure 17). The exceptional
group with regard to aggravated assault are jurisdictions of size 100,000 to 150,000,
which reported declines beginning in 1996 that reached 20% by the year 2000.

With regard to robbery rates, Figure 18 shows that declines among jurisdictions
over 100,000 population began around 1997 and continued through the year 2000 to an
approximate reduction of 20%. Only one other size category (75,000 to 100,000)
reported any robbery reductions during the 19905 and this reduction took place at the

very end of the decade with a slight decline of approximately 4% below 1990 levels.

69

Figure 18. Robbery
(Mean % change by size group)

 

 

—<1o,ooo
A? _ —10-25k

— 50-75k
—75-1(I]<
—1(D-15G(

1991 1993 1995 1997 1999 —>150.00°

 

 

 

 

 

 

Figure 19. Burglary
(Mean % change by size group)

 

1 _ <1QCXI)
. — 10-25(

. _ 5075K
—751oo<
I _ 100150<

 

_>150,r_m
191 1%3 1% 1%7 1999

 

 

 

 

 

With regard to the offense of burglary, Figure 19 shows sharp declines across
virtually every jurisdiction size, beginning around 1992 and continuing through the end
of the decade. By the year 2000, declines in reported burglary rates ranged from
approximately 23% to over 40% when compared with 1990 levels. The trend lines for

larceny rates in Figure 20 indicate a similar but less extreme trend, with reductions

70

occurring somewhat later (after 1995), and ending in 2000. Crime declines for larceny
among the largest jurisdictions in 2000 were slightly over 20%. However among smaller
jurisdictions, declines were more modest, ranging from approximately 5% to 15%.

Figure 20. Larceny
(Mean % change by size group)

 

— <1o,ooo
— 1025K
2560<
—— 5075k
— 751cm
— 100150<
— >150,ooo

 

 

1%1 1%3 1% 1997 1%9

 

 

 

 

Figure 21. Vehicle Theﬁ
(Mean % change by size group)

 

 

—<10,(X'D
—10-25<

—5075k
—75-1(n<
—10015(]<
—>150,000

 

 

 

1991 1993 19% 1997 1999

 

 

 

 

71

The ﬁnal property offense category, auto theft (Figure 21), did not decline below
1990 levels in any of the size categories until approximately 1999, where declines ranged
from 2% among jurisdictions of size 25,000 to 50,000 to approximately 11% among
jurisdictions over 150,000. Interestingly, the 100,000 to 150,000 population size
category never reported vehicle theft rates below 1990 levels throughout the entirety of
the decade. The reason for this atypical pattern among this larger agencies group is
unknown.

Overall Weighted Mean Percentages Nationwide Versus National Summed Rates

The following four ﬁgures (22-24) facilitate comparisons of the trend lines
pertaining to national summed rates with those from weighted mean percentages of
change at nationwide. Figures 22 and 23 indicate the overall weighted mean percentage
changes (perhaps the best single overall measure of trends) from 1990 levels for all
jurisdictions with over 10,000 population. As previously discussed, these mean
percentage changes are weighted by population to provide an indication of the
experiences of the average person across thousands of jurisdictions throughout the 19905.
Although the trend line for the weighted mean of the violent index (Figure 22) resembles
that of the summative national rate shown in Figure 24 with regard to overall shape, the
year in which the trend lines indicates crime decreases below 1990 levels differs
substantially. The summative national rate trend line (Figure 24) drops below the 1990
level around 1994 while the weighted mean percentage change trend line (Figure 22)
does not drop below the 1990 level until approximately 1999. One possible explanation
for this is that the summed national rate more closely reﬂects activities in the largest
jurisdictions while the weighted mean change is more representative of jurisdictions

across the nation.

72

Figure 22. Violent Change Nationwide
(Weighted mean % change)

 

 

I 02
(115+ /\/\

01

0.05 \

,_
0C5

 

 

 

 

 

1

191 1% 1% 1%? 1% 2(D1

 

 

 

Figure 24. National Violent Rate
(Summed offenses/Population)

 

 

 

asses

 

 

l T l l

1990 19% 1994 1% 1% 2111)

 

 

 

Another key difference between the summative national rate and the weighted
mean percentage change is reﬂected in the amount of relative difference between the
levels in the year 2000 as compared with 1990. The national rate as reﬂected in Figure
24 indicates a total nationwide decline of approximately 30% in violent crime. However

3

the mean percentage change shown by Figure 22 indicates an overall decline of only

73

about 2% nationwide. These differences suggest that the bulk of the decline in violent

crime occurred within a relative few large jurisdictions.

Figure 23. Property Change Nationwide
(Weighted mean % change)

 

 

191 1% 1% 1%7 1% 201

 

 

 

Figure 25. National Property Rate
(Summed offenses/Population)

 

 

5500.00
5311(1)“ m\ ’ ' , , ,.
4,5110%»—

4,000.00» , r , \

3500.00
3,CID.CD

 

 

 

 

l I ‘7

1%19Q19941%19% 2000

 

 

 

In contrast to the differences in methodologies with regard to violent crime,

Figures 23 and 25 are strikingly similar. Both the summative national rates and mean

74

percentage changes indicate steep and continuous declines in property crime, beginning
around 1992. Further, with regard to the amount of change between the year 2000 and
the baseline year (1990), both estimation methods result in similar amounts of change.
The national rate of decline illustrated in Figure 25 indicates a 29% decline in property
crime while the mean percentage change reﬂected in Figure 23 indicates a 26% decline
ﬁ'om 1990 levels. The similarity of these approaches with regard to property crime
suggests that declines in these offenses were not limited to large jurisdictions but rather
occurred in the majority of localities across the nation.

National Weighted Mean Percentage Changes by Crime Type

The national weighted mean percentage changes for each type of offense during
the 19908 are illustrated in Figures 26-33 (below). This method of analyzing crime
change indicates that overall nationwide declines in crime were moderate for robbery
(-1 1%), and vehicle theft (-13%) and strong for murder (-32%), manslaughter (-51%),
burglary (-3 7%) and larceny (-19%). Overall rates of aggravated assault remained above
1990 levels as indicated by Figure 29 as did the offense of rape (Figure 28). It was also
noted that nationwide declines in murder, robbery and auto theft occurred primarily after
1995, whereas declines in burglary, larceny and manslaughter occurred in the early
19905.

These weighted mean percentage changes may be compared with the standard
national summative rates, illustrated in appendix A. It is noteworthy, however, that
standard FBI procedure using summative rates reﬂects combined murder and
manslaughter rates. Thus, the downward trend for these offenses begins around 1993,

halfway between the actual downward trend for murder and manslaughter individually.

75

Additionally, the downward trends for robbery and auto theft appear much earlier in the

summative rate approach, indicating that these trends began in larger agencies.

Figure 26. Murder Change Nationwide
(Weighted mean % change)

 

 

 

 

 

 

 

 

 

 

 

 

-O4 . I . r . . . -
1am 1&8 1&51 1&W 1&9 ZEN
Figure 27. Manslaughter
(Weighted mean % change)
01

 

0,,H-,W-,rr-rm,wm-,r.M-tmmr
o1» \ .
02

XMVA * *
-o.5~r r r \

-0-6 . I i . l I l I l
1&N 1&B 1&5 1&W 1&9 ZD1

 

 

 

 

 

 

 

76

Figure 28. Rape Change Nationwide
(Weighted mean % change)

 

 

0.25
0.2

0.15 r" /\ /\
0.1 v 7
0.05 W 77777 \/\/ .-

1991 1993 19% 1997 1999 2031

 

 

 

 

 

 

 

Figure 29. Aggravated Assault Change
(Weighted mean % change)

 

 

0.3
0.25 ~>
0.2 -- *-
0.15
0.1
0.05

Oi.
199119931995199719992001

 

 

 

 

 

 

 

 

 

77

 

 

Figure 30. Robbery Change
(Weighted mean % change)

 

 

 

1&1

1Q31%1Q71% 2131

 

Figure 31. Burglary
(Weighted mean % change)

 

 

o4

 

 

 

 

 

 

 

1%1

 

T f T I l T

1%3 1% 1M 1%

 

78

 

 

Figure 32. Larceny
(Weighted mean % change)

 

 

0.%
O
-O.%
-0.1
-O. 15
-O.2
-0.25

 

1%1 1% 1% 1997 1%

 

Figure 33. Auto Theft
(Weighted mean % change)

 

 

O. 1
0.%

-0.%
0.1
-0. 15

 

1991 1993 1995 1997 1999

 

79

 

 

Descriptive Crime Trends - Summary of Observations

Below is a summary list of observations, based on descriptive analyses of the
UCR crime data that were detailed in the previous section.
During the decade of the 19903:
1. There was an overall nationwide decline in every crime category except rape and
aggravated assault.
2. Weighted mean percentage changes indicate that these overall declines in crime were
moderate for robbery (~11%), and vehicle theft (-13%) and strong for murder (-32%),
manslaughter (-51%), burglary (~37%) and larceny (-19%).
3. In terms of the breadth or pervasiveness of crime declines, murder, manslaughter,
burglary, and larceny decreases were the most widespread. Declines in robbery rates
occurred primarily in larger jurisdictions.
4. Both violent and property offense indexes indicate that crime did not decline in all
places during the same time period. Rather, declines generally began with larger
jurisdictions (over 100,000 population), followed by jurisdictions of medium-size and
eventually reaching the smallest jurisdictions (below 25,000 population).
5. Weighted mean percentage changes indicate that nationwide declines in murder,
robbery and auto theft occurred primarily after 1995. In contrast declines in burglary,
larceny and manslaughter occurred during the early 19903.
6. Subgroup analyses by jurisdiction size categories indicate that the amount of declines
in crime were not equivalent among size categories, even when percentages of change
within each jurisdiction were assessed. For example, among jurisdictions with over

100,000 population, declines in burglary rates were over 40% in the year 2000.

80

However, among jurisdictions below 50,000 population, burglary declines in the same
year ranged from 12% to 15%.
7. The trend lines among jurisdictions below 10,000 population began in 1990 an ended
in the year 2000 at approximately the same level for virtually every offense category with
the exception of aggravated assault (which increased). As a result, these descriptive
analyses suggest that agencies below 10,000 population frequently differ from larger
jurisdictions with regard to causal mechanisms of crime and/or reporting behaviors.

Descriptive Analyses of Control Variables - Systemic Inﬂuences on Crime

A number of researchers have suggested that broad systemic trends such as
employment rates, per capita income, proportion of the population that is aged 15 to 24,
or proportion of the population that is nonwhite (a proxy for disadvantage) are among the
primary driving forces behind changes in crime rates across the nation (Allen, 1996;
Blumstein, Steinberg, Bell, & Berger, 2002;. Fox, 1996; Steffensmeier & Harer, 1999).
Additionally, other researchers have pointed to changes in the number of police officers
per capita as potentially inﬂuential (Levitt, 2002; Marvel & Moody, 1996).
Nevertheless, before the theoretical and empirical rationales behind these

relationships are discussed, annual data pertaining to each of the aforementioned
variables will be presented in an effort to establish the facts regarding what took place
during the 19905. A thorough discussion of data sources pertaining to population
demographics, unemployment, and per capita income may be found in the methodology
section. However, a brief introduction to these data sources is listed in the following
paragraphs to aid in comprehension of the descriptive analyses of changes in these

systemic variables nationwide.

8]

Annual county-level data pertaining to population demographics (i.e. proportion
aged 15 to 24, and proportion nonwhite) are presented below and originate from annual
intercensal estimates that have been revised since the 2000 census. Data pertaining to
annual county level changes in employment and per capita income originate from the
Bureau of Economic Analysis. Finally, data pertaining to per capita police ofﬁcers are
presented annually at the jurisdiction-level and originate from the Police Employee
Master ﬁle that is maintained by the Federal Bureau of Investigation as part of the
Uniform Crime Reporting (UCR) system. With the exception of per capita income, all
ﬁgures below represent mean percentages of annual change from 1990 levels to facilitate
relative comparison of trends among the different variables. Annual changes in per
capita income are expressed in US. dollar amounts.

Figure 33. Population Aged 15-24

 

 

 

 

 

 

 

 

 

(Mean % Change)

0 —— <10,(ID
0.02 _ 1023‘
0'04 255m
-0.(B __5075k
“-03 ‘ " r — 751cm

411 _1oo150<
N “.- <c '\ 0. N

 

 

 

Figure 33 illustrates the trends in mean percentage changes with regard to the
proportion of the population by jurisdictions size that that was aged 15 to 24 and shows a

distinct, non-linear trend with decreases beginning in 1991 ranging from 2% to over 8%

by 1996. However, in 1997 the proportion of citizens aged 15 to 24 began to increase to
levels that approximately equal those in 1991 or 1992.

Figure 34. Percent Employed
(Mean % Change)

 

—<10,(XX)
—10-25(

— 50-75k
— 75100k
— 100-150k
— >150.000

 

 

 

 

 

 

 

Figure 34 shows the trends for proportion of the population that was employed
during the 19905. This ﬁgure indicates a decrease of approximately 2% during the ﬁrst
three years of the decade followed by a steady increase in employment, peaking in the
year 2000 at approximately 6 percentage points above 1990 levels.

Figure 35. Proportion Nonwhite
(Mean % Change)

 

 

— <10,000
— 1025K

— 50-75k

— 751W

— 1m150<
Q

,99 $09 — >150,000

 

 

 

 

 

 

 

83

Figure 35 shows the trends in the proportion of population that was nonwhite (a
proxy for disadvantage). This ﬁgure shows a steady increase in nonwhite population
throughout most of the 19905 with a slightly sharper rate of increase after 1999. Most
jurisdictions experienced increases in nonwhite populations of about 30% by the year
2000. For example, a county with 10% nonwhite population in 1990, under this rate of
increase, could be expected to report a 13% nonwhite population in the year 2000.
Additionally, Figure 36 illustrates the trend in per capita income and shows a steady
increase throughout the 19905.

Figure 36. Per Capita Income
(Mean % Change)

 

 

— <10,(ID
—10-25k

—— 50-75k
—75-100<
—100-15(]<
— >150,000

 

 

 

1991 1993 1995 1997 1999 Z01

 

 

 

 

84

Figure 37. Per Capita Ofﬁcers
(Mean % Change) - Jurisdiction Level

 

 

 

 

 

 

 

 

 

 

 

 

0.3 _ (10,000
025 " —10-25k
2....
° ——50-75k
0.1
0.05, —75-10(]<
o I ' l 1 1 T I I I _1m-1ﬂ
1991 1993 1995 1997 1999 200 —>150.000

 

 

 

 

Figure 37 (above) illustrates trends in ofﬁcers per citizen within each jurisdiction
by size category annually throughout the 19905. These data indicate substantial increases
in officers per capita, beginning around 1994 for jurisdictions of every size. The mean
percentage of increase ranged from approximately 5% in 1994 to around 13% for most
jurisdictions by the end of the decade. However, jurisdictions with populations ranging
from 100,000 to 150,000 reported sharper increases in per capita ofﬁcers ranging from
15% in 1991 to over approximately 27% in 1997. This sharp increase in ofﬁcers was
followed by a steep decline or "drop-off" in per capita ofﬁcers such that the per capita
rate in the year 2000 remained approximately 10% above 1990 levels.

Systemic Changes and Crime Trends - A Review of Literature
Youth Population and Crime Trends
A number of criminal justice researchers have suggested that crime trends
increase as the proportion of young persons (e. g. ages 15-24) in the population increases

(Blumstein et al., 2002;. Fox, 1996; Grogger, 2000). From a theoretical perspective,

85

criminological work that focuses primarily on the individual level of analysis has been
rather contentious with regard to the age-crime relationship. The developmental
perspective has argued that social ties to family, school and law abiding peers are
sometimes weakened among teens (Sampson & Laub, 1993). Consequently,
developmental criminologists have asserted that the decreased inﬂuence of these
mechanisms of social control during youth often leads to delinquency. Additionally,
Warr (1998) studied the age-crime curve at the individual level and provided data
suggesting that mechanisms of social control, such as marriage and full-time employment
may cause delinquent behavior to desist. However, advocates of the self-control
criminological perspective have argued that any apparent age-crime relationship at the
individual level is spurious and results largely from differences in opportunity structure
(Gottfredson & Hirschi, 1990).

Regardless of the underlying causes for delinquent behavior among youth, at
aggregate levels of analysis, studies often begin discussions of the age-crime relationship
by citing facts pertaining to higher overall arrest or self-reported delinquency rates among
youth (Blumstein, 2000; Fox, 2000; Grogger, 2000). The empirical evidence concerning
the relationship between age and crime has been generally supportive, but eﬁ'ect sizes
vary in strength depending on the type of offense and sometimes differ across time eras
or youth cohorts. Blumstein (2000) provided descriptive evidence that the arrest rate for
15-year-olds in 1993 was triple the rate in 1985. Additionally, Fox (2000), a proponent
of demographic explanations of crime, noted a steep increase in homicide offence rates
among persons in the 18 to 24 age category during the early 19903. However, after

analyzing descriptive trends at the national level, Fox concluded that “No one can

86

reasonably assert that all--or even most--of the sharp decline in homicide since 1991
could be linked to demographic change” (Fox, 2000, p. 307).

With regard to the amount of impact that changes in the youth proportion have on
crime, estimates have varied substantially. Cohen and Land (1897) analyzed national
time series data from 1946 to 1984 and reported that decreases in youth population could
account for approximately 58% of declines in homicide and 26% of declines in auto theft.
Additionally, Steffensmeier and Harer (1999) applied age standardization methods to
UCR data in an effort to assess the effects of age composition on index crimes from 1980
to 1996 and estimated that youth population changes resulted in 10% to 15% of changes
in property crime and 3% to 8% of violent offenses in various years during this period.
Finally, Raphael & Winter-Ebmer (2001) conducted a sophisticated econometric analysis
at the state level of analysis from 1971 to 1997 and found that the proportion of juveniles
aged 15-17 had the highest overall impact on crime rates. Various model speciﬁcations
in this analysis ranged from 8% to 14% for property and 5% to 10% for violent crime.

Some researchers have suggested that the impact of youth on crime may vary
during different time periods, depending on individual offense rates and other social
conditions. For example, Steffensmeier and Harer (1999) reported stronger age effects
on crime rates during the 19805 when compared with the early to mid-19905.
Additionally, Fox (2000) and Blumstein and Rosenfeld (1998) have suggested that
individual rates of youth offending increased during the early 19905 as a consequence of
the rapid increase in incarceration for drug dealing during the 1980’s, which may have
resulted in a vacuum where younger dealers replaced adults and were more likely to use

ﬁrearms to protect drug territory.

87

Nevertheless, Bernard (1999) noted that there has been a persistent and erroneous
popular belief throughout recent decades that youth offending is “on the rise” and that the
behavior of teens is no longer like the "good old days" of 30 or 40 years ago (p. 338).

His analysis of UCR arrest and victimization data indicated that, with the exception of
homicide, the general trend among juvenile offending rates has been a decrease of
approximately one third overall with regard to frequency. Furthermore, Levitt (1999)
argued that changes in age structure of the population have only a limited impact on
aggregate crime rates. His analyses of demographic and crime data indicated that even
the very large baby-boom age-trend shifted crime rates by no more than 1% per year.
Unemployment and Crime Trends

Most theoretical arguments asserting a relationship between unemployment and
crime are based on a rational choice perspective that assumes that individuals seek to
obtain resources according to relative costs and beneﬁts (Allen, 1996). However,
Raphael and Winter-Ebmer (2001) note that, although the relationship between
unemployment and crime seems intuitive, research to date has been unable to document a
strong positive effect. They further note that studies of aggregate crime rates have
generally found small and statistically weak unemployment effects with stronger effects
for property crime than violent crime.

Several studies have also found signiﬁcant negative effects of unemployment on
violent crime rates. Cohen, Kluegel and Land (1981) argued that this inverse relationship
may occur during periods of high unemployment because there are more individuals at
home to serve as guardians. Additionally, some researchers have argued that

unemployment may reduce the wealth of crime targets, thus, reducing criminal

88

opportunity (Cohen, F elson & Land,1980). Consequently, Allen (1996) noted that this
debate is not settled because both perspectives have received some measure of empirical
support.

A number of recent studies have found support for a relationship between
unemployment and speciﬁc types of offenses. Monthly time series data in New York
City from 1974—1999 indicated that a 1% decrease in unemployment was associated with
a 16% decrease in burglary (Corman & Mocan, 2002). Additionally, a study of national
level data from 1958 to 1995 indicated signiﬁcant positive effects associated with
property offenses (Ralston, 1999). For the middle year of this analysis (1976), the
reported regression coefﬁcients translated into a 10% to 14% overall decline in property
offenses among various speciﬁcations. Finally, a state level analysis of data from 1971 to
1997 found that a 1% decrease in unemployment was associated with a 2% decrease in
burglary and a 1.5% decrease in larceny (Raphael & Winter-Ebmer, 2001). However, the
estimated effects on violent crime in that study were inconclusive.

Interrelationships between Race, Youth, Wages and Unemployment

Embedded in many theoretical explanations concerning changes in crime,
unemployment, and youth population are implicit and explicit arguments about
interrelationships among racial demographics, youth population, poverty and wages. For
example, Ralston (1999) noted that prior research suggested that whites might be
inﬂuenced to a greater degree by changes in unemployment rates than their black
counterparts due, in part, to differences in life expectations. Consequently, Ralston
included variables to estimate differences among whites and blacks in his models of

unemployment.

89

Measures of poverty, inequality, inﬂation and per-capita income have also been
included in a number of empirical analyses pertaining to unemployment and demographic
changes. In this regard, Imrohoroglu, Merlo & Rupert (2004) found that income
inequality was among the top two determinants of crime rates, explaining as much as
59% of overall increases. Further, Grogger (2000) argued that per-capita income among
youth may be inversely associated with violent offenses that result from increased illegal
drug trade activity which becomes more proﬁtable when wages are low. Finally, it is
noteworthy that race has also been used as a proxy for disadvantage when measures of
poverty and inequality were insufficient (Zhao, Scheider & Thurman, 2002).

Few, if any studies have included all of these variables together and consequently
little is known about how they interrelate. It is noteworthy, for example, that Allen
(1996) did not ﬁnd a signiﬁcant youth effect in his analyses of more than 30 years of
crime data when inﬂation, poverty, and unemployment were entered into the model.
Additionally, lmrohoroglu et a1. (2004) found that changes in unemployment did not
explain changes in crime from 1980-1996 when measures of incarceration, arrest
probability, and income inequality were included in models. These ﬁnding highlight our
ignorance with regard to interactions and interrelationships among race, poverty, wages,
and unemployment. At present, it is unclear whether these variables are capturing
distinct elements of changes in crime rates. Conversely, they may be measuring

essentially the same variation via different mechanisms.

90

Systemic Changes and Crime Trends - A Look at the Data
Youth Demoggaphics Data (1990-2001)

Based on historical data that are now available throughout the entirety of the
19905, revised annual census data (shown in ﬁgure 33) indicate a steady decline in the
proportion of young persons aged 15 to 24 during the early 19905, reaching its lowest
point in 1995. It is noteworthy, however, that violent crime index data at the jurisdiction
level indicate that the majority of agencies between 1991 and 1994 experienced crime
increases during these years. Thus, age demographics alone do not appear to coincide
with declines in crime that occurred in large, then medium, and ﬁnally small
jurisdictions. Further, a substantial 40% or greater portion of jurisdictions reported no
overall decrease in violent crime through 1998 when youth were near their lowest levels.

The decline in proportion of youth did correlate with decreases in property
offenses during the ﬁrst ﬁve years of the 19905. However, after 1995 when the
proportion of young persons again began to rise, property offenses continued to decline
nationwide. Thus, the relationship between age demographics and crime at the county or
jurisdiction level is unclear and remains to be empirically tested.

Unemploment, Income and Racial Demoggaphics Data (1990-2001)

In reviewing the descriptive data pertaining to unemployment and per-capita
income trends, Figures 34 and 36 seem to mirror the patterns property crime, consistent
with prior theoretical and empirical ﬁndings. Simply put, as unemployment and per-
capita income increased, property crime trended downward. However, the relationship

between unemployment and violent crime seems less clear with the exception of murder,

91

which showed a brief increase that coincided with a brief decline in employment around
1992.

In contrast, the trend of nonwhite population (Figure 35) increased throughout the
19905 which, if taken as a measure of disadvantage, should have led to increases in crime
based on prior research. However, with regard to general overall trends for both property
and violent crime, the opposite was true in that both indexes suggest declines from 1990
levels. In this regard, the problem may be related to the large nationwide immigration of
Latinos. Prior research suggests that the occurrence of serious crime and delinquency
among Latinos who recently emigrated, is lower than the population of Latinos that have
lived in the United States for several generations (Sommers, Fagan, & Baskin, 1994).
Consequently, in order to obtain a better measure of the speciﬁc type of disadvantage that
is most likely to lead to crime, the variable pertaining to the percentage of the population
that is nonwhite could be modeled as an interaction with unemployment and/or per-capita
income.

Officers Per Capita - Data (1990-2001)

Finally, with regard to the number of ofﬁcers per capita (Figure 37), descriptive
analyses of trends are inconclusive. The rate of ofﬁcers did increase during the mid-
19905, approximately one or two years before moderately sharp declines in violent
offenses began in larger agencies. However, this officer trend does not appear to
correspond with declines in property crime. Additionally, the step by step downward
trend that migrated from large to smaller agencies with regard to declines in property and

violent crime indexes does not directly correspond with patterns of increases in ofﬁcers.

92

CHAPTER 4
METHODOLOGY
Study Objectives and Hypotheses

A primary objective of this study is to assess the impact of changes in policing
practices on crime during the 19905 to determine the extent to which these efforts
contributed to reported declines in murder, robbery, aggravated assault, burglary, larceny
and auto theft within jurisdictions across the United States. Although a number of
criminal justice researchers have asserted that changes in policing might have contributed
to the nationwide decline (Eck & Maguire, 2000; Lafree, 1998b; Ouimet, 2002; Sherman
et al., 1998), at present, little is known about which speciﬁc changes in policing tactics,
organizational strategies and technologies were most inﬂuential. Additionally, no study
has yet been published that assesses the relative impacts of these changes in policing
using a methodology that allows generalization of ﬁndings to jurisdictions across the
nation. This study attempts to provide information that will be useful to researchers and
practitioners by examining the impact of speciﬁc types of tactics on crime across a
variety of jurisdictional environments and organizational contexts.

The preceding review of literature pertaining to recent changes in policing
suggests that number of ofﬁcers per capita, arrests for minor offenses, problem solving
and other tactics as well as crime analysis may have contributed to crime reductions
across the nation. Thus, the following hypotheses were tested pertaining to rates of
reported murder, robbery, aggravated assault, burglary, larceny and auto theft nationwide

during the 1990s.5

”II-....

S More detail pertaining to the rationale for the hypotheses can be found in a subsequent section of the
methodology.

93

Hypothesis 1:

Hypothesis 2:

Hypothesis 3:

Hypothesis 4:

Increases in the number of officers per capita contributed to
reductions in crime.

Increases in arrests for minor drug and disorder offenses
contributed to reductions in crime.

Increases in the implementation of the broad categories of policing
tactics and strategies tactics (listed below) contributed to
reductions in crime.

Collaboration Eﬂorts, Problem Solving

Citizen Involvement, Code Enforcement

Community Policing Policies, Crime Analysis

Place Oriented Strategies, Eﬂiciency Related Strategies

Increases in the implementation of speciﬁc tactics and strategies

(within the broad classiﬁcations listed in hypothesis 3) contributed
to reductions in crime.

Methodological Overview6

In order to assess the hypotheses, annual changes in crime and policing that were

reported during the years 1990-2001 were included in regressions using a panel data
approach with ﬁxed-effects for both jurisdictions (i.e. places) and year. The
implementation of ﬁxed effects panel data models allow the researcher to assess changes
within each jurisdiction over time (Allison, 2005; Frees, 2004) and this methodology has
recently been utilized by criminal justice researchers (Marvel and Moody, 1996; Zhao,
Scheider and Thurman, 2002). This approach offers a substantial beneﬁt over cross-
sectional research in that omitted variables that are relatively stable over time are no
longer problematic (Benson & Rassmussen, 1998; Frees, 2004). Stated another way, by
comparing each jurisdiction with itself at multiple time periods, stable characteristics that

are unique to each jurisdiction can no longer inﬂuence the model. For example, two

 

6 Images in this dissertation are presented in color.

94

cities of equal size may be very different with regard to a number of characteristics. City
“A” may be located in an industrial or factory based economy and city “B” may be
located in a service based economy. In traditional cross-sectional analysis, these
differences would be problematic unless a variable pertaining to “economy-type” were
included. However, by using ﬁxed place effects, a dummy variable is entered for each
speciﬁc jurisdiction, resulting in a differencing of all modeled values ﬁom time-l to
time-2 (Benson & Rassmussen, 1998). This process is illustrated in Figure 38, where the
time-l values are subtracted from time-2 values such that stable city characteristics are
eliminated from analyses.

Figure 38. Illustration of the “Canceling Effects” of Fixed Effects Regression.

 

City “A” 1991 Crime91 = ofﬁcer_rate9l + unemployment91 + fM economy91

City “A” (subtract) 1990 Crime90 = ofﬁcer_rate90 + unemployment90 + fMeconomWO

 

In the case of “City A” the stable characteristic of factory based economy cancels out of
the equation because the values in 1990 are equal to the values in 1991. This approach
removes the necessity to include variables for other relatively stable jurisdictional
characteristics such as region, jurisdiction size and type of government (i.e. municipal or
county). Similarly, trends that effect all agencies in a given year or era can be controlled
for by including a year ﬁxed effect or variable in the model (Allison, 2005). As a result,
only variables that change -- and correlate with independent variables in the model may

be inﬂuential and need to be entered into the analysis (Frees, 2004).

95

Population of Inference and Database Overview

The population of inference for this study is comprised of all ofﬁcially reported
crime at the national level during the years 1990-2001. Two primary databases (detailed
below) were utilized to assess changes and trends in crime during the 19905. First, arrest
and ofﬁcer count data, maintained by the Federal Bureau of investigation, were combined
with crime data that include over 89% of ofﬁcially reported felony offenses. This
database allowed an assessment of the relationships between the number of ofﬁcers per
capita, as well as crime rates and minor misdemeanor arrests.

A second, a database (detailed below) was created by combining data from a
national policing survey with the FBI crime data and includes information pertaining to
nearly one-half (49%) of ofﬁcially reported crime during the years 1993 and 1997. This
policing survey database allowed the assessment of relationships between speciﬁc
policing practices, such as foot patrol, problem solving and crime analysis with crime
reduction that occurred from 1993 to 1997.

Primary Data Sources
Changes in Crime

The Federal Bureau of Investigation (FBI) collects and maintains a database,
known as the Uniform Crime Report, which contains the number of reported murders,
aggravated assaults, robberies, rapes, arsons, burglaries, vehicle thefts and larcenies
within individual jurisdictions across the nation (FBI, 2002). All of these crimes are of a
serious nature and are classiﬁed as felonies for which the penalty may include over one
year of incarceration. The Uniform Crime Reports have been collected since the 1930’s

and, although reporting is voluntary, the vast majority of full service (24 hour) law

96

enforcement agencies reported at least partial crime data during the (FBI, 2002). For
example, in 1990 17,608 of the estimated 18,778 law enforcement agencies were listed in
the UCR database.

State police and campus, airport and other special police agencies were removed
from analyses because their jurisdictions overlap with other law enforcement agencies
that have primary responsibility for policing these populations. The UCR program refers
to state and special police agencies as “zero-population agencies” (FBI, 2002). Thus,
calculation of crime, arrests and ofﬁcer rates per capita are problematic for these
jurisdictions. Additionally, following the methodologies of recent work by Zhao,
Schieder and Thurman (2002) and the US. Government Accountability Ofﬁce (2005a),
crime data that do not include a complete 12 months of crime data in a given year were
removed from regressions as well. The vast majority of these partial crime reports were
originated by small agencies with jurisdiction sizes of less than 10,000 citizens. Overall,
12,975 jurisdictions reported at least one complete year of data during the 1990-2001
period and were not zero-population agencies. Reports from these agencies represent
approximately 93% of all reported crime in the US. (GAO, 2005b).

Changes in Arrests and Ofﬁcer Rates

Variables pertaining to changes in arrests and ofﬁcers per capita are available
annually from 1990-2001 via two databases that are produced by the Federal Bureau of
Investigation. The number of sworn ofﬁcers per 10,000 citizens is contained in the
Police Employee Master Files. All of the agencies that reported at least one complete

year of crime data during the are also found in the Police Employee Master File.

97

Data pertaining to annual arrests in each jurisdiction for 37 types of felonies and
misdemeanors are contained the Arrest Master File. As with the UCR crime data,
reporting of arrests is voluntary and very small police agencies are less likely to
consistently report. However, the vast majority (over 89%) of the 12,975 agencies that
reported crime to the UCR also reported arrest data during the 1990-2001 time period.
Annual arrest reports that did not include a full 12 months of data were removed from
analyses. All data pertaining to changes in crime, arrests and the number of ofﬁcers were
obtained directly ﬁom the Federal Bureau of Investigation via the Criminal Justice
Information System (CJ IS) headquarters in Clarksburg, West Virginia.

Changes in Policing Practices

Two waves of a National Survey of Community Policing Strategies were
conducted during the 19905 to assess the extent to which various policing tactics and
strategies were being implemented by local police agencies (Rosenthal & Fridell, 2002).
The ﬁrst survey wave was administered by the Police Foundation in 1993 and the second
was implemented in 1997 via a partnership between Opinion Research Center (ORC)
Macro and the Police Executive Research Forum (Rosenthal & Fridell, 2002). Both
policing survey waves were designed to be nationally representative and consisted of
random samples of over 1,600 local police agencies, stratiﬁed by size category.

The 1993 and 1997 survey waves asked law enforcement ofﬁcials the same
questions about the implementation of speciﬁc policing tactics and strategies. More
speciﬁcally, 80 question items pertaining to problem-solving, community collaboration,
place or “hot spot” oriented tactics, crime analysis and other practices were inquired in

1993 and again in 1997.

98

Questionnaire items were coded in a binary fashion such that “0” (zero) indicated that the
activity was not performed and a “1” indicated that the activity was performed. For
example, an agency that began training citizens in problem solving in 1997 received a
“O” for this item in 1993 and a “1” in 1997. All 80 question items that were utilized in
this study are listed in Table 1 in appendix B.
Control Variables

Independent variables pertaining to changes in ofﬁcer rates, arrests and policing
practices were also combined with demographic and employment data to control for
contextual changes that occurred annually during the 19905 that may have effected crime
rates. County level employment rates and per capita income from the Bureau of
Economic Analysis were included as control variables. Additionally, the percentage of
the population that was aged 15 to 24 and percentage of the population that was non-
white (a proxy measure for disadvantage) were included as control variables from the
US. Census annual intercensal estimates. To correct for within county or infra-cluster
correlation, the cluster option was selected in regressions of ofﬁcer and arrest effects
using STATA software.
Linking Stratggy

Independent variables were linked to control variables via F IPS (Federal
Information Processing Standard) codes that indicate state, county and jurisdiction (or
place) location. Additionally, survey data from the national community policing
strategies study was linked to FBI Uniform Crime Report (U CR) data via agency ORI
(Originating Agency Identiﬁer) codes. Unique ORI codes are assigned to every police

agency that reports crime data to the Federal Bureau of Investigation. Data from the FBI

99

Police Employee Master Files and the Arrest Master Files were linked to police survey
records and UCR crime reports via ORI code as well.
Rationale for Hypotheses and Creation of Independent Variables

One of the primary objectives of this study is to identify relationships between
speciﬁc policing practices and speciﬁc types of crime. Thus, each of the seven primary
Index I offenses contained in the UCR (murder, robbery, aggravated assault, rape, vehicle
theft, burglary and larceny) were analyzed separately.7 Although it is fairly common in
criminal justice literature for offense types to be combined into broad indexes of property
and violent crime, the index approach is problematic in that it assumes that underlying
causal mechanisms are the same for different offenses. Each of the seven offense
categories were standardized into rates per 100,000 population for all analyses.

Crime rates and population size are closely related. For example, the rates of
robberies in Detroit or Los Angeles are typically higher than in smaller jurisdictions. As
a result, regression analyses of crime rates frequently include heteroskedastic error. That
is, regression estimates are more precise among jurisdictions with large populations and
are less precise for smaller jurisdictions. In the past, some econometric researchers have
performed logarithmic transformation of crime rates in an effort to reduce
heteroskedasticity (Harmon & Knapp, 2003). However, this approach has been strongly
criticized by criminologists who argue that these transformations may introduce bias by
limiting or “pinching off” the crime rates among the largest jurisdictions (Harmon &
Knapp, 2003). Hannon and Knapp (2003) have recommended an alternative approach

wherein regressions are weighted by population and standard errors are adjusted through

‘

7 Arson was excluded from analyses due to concerns regarding data reliability and because it is an offense
that often falls under the purview of ﬁre officials, rather than local police.

100

the implementation of a heteroskedasticity consistent covariance matrix (HCCM)
method. All panel data regressions in this study followed the Hannon an Knapp (2003)
approach. Regressions were weighted by jurisdiction population and robust standard
errors were calculated through an HCCM technique that is available through STATA and
SUDAAN software. Analyses of ofﬁcer and arrest data was performed using STATA
software. Analyses involving police survey data were conducted via SAS callable
SUDAAN software.
Increases in Ofﬁcers Per Capita - Hymthesis 1

Although the relationship between police force strength and jurisdiction crime
rates has been somewhat contentious, recent econometric research suggests that increases
in the number of ofﬁcers per capita are associated with reductions in crime (Levitt, 1997;
Marvel & Moody, 1996). From a theoretical standpoint, these crime reductions could
result from increases in arrests and incapacitation due to ofﬁcer availability. However,
crime reduction may also occur through a deterrent effect that emanates from increased
officer presence (N agin, 1998). As a result, it is hypothesized that increases in the
number of ofﬁcers per capita contributed to reductions in crime (hypothesis 1). To test
whether increases in officers contributed to reductions in crime during the 19905, a
variable containing ofﬁcer rates per 10,000 citizens was calculated from annual counts of
sworn personnel and reported jurisdiction population in the FBI Police Employee Master
data.
Increases in Enforcement of Minor Offenses - Hypothesis 2

One of the principles of the “Broken Windows” hypothesis, advanced by Wilson

and Kelling (1982) is that focusing on minor offenses may decrease citizen fear and

101

increase perceptions that the community expects individuals to behave in an appropriate
manner. Policing literature indicates that during the 19905 a number of large police
agencies were inﬂuenced by this perspective and implemented order maintenance or
“zero-tolerance” strategies (Green, 1999; Sherman, et al., 1998). New York City police,
for example targeted graffiti artists, public drunkenness and other disorderly activities
such as panhandling and public urination (Green, 1999; Kelling & Sousa, 2001).
Additionally, some empirical support has been highlighted pertaining to the crime
reducing effects of increased enforcement of minor offenses. After an in depth analysis
of New York’s order maintenance tactics during the 19905, Kelling and Sousa (2001)
concluded that this policing approach had a direct crime reduction effect. As a result, it is
hypothesized that increases in arrests for minor drug and disorder offenses contributed to
reductions in crime during the 19905 (hypothesis 2).

Two arrest indexes were created to test whether minor drug and disorder arrests
contributed to reductions in any of the seven felony offenses found in the UCR. The
Arrest Master Files list 37 types of felonies and misdemeanor offenses. Individual items
that were utilized to create both drug and disorder arrest indexes were selected only if
they were relatively common non-violent misdemeanor offenses for which ofﬁcers may
have some measure of discretion with regard to the aggressiveness of enforcement.
Offenses that are often classiﬁed as felonies (e. g. drug sale or trafficking) or that
frequently cause serious harm or loss to victims were eliminated. According to the
Broken Windows perspective, increasing focus on minor offenses may reduce crime.
Thus, a minor drug arrest index was created by summing arrests for drug abuse and drug

possession. Additionally an index of minor disorder offenses was created by summing

102

arrests for curfew violation, disorderly conduct, drunkenness, prostitution, runaways and
vagrancy. Annual arrest totals for each jurisdiction were then converted into rates per
1,000 citizens.

Figure 39. Composition of Summative Arrest Indexes

 

Minor Drug Arrest Index: (Drug abuse, Drug possession)

Minor Disorder Arrest Index: (Curfew, Disorderly conduct, Drunkenness,
Prostitution, Runaway, Vagrancy)

 

Arrest indexes are rates per 1,000 citizens

Policing Practices - Hypotheses 3 and 4

 

During the 19905 a number of trends in policing practices occurred that may have
impacted crime rates. A substantial quantity of policing literature, for example, has
documented the widespread adoption of community policing and problem solving
strategies (MacDonald, 2002; Scott, 2000). Although empirical assessments of the
impact of community policing and problem solving have yielded mixed results, most
prior analyses have involved case studies or regional subsamples. Nevertheless, from a
theoretical standpoint speciﬁc aspects of these strategies may have contributed to the
decline in crime as a result of increased police attention to community concerns,
increased information from citizens or the more permanent assignment of personnel to
speciﬁc geographic areas or problems. As a result, it is asserted that increases in the
implementation of community policing policies, citizen involvement and problem solving
may have contributed to reductions in crime.

A number of technological, efﬁciency and enforcement based changes also

occurred during the 19905. For example, the implementation of crime analysis increased,

103

in part due to the availability of personal computers (Mullen, 1996). Additionally, police
organizational efﬁciency improved via the implementation of cellular phones and crime
tip lines (Nunn & Quinet, 2002). Policing literature also suggests that increased attention
was focused on recurring crime places or “hot spots” as well as increased enforcement of
civil and regulatory codes during this era (Mazzerole & Rohl, 1998). Any of these
efﬁciency and enforcement policing trends during the 19905 may have impacted crime.
At present, the primary mechanisms through which changes in policing practice
may have impacted crime are unknown. It is possible, for example, that increases in
breadth or extent of implementation of problem solving reduced crime because more
problems were being identiﬁed and addressed by local police through a wider array of
tactics. However, it is also possible only certain types of problem solving activities were
effective and that the breadth of implementation of the problem oriented policing concept
was not of primary importance. Consequently, the methodology employed in this study
assessed the crime impact of both the breadth of implementation of broad categories of
changes through the creation of policing practice indexes as well as the impact of speciﬁc
policing tactics. Two additional hypotheses are therefore asserted. First, increases in the
implementation of broad categories of Collaboration Eﬂorts, Problem Solving, Citizen
Involvement, Code Enforcement, Community Policing Policy, Crime Analysis, Place
Oriented and Efficiency Oriented policing tactics and strategies tactics contributed to
reductions in crime (hypothesis 3). Second, increases in speciﬁc tactics and strategies

within the broad classiﬁcations listed in hypothesis 3 contributed to reductions in crime

(hypothesis 4).

104

Creation and Analyses of Policing Practice Indexes

Data from the National Survey of Community Policing Strategies contain
information pertaining to changes in the reported implementation of a broad array of
policing practices. More speciﬁcally, the survey questionnaires that were administered in
1993 and 1997 included 80 individual question items pertaining to efforts to collaborate
with citizens, problem solving, civil code enforcement, place or "hot spot" oriented
activities and other tactics and strategies. Each of the 80 variables were classiﬁed into an
index that corresponds to the breadth of reported implementation of speciﬁc types of
policing practicesg. For example, each item that mentioned "problem solving" was
classiﬁed into a problem solving index. Items that described the implementation of crime
analysis were placed into a crime analysis index and a similar process was followed with
regard to the remaining questionnaire items.

Eight primary indexes were created by summing together the items that were of a
similar nature. Items that described efforts to increase routine citizen contact or
feedback to the police were summed into a “Collaboration Effort” index. All items that
mentioned problem identiﬁcation or resolution were summed into a problem solving
index. Items that described citizen actions to assist the police that did not fall into
collaboration or problem solving were summed into a citizen involvement index.
Additionally, a “Civil Code Enforcement” index was created by summing items that
described efforts to utilize civil or regulatory codes to reduce crime. A “Community

Policing Policies” index was created by summing items that described new policies or

 

8 One item, v16, was removed because it did not ﬁt clearly into any single index.

105

legislation to support community policing efforts. A “Crime Analysis” index was created
by summing items that described efforts to identify crime patterns within the jurisdiction.
An index of “Place Oriented Tactics” was created by summing items that described
efforts to increase enforcement or management in geographic areas or speciﬁc types of
places. Finally, an “Efﬁciency Efforts” index was created by summing items that
described efforts to streamline organizational design or management. These individual
items and their classiﬁcation indexes can be found in Appendix B.

However, the community collaboration and problem solving indexes were further
sub-divided to differentiate items that referred to actions by line-level or patrol ofﬁcers
and their immediate supervisors (e. g. sergeants) from actions that were carried out by the
organization as a whole. Thus, community collaboration was divided into two sub-
indexes: community collaboration (line level) and community collaboration
(organization level). The same procedure was followed to create two problem solving
sub-indexes: problem solving (line level) and problem solving (organization level). The
rationale for these further subdivisions is detailed in the following paragraph.

The National Survey of Community Policing Strategies was designed to assess
the implementation of policing practices at multiple levels of the organization,
particularly with regard to collaboration efforts and problem solving (Rosenthal &
Fridell, 2002). Consequently, question items of a similar nature were asked in reference
to a speciﬁc level of the organization. For example, questionnaire item number 15 in
Table 1 (Appendix B) inquired as to whether the agm trains citizens in problem
identiﬁcation or resolution. However, a similar item (number 52 in Appendix B) inquires

as to whether officers teach residents how to address community problems. As a result, a

 

106

substantial portion of the 16 problem solving items were similar with regard to content
but pertain to either the police organization as a whole or line level officers and
supervisors.

At present, it is unknown whether problem solving and collaboration efforts at the
line or organization levels are most likely result to in crime reduction. Therefore, ad hoc
removal of any survey items was undesirable. Thus, to avoid duplication in the problem
solving and collaboration effort categories, survey questions were subdivided into two
groups as described above. This questionnaire levels and duplication issue was not
problematic for any of the other indexes such as crime analysis civil or code enforcement.
Therefore only single indexes were created for policing practices that did not fall into the
collaboration for problem solving categories.

Figure 40. Composition of Summative Indexes Pertaining to Policing Practices

 

Collaboration Effort: Items that describe efforts to increase routine citizen contact or
feedback to the police.

Organization-Level: v7,v1 1,v12,v13,v25,v26,v3 1,v33 ,v65,v74 (range 0-10)
Line-Level: v46,v48,v49,v51,v55,v58 (range 0-6)

Problem Solving: Items mentioning problem identiﬁcation or resolution

Organization-Level: v3, v14, v15, v22, v35, v45, v66 ,v79 (range 0-8)
Line Level: v47, v50, v52, v57, v60, v61, v62, v63 (range 0-8)

Citizen Involvement: Items that describe citizen actions to assist the police that do not
fall into collaboration or problem solving.

Index: v67, v68, v69, v70, v71, v72, v73, v75, v76, v77 (range 0-10)

Code Enforcement: Items that describe efforts to utilize civil or regulatory codes to
reduce crime.

Index: v19, v20, v37, v56 (range 0-4)

107

Community Policing Policies: Items that describe new policies or legislation to support
community policing efforts.

Index: v1, v2, v4, v5, v80 (range 0-5)
Crime Analysis: Items that describe efforts to identify recurring crime patterns.
Index: v21, v43, v44, v54, v64 (range 0-5)

Place Oriented Tactics: Items that describe efforts to increase enforcement or
management in geographic areas or speciﬁc types of places.

Index: v9, v10, v18, v28, v29, v30, v32, v38, v39, v40 (0-10)

Efficiency Efforts: Items that describe efforts to streamline organizational design or
management.

Index: v6, v8, v17, v23, v24, v27, v34, v36, v41, v42, v53, v59 (range 0—12)

 

Creation and Analyses of Individual Policing Practice Variables

Although analyses of changes in the number of problem solving, place oriented or
other groups of tactics may increase our understanding of relationships among broad
classiﬁcations of policing practices and crime, it is possible that only speciﬁc individual
policing practices may be effective. Therefore, regressions for each individual survey
item were conducted to aid in identifying speciﬁc policing practices that were associated
with crime reduction. Each of the 80 survey items, listed in Appendix B pertaining to
speciﬁc policing practices was coded such that the change in implementation from 1993
to 1997 ranged from -1 to +1. That is, an agency that reported implementing a policing
practice in 1993 and discontinued this practice by 1997 would be assigned a -1 for this
survey item. In contrast, an agency that began implementing a policing practice after
1993 received a score of +1. Finally, agencies where implementation of a speciﬁc

policing practice did not change received a score of zero. The ﬁxed effects panel data

108

regression models utilized in this study are designed to assess within subject change,
rather than cross-sectional differences in levels. Therefore, the net change within a
jurisdiction that continued to implement a policing practice from 1993 through 1997 is
zero.

Because there is substantial similarity with regard to subject matter among the 80
policing survey items, it was expected that reported implementation of many of these
individual policing practices would be correlated. For example, an item that inquired as
to whether police ofﬁcers train citizens in problem solving may be correlated with
practices pertaining to collaboration efforts by the agency. Consequently, to identify
items that explained a unique portion of the variation in crime, subsequent regressions of
individual policing practice items together (simultaneously) in a single equation were
performed.

Individual policing practices that demonstrated statistical signiﬁcance with regard
to crime reduction were simultaneously included in a regression to determine which items
were most strongly or closely associated with crime reduction. For example, if policing
survey items, v1, v2, v3 and v4 were signiﬁcant in individual crime regressions, these
four items would be included together as explanatory variables in a subsequent
regression. Next, the weakest, non- signiﬁcant item was removed and the regression was
performed again. Nonsigniﬁcant policing practices were continually be removed from
the regression, one at a time, until only practices that were signiﬁcant at the .10 level or
lower remained.

The simultaneous regression modeling of individual policing practices was intend

to identify activities that were most strongly associated with crime reduction. However,

109

this methodology also resulted in a data reduction process that aided in interpretation of
the results. The ﬁnal analytic step in this procedure involved the identiﬁcation of the
similarities and differences among policing activities that demonstrated statistical
signiﬁcance when included in regressions with other policing practices. For example,
individual policing practices that retained statistical signiﬁcance with regard to robbery
might all relate to ofﬁcer training but items that explain burglary might result from items
that described increases in patrol visibility. Thus, a goal of this methodology was to
identify the nature or types of practices that were most closely associated reduction of
speciﬁc types of offenses.
Panel Data Regression Procedures

Analysis of Officers and Arrests

Research suggests that the relationship between the number of ofﬁcers and crime
rates may be reciprocal (Levitt, 2002; Marvel & Moody, 1996). For example, increases
in crime may lead to the hiring of additional ofﬁcers and falling crime may lead to
reductions in police force strength. Additionally, it is possible that the relationship
between minor disorder arrests and felonies may also be reciprocal since increases in
serious offenses may reduce the
amount of discretionary oﬁicer time that can be devoted to lesser offenses.
Consequently, using ordinary regression to determine which event is causing the other is
problematic. However, one approach that has been utilized to address this problem is a
vector autoregressive method known as Granger causality testing (Granger, 1969;

Paternoster & Bushway, 2001).

110

Recent work by Marvel and Moody (1996) utilized the Granger approach to
assess the relationship between ofﬁcers and crime, ﬁnding a small effect of crime on
ofﬁcers but a substantial effect of additional ofﬁcers on crime. To assess causal direction
under the Granger method, two separate sets of regressions were required. First, crime
was regressed on lags of itself and then lagged values of ofﬁcer rates were added and an
F test of the residuals from both models was utilized to determine whether the added
variables (i.e. ofﬁcers) signiﬁcantly increased the explanatory power of the model.
Second, officer rates were regressed on lagged ofﬁcer rates and then and then lagged
values of crime were added and an F test was utilized to determine whether the added
variables (i.e. crime) signiﬁcantly increased the explanatory power of the model. The
rationale behind the Granger approach was described by Loftin and McDowall (1982,
395)

“Granger [1969] argues that if one variable (X) causes another variable (Y), then X

should provide a more accurate prediction of Y’s present value than could be obtained
by using past values of Y alone.”

The ﬁrst step that was conducted pertaining to the Granger analyses involved the
assessment of the nature and direction of the relationships between crime and the
variables of interest. This step was carried out by regressing crime on lags of itself and
then adding lagged ofﬁcers into the model to determine whether the inclusion of the
ofﬁcers variables signiﬁcantly increased the explanation of crime. Regression residuals
from the model containing crime as the dependent variable with lagged crime values as
independent variables were then compared with the residuals from the model where
lagged ofﬁcer rates were included as independent variables using an F-test.

The possibility of reverse causal direction (i.e. that crime effects the number of

ofﬁcers) was also explored by regressing officer rates on prior (lagged) ofﬁcer rates and

Ill

comparing these residuals with models that included lagged values of crime as
explanatory variables. These steps were repeated for minor drug and misdemeanor
arrests as well. For these preliminary analyses, no control variables other than ﬁxed
effects for jurisdiction (place) and year were included.
Initial Models of Officers and Minor Arrests

After verifying that ofﬁcer rates and minor arrest rates effect each of the seven
crime types via the F-test, individual regressions of each crime on ofﬁcer rates, drug
arrest and misdemeanor arrest rates were conducted utilizing lagged values of these
independent variables as indicated by the models in Figure 41 below.

Figure 41. Initial Models of Crime as Explained by Ofﬁcers, Drug and Disorder Arrests

 

Crimea, = 115+ 71 + (1| Ofﬁcers (it-l)+ 0.2 Crimea”) + (13 Crime (“4) + B'Xit'i’ Sit
Crime“ = u, + y, + a. Drug Arrests (it-I) + a2 Cnme(it-1)+ a4 Crrmemz) + B'xiﬁ 8,,
Crime“ = u, + y, + or. Misd Arrests (M) + a2 Crimea.-.) + a4 Crimemz) + B'xi,+ sit

 

In the above models, Crime,t = represents the reported annual murder, aggravated
assault, robbery, rape, burglary, larceny and vehicle theft rates each jurisdiction from
during year t. The n; represents a vector of jurisdiction (place) ﬁxed effects and the 7,
represents the years 1992 through 2001. The ﬁrst year of the Granger analyses is 1992
due to the inclusion of a two-year lag of the dependent variable. The Ofﬁcers (n-1,, Drug
Arrests (“-1 ), and Misd Arrests (“-1 ) represent a one year lag of these variables. Officers
were modeled in rates per 10,000 citizens while drug and misdemeanor arrests were
modeled as rates per 1,000 citizens to reduce the number of leading zeros in regression
coefﬁcients. Crime m-” and Crime m4) represent one and two-year lagged values of the

dependent variable. Finally, the B'Xn reﬂects a vector of control variables pertaining to

112

changes in the percentage of population that is aged 15-24, nonwhite, employed as well
as the level of per capita income.

This analysis does not require that reciprocal relationships between ofﬁcers and
crime or arrests and crime be isolated or eliminated. However, by including lags of these
independent variables rather than current year values, the portion of the relationship that
is being assessed is the effect of independent variables in previous time periods on
current rates of crime. Marvel and Moody (1996) argued that no current year effect
between ofﬁcers and crime is likely since training and administrative delays would limit
the impact in the initial year of hire. However, it is possible that a simultaneous current
year effect between arrests and crime may occur and this methodology is unable to assess
this effect. Consequently, the effect of minor arrests on more serious felonies may be
stronger than indicated by Granger regressions.

Effect sizes in Granger analyses must be adjusted for the inﬂuence of lagged
dependent variables. The formula for this adjustment is relatively simple and requires a
summation of the beta coefﬁcients for the one and two-year lagged independent
variables, divided by one minus the summation of the beta coefﬁcients of the one and
two-year lagged dependent variables as shown in Figure 42 (Marvel and Moody, 1996).

Figure 42. Formula for Calculation of Effect Sizes in Granger Analyses.

 

(independent var. lagl + independent var. lag2)

 

(1 - (dependent var. lagl + dependent var. lag2))

 

113

Final Model (Officers and Minor Arrests)

Changes in ofﬁcer rates, drug arrest and minor disorder arrest rates during the
19905 were correlated as indicated by Table 5 below. Consequently it was necessary to
determine which of these variables explained unique variation in crime. After assessing
the crime effects of ofﬁcers, drug and disorder arrest individually, these variables were
included together in a single regression along with all ﬁxed effects and control variables.
Results from this ﬁnal model allowed an assessment of the amount of crime that is
explained by changes in ofﬁcer rates while controlling for minor drug and disorder
offenses arrest rates and vice versa. The ﬁnal regression model pertaining to ofﬁcers and

minor arrests is listed below:

Figure 43. Final Model of Crime as Explained by Ofﬁcers, Drug and Disorder Arrests

 

Crime“ = u; + y, + (1. Ofﬁcers m-”+ (12 Drug Arrests m-” + a3 Misd Arrests m-” +
a4 Crimean) + asCrime (it-2) + B'xit + an

 

Rggression of Policing Practices Indexes vs. Individual Items

Analyses of changes in policing practices (i.e. tactics and strategies other than
arrest) were conducted in two parts. The ﬁrst portion assessed changes among broad
categories of practices through the creation of summative indexes of individual survey
items that correspond to types or classes of tactics. The second part assessed the
inﬂuence of changes in the reported implementation of each individual tactic. The
National Survey of Community Policing Strategies was administered in two waves (1993
and 1997). Both waves consisted of 80 items pertaining to the current implementation of

speciﬁc tactics and strategies by each police organization.

114

Analyses of reported changes among summative indexes of groups of tactics, such
as problem solving or community collaboration fundamentally differ ﬁ'om analysis of
speciﬁc individual tactics in a number of important respects. First, summative indexes of
tactics measure the breadth or range of implementation based on generic types or
classiﬁcations. For example, an agency may report implementing 3 problem solving
tactics in 1993, followed by 10 problem solving tactics in 1997. Thus, the problem
solving index would measure a broadening of the problem solving efforts through the
addition of 7 new tactics. Measures of the breadth of implementation are useful for
assessing overall trends within jurisdictions but offer little information pertaining to
speciﬁc mechanisms through which any resulting crime reduction occurred.

In contrast, analyses of individual tactics provide more speciﬁc information as to
the nature of the policing practice (e. g. increased foot patrol) that was implemented. This
approach may be more effective in identifying effective tactics than summative indexes
under certain conditions. For example, if the majority of problem solving tactics are
ineffective, but a few are excellent methods of crime reduction, measures of the problem
solving index would likely fail to reach statistical signiﬁcance. However, measures of
individual items should prove useful under this condition. Nevertheless, if the generic
broadening of problem solving or community collaboration by police jurisdictions results
in crime reduction, the summative index approach may be useful. Consequently, both
approaches were conducted in this study.

Regr_'ession Models for Policing Practices
The Granger approach cannot be conducted with only two years or panels of data

as is the case with available information pertaining to policing tactics and strategies.

115

However, a similar method was utilized that includes information about crime levels and
trends prior to the ﬁrst wave that was conducted in 1993. The National Survey of
Community Policing Strategies includes information about changes that occurred after
the baseline year in 1993. However, crime levels or trends in years just prior to 1993
may have inﬂuenced police administrative decisions in 1993 or thereafter. For example,
a high level of drug offenses in 1990 may have increased the likelihood of participation
in an anti-drug task force. Additionally, a rapid increase in crime during the years
immediately prior to 1993 may have increased the adoption of crime analysis techniques.
Thus, it is important to include this information in models of the effects of policing
changes on crime.

Police organizational changes do not occur instantaneously for a number of
reasons. First, annual crime information and analysis from a given year may not be
available until several months into the following year. Second, organizational planning
and decision making in response to recent crime information may require substantial
time. Finally, the procurement of funds and re-organization of personnel or resources
require additional time before new methods are implemented at the line-level. As a
result, the models of policing changes on crime included both the crime level from 1990
(3 years prior) as well as the change or trend from 1990 to 1993. The objective of
including these two variables as predictors is to utilize the maximum amount of crime
information that may have inﬂuenced levels of problem solving, crime analysis,
collaboration and other practices that were reported in the baseline year (1993) and

thereafter.

116

 

Each of the seven primary crime variables were included as dependent variables
in regressions for the corresponding survey years 1993 and 1997. Additionally, the
inclusion of ﬁxed effects for jurisdiction (i.e. place) and year renders regressions that are
equivalent to change scores in the two-period panel models (Allison, 1995).
Consequently, with regard to robbery rates in each jurisdiction, the change in the robbery
rate ﬁom 1993 to 1997 was regressed. Change scores were also utilized for the other
independent and dependent variables in the two wave panel. For example, since the
index of place oriented tactics ranged from 0 to 10, the resulting change score for any
given jurisdiction may range from -10 to 10. Thus, if an agency reported implementation
of 3 place oriented tactics in 1993 and discontinued all of these tactics by 1997, a change-
score of -3 would be assigned. Conversely, if an agency reported implementing one
place oriented tactic in 1993 and ﬁve of these tactics in 1997, the change score would be
+4. Control variables pertaining to changes in the percentage of population that is aged
15-24, nonwhite, employed and level of per capita income were also included in the
regression models. The initial regression model pertaining to the summative policing
practice indexes is listed below in Figure 44.

Figure 44. Initial Model of Crime as Explained by Policing Practice Indexes

 

ACrimek = at + (12 APolicing Index k + a3 Crimekquo) + a4 ACrimemqg34990) + B'Axk + 3k

 

In a two wave panel design such as the National Community Policing Strategies
Survey, regressions using change scores (i.e. ﬁrst diﬁerenced values) from wave-1 to
wave-2 yield identical results to regressions that utilize ﬁxed effects for place and year.

Thus, the more simple change score method was utilized. As illustrated above, the

117

ACrimek represents the change in reported annual murder, aggravated assault, robbery,
rape, burglary, larceny and vehicle theft rates will in each jurisdiction from 1993 to 1997
(e.g. the 1993 robbery rate subtracted from the 1997 robbery rate). The APolicing Index k
represents the change score for each individual policing index (e. g. the place oriented
index value from 1997 minus the place oriented index value from 1993). The
Cﬁmekugoo) represents the level or rate of the dependent variable in 1990 and the
ACrimek(1993-.990, represents the change in the dependent variable from 1990 to 1993 (e. g.
the 1990 robbery rate subtracted from the 1993 robbery rate). As previously discussed,
the inclusion of information pertaining to crime prior to 1993 will reduce the effects of
autocorrelation and provide additional detail pertaining to crime events that may have
inﬂuenced organizational decisions pertaining to policing practices. Finally, the B'xi.
represents a vector of control variables pertaining to changes in the percentage of
population that is aged 15-24, nonwhite, employed and the level of per capita income.
Data Reduction and Final Model (Policing Indexes)

Several of the policing practice indexes were correlated as indicated by Table 6
below. Consequently it was necessary to determine which indexes explained unique
variation in crime and remove those that did not add to the crime reduction model. After
assessing the crime effects of each summative index individually, all indexes that
demonstrated statistical signiﬁcance were entered into a regression simultaneously.
Indexes that did not explain a signiﬁcant amount of variation when entered into the full
models of crime were removed from subsequent regressions, one at a time, beginning
with the least inﬂuential (i.e. least signiﬁcant) index. This procedure was followed until

all remaining indexes explained variation in crime at the .10 level or lower. The ﬁnal or

118

“full model” pertaining to regressions of policing practice indexes is listed in Figure 45
below.

Figure 45. Final Model of Crime as Explained by Policing Practice Indexes

 

ACrimek = a1 + a2 APolicing Indexl k + a3 APolicing Index2 k + a4APolicing Index3 k +
as RObbBI'quqoo) + (14 ARObberka993-I990) + B’Axk + 8k

 

The ﬁnal model shown in Figure 45 above is nearly identical to the initial model but
includes all of the relevant policing practice indexes together as explanatory variables in
a single regression.9
Analyses of Individual Policing Practice Items

After the changes in breadth of policing strategies and tactics are assessed, the
effects of each of the individual (speciﬁc) policing practices were examined utilizing the
same modeling approach described above. As a ﬁrst step, each of the crime seven types
were regressed against each of the 80 individual survey items separately with control
variables and ﬁxed effects included. Analyses of individual tactics followed the equation

listed in Figure 46 below.

Figure 46. Initial Model of Crime as Explained by Individual Policing Practice Items

 

ACrimek = at + a2 APolicing Practicel k + a3 Crimekquo) + a4 ACI’ika(|993-|990) + B'Axk
+ 8k

 

 

9 The term “relevant indexes” refers to those that demonstrated a signiﬁcant effect on crime in the previous
individual regressions.

119

Data Reduction and Final Model (Individual Policing Items)

The following step involved a data reduction procedure that was similar to the one
implemented with the summative indexes. Each of the items that signiﬁcantly explained
crime were simultaneously included in a regression model and the weakest non-
signiﬁcant items were removed one at a time until only signiﬁcant policing practices

remained.

Figure 47. F inal Model of Crime as Explained by Individual Policing Practice Items

 

ACrimek = a. + a2 APolicing Practicel k + a3 APolicing Practice2 k + a3 APolicing
Practice3 k +
04 Cﬁmekmoo) + (15 ACYimekt l993-I990) + B’Axk + 8k

 

The ﬁnal model shown in Figure 47 above is nearly identical to the initial model of
individual policing practice items but includes all of the relevant individual policing
practices together as explanatory variables in a single regression.
Methodological Challenges

Over-Reporting of Innovative Policing Practices

Some researchers have suggested that political pressures emanating from the
popularity of community policing and problem solving concepts have led to exaggerated
or inaccurate claims by police officials in survey formats (Maguire & Katz, 2002).
Consequently, if reported problem solving or other policing activities are not truly being
conducted, the likelihood of ﬁnding an effect is reduced. Furthermore, differences in the
intensity or quality of implementation across agencies may also tend to decrease the size

or signiﬁcance of the policing effect.

120

Nevertheless, these potential sources of error tend to strengthen or bolster the
validity of ﬁndings since the econometric models must be strong enough overcome these
challenges in order to demonstrate statistical signiﬁcance. There is no reason to believe
that responding police ofﬁcials from jurisdictions where crime was declining were more
likely than others to exaggerate about community policing or problem solving. Indeed,
the greatest political pressure for reform and over-reporting of innovative activities would
occur in jurisdictions where crime had remained at high levels. If agencies where crime
was not declining tended to over-report innovative activities, the effects of policing
practice variables on crime would be underestimated. Consequently, to ensure that
important relationships between policing changes and crime were not overlooked,
variables that were signiﬁcant at or below the .10 level are highlighted and discussed in
subsequent sections.

Arrest Data Standardization Issues

One potential concern that has been noted by researchers pertaining to data in the
Arrest Master Files is the lack of consistency across agencies with regard to reporting
standards. Some agencies, for example, report an arrest only when a suspect is taken to
the police station or a jail facility. In other police jurisdictions, however, an arrest may
be reported whenever a court appearance citation or ticket for a misdemeanor offense is
issued. Nevertheless, although these differences across agencies are problematic to
cross-sectional analyses, they are not at all problematic to assessments of changes within
each jurisdiction. The panel data analyses conducted in this study measure within subject
changes. In this case, for example, two agencies of approximately equal size and

population demographics may have different criteria for reporting arrests such that

121

“agency A” reports every appearance citation as an arrest and averages 2,000 arrests per
year while “agency B” utilizes a different standard that results in an average of only 500
arrests. In a cross-sectional regression this issue would be problematic unless a variable
were added to account for the different reporting standards. However, since these panel
data analyses in this study assess change, arrests are measured with respect to the
baseline level in each jurisdiction. Thus, change scores for each agency are relative to
local conditions and not subject to standardization error in reporting.
Omitted Variables and Threats to Validity

As described in the previous section, routine cross-sectional regressions are
susceptible to differences across jurisdictions that explain the dependent variable. Thus,
any difference that is relatively stable across time between jurisdictions and associated
with crime should be entered into regressions. Consequently, the list of potential
variables that should be included in cross sectional regressions is lengthy. For example,
one city may have a longstanding gang problem, another may be located near a large
urban area and yet another may have inadequate jail or law enforcement facilities. These
relatively stable characteristics are important for explaining differences in crime levels.
However, by assessing change through panel data analyses with ﬁxed effects, only
variables that changed and trended with crime across a large number of jurisdictions need
to be added to regressions (Frees, 2004). Additionally, given that crime did not decline in
most jurisdictions at the same time or in precisely the same year, the list of inﬂuential
variables is shortened. Stable characteristics cannot explain changes in crime and are not

necessary in panel data regressions of change within jurisdictions.

122

Robustness Checks

After initial regression results were obtained, a series of additional regressions to
assess the robustness of ﬁndings pertaining to changes in ofﬁcers, arrests and tactics were
conducted. To ensure that the results were not being driven entirely by large or small
places, jurisdictions with over 250,000 citizens and those with less than 25,000 citizens
were removed to determine whether regression results continued to be statistically
signiﬁcant. Additionally, to ensure that the policing effects due to changes in policing
tactics and strategies were not being driven by changes in ofﬁcers or minor arrest rates,
these variables were entered into regressions along with policing practice items as a ﬁnal
robustness check.

Variable Descriptives Pertaining to the Officer and Arrest Database

Table 1a shows the amount and distribution of total crime that was reported to the
Federal Bureau of Investigation via the UCR system during the 1990 to 2001 time period.
Overall, 12,975 police agencies reported at least one complete year of crime data. The
vast majority of these agencies (11,625) also reported at least one year of arrest data.
However, the Granger methodology utilized in this study requires that at least three
consecutive years of data be present due to the inclusion of two lags the crime variables.
In total, the arrest data subsample includes over 89% of the total reported crime in the
US. by 9,362 jurisdictions.

Consistent with UCR population demographics, the arrest data indicate that the
largest portion of crime was reported in Southern region (37.4%), followed by the West

(27.6%), Midwest (18.1%) and Northeast (16.9%). Additionally, both UCR and Arrest

123

Master File data indicate that over 81% of reported crime occurred in municipal

jurisdictions, with remainder (17.3% to 18.5%) being reported by county agencies.

Table 1a. Comparison of UCR Population and Arrest Sample

 

 

UCR Population Arrest Sample
Total Crime Percent Total Crime Percent
Reported Crimes 140,501,109 100 125,160,314 89.1
Region

Northeast 21,579,005 15.4 21 ,087,1 13 16.9

Midwest 26,946,393 19.2 22,716,463 18.1

South 56,455,087 40.2 46,797,531 37.4

West 35,520,624 25.3 34,559,206 27.6
Government Type

Municipal 1 14,543,782 81.5 103,458,992 82.7

County 25,957,327 18.5 21,701,321 17.3

Jurisdictions 12,975 100 9,362 72.2

 

Table 2 details the mean levels and changes in ofﬁcer rates, drug arrest, and minor
disorder arrest rates during the 19905. The nationwide mean ofﬁcer rate per 10,000
citizens was 20.2 % in 1990 but increased by a substantial 16.4% by the year 2000. As a
result, the national officer rate increased more than 3.3 ofﬁcers per 10,000 citizens by the
end of the decade. Additionally, the national drug arrest rate per 1,000 citizens was 4.85
in the initial year (1990). However, it is noteworthy that this minor drug arrest rate
increased more than 106% by the year 2000. Finally, the national arrest rate for minor
disorder offenses was more than twice that of the drug arrest rate in 1990 (10.41 ).
However, by the year 2000 the disorder arrest rate declined substantially (-21 .9%) such

that arrests for minor drug offenses were more frequent than those for minor disorder.

124

Table 2. Changes in Ofﬁcer and Arrests Rates from FBI data

 

 

 

FBI Arrest Data Mean Standard Change % Change

1 990 Deviation 2000 2000
Ofﬁcer Rate (per 10,000) 20.18 39.01 3.31 16.4
Drug Arrest Rate (per 1,000) 4.85 9.24 5.18 106.8
Disorder Arrest Rate (per 1,000) 10.41 16.89 -2.28 ~21.9
N=9,362

Variable Descriptives Pertaining to Policing the Practices Database

The National Survey of Community Policing Strategies, was conducted in 1993
and 1997. In total, 1,138 agencies responded to both survey waves and possessed an ORI
code that allowed them to report ofﬁcial crime data to the FBI. However, approximately
77% of these agencies (868) reported complete crime data in both survey years and were
included in analyses. The vast majority of agencies that did not report complete crime
data in both 1993 and 1997 represent smaller jurisdictions. Fully three quarters of these
agencies police populations of less than 50,000 citizens. Consequently, the estimated
amount of crime that these incomplete reporting agencies represent is less than 1.9% of
the total crime that was reported by the full 1,138 agencies. The impact of this
underrepresentation of jurisdictions with less than 50,000 citizens is small in terms of
total crime in the United States. This issue is discussed in detail in the limitations
section.

Although the policing survey sample is much smaller than that of the arrest data
in terms of the number of reporting jurisdictions (868), these data include nearly one half
of all reported crime in United States during the years 1993 and 1997 (48.9%).
Moreover, the distribution of reported crime that is represented by policing survey

respondents is similar to that of the broader UCR population. Consistent with crime

125

reports in the UCR data, the policing survey subsample indicates that slightly more than
40% of the total reported crime in the US. originated from the Southern region. The
second-highest level of reported crime occurred in the West (31.5%), followed by the
Midwest (17.6%), and Northeast (9.4%). Finally, the distribution of crime with regard to
level of government was similar among the policing survey subsample and the broader
UCR population with more than 81% of crime originating from municipal agencies.

The proportion of total crime in the West among the policing survey subsample is
slightly higher than that of the UCR population. Additionally, the amount of crime
reported in the northeastern region among the policing survey data was approximately
6% below that of the UCR population. This underrepresentation of crime ffom the
Northeast is a minor data limitation and is discussed in the subsequent limitations section.
Nevertheless, the overall representation of crime among the policing survey subsample is

sufﬁciently similar to allow inference to the broader UCR population.

Table 3. Comparison of UCR Population and Police Survey Subsample

 

 

UCR Population Police Survey Sample
Total Crime Percent Total Crime Percent
Reported Crimes 23,703,384 100 11,593,198 48.9
Region

Northeast 3,621,852 15.3 1,086,608 9.4

Midwest 4,343,194 18.3 2,045,] 16 17.6

South 9,648,825 40.7 4,805,422 41.5

West 6,089,513 25.7 3,656,052 31.5
Government Type

Municipal 19,272,282 81.3 9,742,602 84.0

County 4,431,102 18.7 1,850,596 16.0

Jurisdictions 10,270 100 868 8.5

 

126

Table 4 details the mean levels and changes among the eight primary policing
practice indexes from 1993 to 1997.'0 Overall, reported policing practices among all of
the indexes increased during the mid-19905. The mean number of community policing
policies in 1993 was only 1.3 out of a possible 5 total. However, community policing
policies increased to a total of 2.3 on average by 1997. Place oriented tactics increased
ffom 3.5 to 4.9 on a scale of zero to 10. The mean number of citizen involvement
practices also increased from 2.2 to 3.0. Finally, problem solving practices increased by
nearly 39% from 6.6 to 8.5 on a scale ranging from zero to 16.

Table 4. Changes in Policing Practices from the National Survey of Community Policing
Strategies

 

 

Policing Survey Mean Index Standard Change % Change

1993 Range Deviation l 997
Collaboration 7.2 0 - 16 2.96 1.2 16.7
Problem Solving 6.6 0 - 16 3.38 1.9 28.8
Citizen Involvement 2.2 0 - 10 1.87 0.8 36.4
Code Enforcement 1.7 0 - 4 1.25 0.5 29.4
CP Policy 1.3 0 - 5 1.46 1.0 76.9
Crime Analysis 1.7 0 - 5 1.23 0.1 6.0
Place Oriented 3.5 0 - 10 2.39 1.4 40.0
Efﬁciency 6.5 0 - 12 2.42 0.8 12.3
N=868

Although changes in ofﬁcer rates, drug arrest and minor disorder arrest rates from
1993 to 1997 were signiﬁcantly correlated (Table 5), none of these variables is highly
correlated with changes in policing practices. In fact, only two of the eight indexes
(place oriented and efﬁciency efforts) are weakly correlated with changes in ofﬁcers and
arrests. This lack of correlation is noteworthy because ofﬁcer and arrest rates can not be
modeled with the two wave policing survey data using a Granger methodology. If officer
and arrest rates were highly correlated with changes in policing practices, their exclusion

from regression analyses would introduce omitted variable bias. That is, the relationships

 

'0 See Appendix D for descriptive statistics pertaining to each of the 80 individual policing practices.

127

between policing practices and crime might actually be due to the fact that practices were
correlated with officers or arrests. Nevertheless, ofﬁcer and arrest rates were included as
checks of robustness after regression of policing practices were carried out (see Tables
43-36).

The most substantial correlation shown in Table 5 below relates to ofﬁcer and
drug arrest rates (.60). Interestingly however, the correlation between ofﬁcers and
misdemeanor arrests is substantially lower (.21). Consequently, the addition of ofﬁcers
to a police agency may not automatically lead to equal amounts of increases in all types
of arrests. Nevertheless, these correlations require that ofﬁcers, drug arrest, and minor
disorder arrests be modeled together in a single equation to estimate the unique effect of
each variable. Finally, the Chronbach’s alpha statistics indicate that the individual items
that comprise the indexes for minor drug arrests (.87) and disorder arrests (.61) are
sufficiently similar to be included together.

Table 5. Correlation of Ofﬁcers and Minor Arrests and Policing Indexes

 

 

Tactic Ofﬁcer Rate Drug Arrest Rate Disorder Arrest Rate
Ofﬁcer Rate -- .60 .21
Drug Arrest Rate .60 -- .19
Misd. Arrest Rate .21 .19 --
Citizen Involvement n.s. n.s. n.s.
Code Enforcement n.s. n.s. n.s.
Collaboration Efforts n.s. n.s. n.s.
CP Policy n.s. n.s. n.s.
Crime Analysis n.s. n.s. n.s.
Problem Solving n.s. n.s. us.
Place Oriented -.07 n.s. n.s.
Efﬁciency .09 us. n.s.
Alpha -- a=.87 a=.61

 

Table 6 below indicates that policing practice indexes are correlated at varying
levels. For example, citizen involvement is only weakly correlated with code

enforcement (.14) and crime analysis (.17). However, changes in the number of citizen

128

involvement practices demonstrate a moderately strong correlation with problem-solving
(.41) and collaboration efforts (.32). As a result, the unique contribution of each of these
concepts to crime reduction was obtained by modeling relevant indexes together in a
single regression. Finally, the Chronbach’s alpha statistics indicate that the individual
items that comprise the indexes of policing practices are sufﬁciently similar to be

included together.

129

 

 

wewnz
more Que Ste onus its more «One more 2%
-- mm. 2.. om. mm. 2. mm. on. aeosem
mm. -- 3.. mm. mm. 3. a. mm. 355 8a:
S. 2.. -- 2. 9.. No S. 3. 220m Bose;
am. .8. R. -- 2. am. 2. S. were? 256
2. mm. 9.. t. -- :H. o... a. 3:0; .5
2. X. mm. em. mm. -- mm. mm. 385 83882.8
2. 2. R. 2. 2. mm. -- 3. .aoeueeem 88
on. 3. :.. t. a. 2. z. -- 605295 555
p855 wag—om mix—£2 mtobm Seem .302:
5865.”. 82m 8038a 0&5 3:8 .6 .2230 88 85.6

 

$3780: 852.. 88?. was; 2: 95:2 825050 .0 29?

130

CHAPTER 5
RESULTS
Section I - The Inﬂuence of Ofﬁcers, Drug and Disorder Arrests

Although hypotheses l and 2 in this study focus on the effects of the number of
ofﬁcers and minor arrests on crime rates, it was possible that the causal direction of this
relationship could be opposite of the expected direction. More speciﬁcally, crime could
have exerted a causal impact on officer and minor arrest rates. Additionally, the
relationships between crime and the independent variables could have been reciprocal.
Consequently, a preliminary step that was required in Granger analyses involved the
assessment of the nature and direction of the relationships between crime and the
variables of interest. As discussed in the methodology, this step was carried out by
regressing crime on lags of itself and then adding lagged ofﬁcers into the model to
determine whether the inclusion of the officers variables signiﬁcantly increased the
explanation of crime via an F-test of the residuals. These steps were repeated for minor
drug and disorder arrests as well. For these preliminary analyses, no control variables
other than ﬁxed effects for jurisdiction (place) and year were included.
Preliminary Granger Causality Assessments

The results of the regressions and F-tests are shown in Tables 7 -12. These data
clearly suggest that during the 1990s, ofﬁcer rates had an impact on crime and that the
relationship was unidirectional.ll Crime did not signiﬁcantly effect the number officers
in police jurisdictions across the nation. More speciﬁcally, the values of the F-tests in

Table 7 indicated that rates of murder, robbery, burglary and other crime types did not

 

” Reciprocal causality, however, would not be problematic as the Granger method is speciﬁcally designed
for use among variables with causality in both directions.

131

explain subsequent changes in ofﬁcer rates. However, the number of ofﬁcers per capita
did have a signiﬁcant impact on robbery, aggravated assault, burglary, larceny and
vehicle theft (but not murder or rape) according to F-test values (Table 8). When no
other explanatory or control variables were entered into the regression model, R-squared
values indicated that changes in ofﬁcer rates explained 2.6% of the variation in larceny
and slightly over 1% of the variation in aggravated assault (.014), burglary (.011) and
vehicle theft (.012). To aid in the comparison of the explained variance (R-squared)
between ofﬁcer and misdemeanor arrests, agencies that did not report both ofﬁcers and
minor arrests were removed from these preliminary analyses and therefore, the number or
jurisdictions (n=8,050) is slightly smaller for the initial Granger testing than for
subsequent regressions. The resulting sample size 47,097 indicates the total repeated
observations for each year during the 19903.

Table 7. Granger Analyses of the Impact of Crime on the Number of Police

 

 

 

Independent F F R2 R2
Variables Value Probability Value Increase
Police -- -- .04 --
Murder 1.00 .81 .04 n.s.
Rape 1.00 .99 .04 n.s.
Robbery 1.00 .99 .04 n.s.
Assault 1 .00 .99 .04 n.s.
Burglary 1.00 .97 .04 n.s.
Larceny 1 .00 .96 .04 n.s.
Vehicle theft 1.00 .99 .04 n.s.
Sample Size 47,097

D.F. 47,092

Agency (N )=8,050

132

Table 8. Granger Analyses of the Impact of the Number of Police on Crime

 

 

Police effects F F R2 R2
on Crime Value Probability Value Increase
Murder 1.01 .547 .119 n.s. +
Rape 1.04 .000 .063 n.s. +
Robbery 1 .06 .000 .746 .0001
Assault 1.12 .000 .41 1 .014
Burglary 1.05 .000 .530 .01 1
Larceny 1 .3 8 .000 .460 .026
Vehicle theft 1.12 .000 .605 .012
Sample Size 47,097

D.F. 47,092

 

i t-value not signiﬁcant for one year lag of officers
Agency (N)=8,050

The initial Granger assessment of causal direction with regard to drug and
disorder arrests suggested that minor arrests impacted felony offense rates and that these
relationships during the 19905 were also unidirectional. Lagged values of crime did not
signiﬁcantly effect the number minor arrests in police jurisdictions across the nation.
More speciﬁcally, the values of the F-tests in Tables 9 and 11 indicated that rates of
murder, robbery, burglary and other crime types did not explain subsequent changes in
minor drug or disorder arrest rates.

The number of minor drug arrests per 1,000 citizens as indicated in Table 10
signiﬁcantly impacted all crime categories with the exception of rape according to F-test
values. When no other explanatory or control variables were entered into the regression
model, R-squared values indicated that changes in drug arrest rates explained 2% of the
variation in larceny, slightly over 1% of the variation in burglary (.014) and vehicle theft

(.014) and less then 1% of the variation in robbery (.004) and murder (.005).

133

Table 9. Granger Analyses of the Impact of Crime on Misdemeanor
Drug Arrests

 

 

Independent F F R2 R2
Variables Value Probability Value Increase
Drug Arrests -- -- .126 --
Murder 1.00 .807 .126 n.s.
Rape 1.00 .930 .127 n.s.
Robbery 1.01 .169 .131 n.s.
Assault 1.01 .377 .130 n.s.
Burglary 1.01 .392 .129 n.s.
Larceny 1.01 .392 .129 n.s.
Vehicle theft 1.00 .818 .127 n.s.
Sample Size 47,097

D.F. 47,092

 

Agency (N )=8,050

Table 10. Granger Analyses of the Impact of Misdemeanor Drug Arrests
on Crime

 

 

Drug Arrest effects F F R2 R2
on Crime Value Probability Value Increase
Murder 1 .07 .000 .124 .005
Rape 1.05 .000 .065 n.s. "‘
Robbery 1 .47 .000 .748 .004
Assault 1.01 .250 .409 .012
Burglary 1.49 .000 .533 .014
Larceny l .17 .000 .453 .020
Vehicle theft 1.39 .000 .610 .014
Sample Size 47,097

D.F. 47,092

 

+ t-value not signiﬁcant for one year lag of minor drug arrests

Agency (N )=8,050

The number of minor, non-drug related disorder arrests per 1,000 citizens as
indicated in Table 12 signiﬁcantly impacted robbery, burglary and vehicle theft according
to F-test values. However, the effects of disorder arrest rates on crime were small. When
no other explanatory or control variables were entered in to the regression model, R-
squared values indicated that changes in these disorder arrest rates explained less then 1%

of the variation in robbery (.002), burglary (.003) and vehicle theft (.002).

134

Table 11. Granger Analyses of the Impact of Crime on Misdemeanor
Disorder Arrests

 

 

 

Independent F F R2 R2
Variables Value Probability Value Increase
Disorder Arrests -- -- .328 --
Murder 1.000 .900 .329 n.s.
Rape 1.00 .877 .329 n.s.
Robbery 1.00 .624 .330 n.s.
Assault 1.01 .388 .329 n.s.
Burglary 1.00 .648 .329 n.s.
Larceny 1.00 .813 .329 n.s.
Vehicle theft 1.00 .761 .329 n.s.
Sample Size 47,097

D.F. 47,092

N=8,050

Table 12. Granger Analyses of the Impact of Misdemeanor Disorder Arrests
on Crime

 

 

Disorder Arrest F F R2 R2
Effects on Crime Value Probability Value Increase
Murder 1.02 .011 .121 n.s. 1‘
Rape 1.02 .053 .063 n.s. +
Robbery 1 .09 .000 .747 .002
Assault 1.02 .025 .397 n.s. +
Burglary 1.02 .016 .522 .003
Larceny 1.02 .079 .435 n.s. *
Vehicle theft 1.03 .000 .592 .002
Sample Size 47,097

D.F. 47,092

 

+ t-value not signiﬁcant for one year lag of misdemeanor arrests.

Agency (N)=8,050

With crime as the dependent variable, Tables 13-15 (below) show the beta
coeﬁicients and t-statistic values for the one and two-year lags of ofﬁcers and minor
arrests in the preliminary Granger causality regressions. The vast majority of the
signiﬁcant effects of ofﬁcers and minor arrests occurred in the previous year (i.e. the ﬁrst
lag) as indicated by t values greater than, or less than 1.65. However, effect sizes as
indicated by beta coefﬁcients in Granger analyses cannot be directly interpreted as in

standard OLS regression. Interpretation requires adjustment for the inclusion of lagged

135

dependent variables. Tables 16 through 18 provide ﬁirther detail pertaining to effect
sizes.

Table 13. One and Two-year Beta Coefﬁcients for the Effects of Ofﬁcer Rates on Crime

 

 

One-Year Lag Two-Year Lag
Ofﬁcer Effects F F
on Crime Coef. t Coef. t Value Probability
Murder -.0013 -1.15 -.0009 -1.35 1.01 .547
Rape -.0155 -1.60 -.0058 -0.99 1.04 .000
Robbery -. 1003 -3.13 -.0130 -1 .18 1.06 .000
Assault -.6264 -12.23 .0331 0.81 1.12 .000
Burglary -.9815 -1 1.77 -.0240 -0.35 1.05 .000
Larceny -3.5027 -10.74 .0781 0.40 1.38 .000
Vehicle theft -.8329 -12.57 .0464 0.92 1.12 .000
Sample Size 47,097
D.F. 47,092

 

Agency (N)=8,050

Table 14. One and Two-Year Beta Coefﬁcients for the Effects of Drug Arrests on Crime

 

 

One-Year Lag Two-Year Lag
Drug Arrest F F
Effects on Crime Coef. t Coef. t Value Probability
Murder -.0299 -l .77 -.0133 -1.44 1.07 .000
Rape -.0587 -0.85 -.0749 -3.18 1.05 .000
Robbery -.8973 -3.43 -.1859 -l .1 l 1.47 .000
Assault -2.801 -2.89 .4552 1.36 1.01 .250
Burglary -4.947 -3.59 -.2030 -0.44 1.49 .000
Larceny -14.867 -2.62 1.80 1.13 1.17 .000
Vehicle theft -3.9503 -3.05 .62 1.20 1.39 .000
Sample Size 47,097
D.F. 47,092

 

Agency (N)=8,050

136

Table 15. One and Two-Year Beta Coefﬁcients for the Effects of Disorder Arrests on
Crime

 

 

One-Year Lag Two-Year Lag
Disorder Arrest F F
Effects on Crime Coef. t Coef. t Value Probability
Murder -.0209 -l .15 .0044 0.28 1.02 .01 1
Rape .1122 1.78 -.0182 -0.29 1.02 .053
Robbery -l .177 -4.24 0.3459 1.36 1.09 .000
Assault -.9002 -1.06 1.1025 1.98 1.02 .025
Burglary -2.647 -1.92 1.3092 1.33 1.02 .016
Larceny -4.400 -1 .24 1.9353 0.86 1.02 .079
Vehicle theft -2.230 -1 .82 1.6303 2.05 1.03 .000
Sample Size 47,097
D.F. 47,092

 

Agency (N )=8,050

Effect sizes in Granger analyses must be adjusted for the inﬂuence of lagged
dependent variables. The formula for this adjustment is shown in Tables 16-18 along
with beta coefﬁcients for one and two-year lags of the independent variables. When no
other explanatory or control variables were entered in to the regression model, results
indicated that an increase of 10 ofﬁcers was associated with a reduction of over 8
larcenies (-8.13), and more than 2 burglaries (-2.49) and vehicle thefts (-2.88) per

100,000 citizens.

137

Table 16. Adjusted Effect Sizes for Initial Granger Regressions of Ofﬁcers and Crime

 

Indep. Variable

Dependent Variable Calculated Total

 

Effects Effects Effect
Ofﬁcer Effects Lag l Lag 2 Lag 1 Lag 2 (x1 + x2) /
on Crime (1 - (yl + y2))
Murder -- -- -- -- n.s.
Rape -- -- -- -- n.s.
Robbery -.100 n.s. .984 -.121 - 0.73
Assault -.626 n.s. .652 -.101 - 1.40
Burglary -.982 n.s. .647 -.041 - 2.49
Larceny -3.502 n.s. .665 -.096 - 8.13
Vehicle theft -.833 n.s. .828 -.117 - 2.88

x1 x2 yl y2

 

Ofﬁcer rates are per 10,000 citizens and crime rates are per 100,000.

Table 17 (below) indicates that each additional 100 minor drug arrests by police

agencies were associated with a reduction of 34 larcenies (-34.09), over 13 burglaries (-

13.14) and vehicle thefts (-13.62), as well as 6.45 robberies per 100,000 citizens.

Additionally, Table 18 indicates that each additional 100 minor disorder arrests by police

agencies were associated with a reduction of 8.5 robberies (-8.54) and 6.5 burglaries (-

6.51) and .61 vehicle thefts per 100,000 citizens when no other explanatory or control

variables were entered in to the regression model.

138

Table 17. Adjusted Effect Sizes for Initial Granger Regressions of Drug Arrests and

 

 

Crime
Indep. Variable Dependent Variable Calculated Total
Effects Effects Effect
Drug Arrest Lag 1 Lag 2 Lag 1 Lag 2 (x1 + x2)/
Effects on Crime (1 - (yl +y2))
Murder -.030 n.s. .158 .077 -0.039
Rape -- -- -- -- n.s.
Robbery -.897 n.s. .980 -.120 -6.45
Assault -- -- -- -- n.s.
Burglary -4.947 n.s. .642 -.034 -l3.14
Larceny -14.861 n.s. .647 -.083 -34.09
Vehicle theft -3.950 n.s. .822 -.112 -l3.62
x1 x2 yl y2

 

Drug arrest rates are per 1,000 citizens and crime rates are per 100,000.

Table 18. Adjusted Effect Sizes for Initial Granger Regressions of Disorder Arrests and

 

 

Crime
Indep. Variable Dependent Variable Calculated Total

Effects Effects Effect
Disorder Arrest Lag l Lag 2 Lag l Lag 2 (x1 + x2) /
Effects on Crime (1 - (yl +y2))
Murder -- -- -- -- n.s.
Rape -- -- -- -- n.s.
Robbery -1.17 n.s. .984 -.121 -8.54
Assault -- -- -- -- n.s.
Burglary -2.648 n.s. .618 -.025 -6.51
Larceny -- -- -- -- n.s.
Vehicle theft -2.230 1.630 .795 -.093 -0.61

x1 x2 yl y2

 

Disorder arrest rates are per 1,000 citizens and crime rates are per 100,000.

The preliminary Granger analyses shown in Tables 16-18 (above) suggest that

during the 19905, changes in the number of ofﬁcers and minor arrests in jurisdictions

across the nation impacted crime and that the primary effect of these variables occurred

through a one-year lag. Consequently, subsequent analyses were conducted to more

closely examine the effects of changes in ofﬁcer rates and minor arrests on each of the

139

seven felony index offenses utilizing a one-year lag with control variables included as
described in the methodology and shown in the equation in Figure 41.
Full Models Using Granger Regression

To more accurately determine the overall policing effect on crime during the
19908, ofﬁcer, drug and disorder arrest rates were simultaneously entered into regression
models. As previously reported in Table 5, ofﬁcer rates, drug arrest and disorder arrest
rates are correlated. Thus, by entering these variables into together into regression, it was
possible to determine the total amount of unique variation that was attributed to each

independent variable.

Table 19. Final Granger Regression for Murder (Full Model)

 

 

 

Murder Coef. Robust S.E. t Probability
Ofﬁcer Rate (lagged) .0048 .0026 1.86 .063
Drug Arrests (lagged) -.0394 .0209 -1.88 .060
Disorder Arrests (lagged) -.0044 .0173 -0.25 .799
Per Capita Income * -.1713 .1094 -1.57 .117
Age 15-24 % .2281 .2061 1.11 .269
Nonwhite % .1506 .1425 1.06 .291
Employment Rate * -.0079 .0031 -2.53 .012
Dependent Var. lag 1 .1175 .0260 4.51 .000
Dependent Var. lag 2 .0553 .0150 3.67 .000
R-square change (ofc., arrests) n.s.

R-square .1519

N=9,362 * multiplied by 1,000 to reduce leading zeros.

140

Table 20. Final Granger Regression for Rape (Full Model)

 

 

 

Rape Coef. Robust S.E. t ProbabilitL
Ofﬁcer Rate (lagged) -.0039 .0080 -0.49 .625
Drug Arrests (lagged) -.0828 .0523 -1.58 .113
Disorder Arrests (lagged) .0805 .0495 1.62 .105
Per Capita Income * .0870 .1029 -0.85 .398
Age 15-24 % .7870 .3235 2.43 .015
Nonwhite % -.7434 .2948 -2.52 .012
Employment Rate * -.0140 .0061 -2.30 .022
Dependent Var. lag l .1917 .0200 9.57 .000
Dependent Var. lag 2 .0244 .0117 2.08 .038
R-square change (ofc., arrests) n.s.
R-square .0962

=9,362 "‘ multiplied by 1,000 to reduce leading zeros.

Murder, Rape and Robbery

Consistent with the preliminary Granger analyses, Tables 19 and 20 (above)
indicate that changes in jurisdictional officer rates, as well as minor drug and disorder
arrests rates were not strongly associated with declines in murder or rape. Although the t-
value for drug arrests (-l .88) was marginally signiﬁcant for murder, further analyses
indicated that the effect size was so minute that meaningful interpretation was not
feasible. Additionally, the reason for the positive, but marginally signiﬁcant t-value
pertaining to the relationship between ofﬁcers (1.86) and murder is unknown. With
regard to control variables, only the employment rate (-2.53) was associated murder.
This analysis indicates that increases in employment were associated with declines in
murder. However, the effect size of this control variable is not directly interpretable in
Granger regression and is beyond the scope of this study.

With regard to the offense of rape, none of the t-values pertaining to changes in

ofﬁcers (-.49), drug arrests (-1.58) or disorder arrests (1 .62) approached statistical

141

signiﬁcance at the .05 level. With regard to control variables, this analysis indicates that
as the proportion of the population that was aged 15 to 24 increased, the number of
reported rapes also increased as indicated by the t-value 2.43. Additionally, the
proportion of population that was nonwhite (-2.52) and employment rate (-2.30) were
inversely associated with the number of reported rapes. However, the effect sizes
pertaining to control variables are not directly interpretable in Granger regression and are
beyond the scope of this study.

Table 21 (below) indicates that increases in both drug and disorder arrest rates
were associated with decreases in robbery rates. Although effect sizes will be discussed
in a subsequent section, these data indicate that minor drug arrests were more strongly
associated with robbery than disorder arrests. It is also noteworthy that when the officer
rate variable was simultaneously entered into the model with drug and disorder arrest
rates, the ofﬁcer-effect on robbery (from preliminary regressions) became non-
signiﬁcant. The number of ofﬁcers per capita was not associated with robbery when
information pertaining to drug and disorder arrests were included in the model. Lastly,
these data indicate that none of the control variables signiﬁcantly explained jurisdictional
trends in robbery during the 19905 at the .05 level. However, it was noted that per capita
income was signiﬁcantly and inversely associated with reported rates of robbery at the

.10 level.

142

Table 21 . Final Granger Regression for Robbery (Full Model)

 

 

Robbery Coef. Robust S.E. t Probability
Ofﬁcer rate (lagged) .0409 .0468 0.87 .382
Drug Arrests (lagged) -.9269 .2461 -3.77 .000
Disorder Arrests (lagged) -.5783 .2406 -2.40 .016
Per Capita Income * -3.713 2.173 -1.71 .088
Age 15-24 % 6.987 4.342 1.61 .108
Nonwhite % 2.136 2.459 0.87 .385
Employment Rate * .0034 .0406 0.08 .933
Dependent Var. lag 1 .9378 .0820 11.44 .000
Dependent Var. lag 2 -. l 791 .0793 -2.26 .024
R-square change (ofc., arrests) .0112

R-square .7564

 

N=9,362

Aggravated Assault

* multiplied by 1,000 to reduce leading zeros

Increases in the number of ofﬁcers per 10,000 citizens were associated with

reductions in aggravated assault as indicated by the t-value of -6.84 in Table 22 (below).

Additionally, increases in drug arrests were marginally signiﬁcant (-1.72). However,

changes in jurisdictional disorder arrests were not associated with annual aggravated

assault rates. Effect sizes pertaining to officers and drug arrests will be discussed in a

subsequent section. With regard to control variables, these data indicate that increases in

employment (-3.5) and per capita income (-2. 1) were associated with declines in

aggravated assault.

143

Table 22. Final Granger Regression for Aggravated Assault (Full Model)

 

 

 

Aggravated Assault Coef. Robust t Probability
S.E.

Ofﬁcer Rate (lagged) -.4922 .0720 -6.84 .000
Drug Arrests (lagged) -.7923 .4603 -1.72 .085
Disorder Arrests (lagged) .3652 .4231 0.86 .3 88
Per Capita Income * -l .982 .9424 -2.10 .035
Age 15-24 % 1.947 2.269 0.45 .656
Nonwhite % 2.712 2.269 1.20 .232
Employment Rate * -.0929 .0266 -3.50 .000
Dependent Var. lag l .6328 .0191 33.09 .000
Dependent Var. lag 2 -.0892 .0153 -5.83 .000
R-square change (ofc., arrests) .0223

R-square .4178

N=9,362 * multiplied by 1,000 to reduce leading zeros.

Burggary

Increases in the number of ofﬁcers per 10,000 citizens were associated with
reductions in burglary as indicated by the t-value of -2. 1 7 in Table 23 (below).
Additionally, increases in drug arrests were also signiﬁcantly (-2.96) associated with
declines in burglary. However, changes in jurisdictional disorder arrests were not
associated with annual burglary rates. Effect sizes pertaining to changes in ofﬁcers and
drug arrests will be discussed in a subsequent section. With regard to control variables,
these data indicate that increases in employment and per capita income were associated

with declines in burglary.

144

Table 23. Final Granger Regression for Burglary (Full Model)

 

 

 

Burglary Coef. Robust S.E. t Probability
Ofﬁcer Rate (lagged) -.3611 .1670 -2.17 .030
Drug Arrests (lagged) -3.868 1.305 -2.96 .003
Disorder Arrests (lagged) -.6397 .9202 -0.70 .487
Per Capita Income * -5.165 2.175 -2.37 .018
Age 15-24 % -3.238 8.835 -0.37 .714
Nonwhite % -6.681 4.016 -1 .66 .096
Employment Rate * .2355 .0681 -3.46 .001
Dependent Var. lag 1 .6041 .0200 30.18 .000
Dependent Var. lag 2 -.O662 .0270 -2.45 .005
R-square change (ofc., arrests) .0185

R-square .5591

=9,362 "' multiplied by 1,000 to reduce leading zeros.

Larceny

Increases in the number of ofﬁcers per 10,000 citizens were associated with

reductions in larceny as indicated by the t-value of -5.10 in Table 24 (below). However,

changes in jurisdictional drug (-1.47) and disorder arrest (-1.13) rates were not associated

with annual larceny rates. The effect size of changes in ofﬁcer rates pertaining to larceny

will be discussed in a subsequent section. With regard to control variables, these data

indicate that increases in employment (-2.59) and per capita income (-2.3 7) were

associated with declines in larceny. Additionally, the proportion of population that was

nonwhite (-2.46) and proportion of the population that was aged 15 to 24 (-1.9) were

inversely associated with larceny.

145

Table 24. Final Granger Regression for Larceny (Full Model)

 

 

 

Larceny Coef. Robust S.E. t Probability
Ofﬁcer rate (lagged) -2.52 .4944 -5.10 .000
Drug Arrests (lagged) -4.73 3.220 -1.47 .142
Disorder Arrests (lagged) -4.11 3.646 -1.13 .259
Per Capita Income * -5.17 2.175 -2.37 .018
Age 15-24 % -27.69 14.567 -1.90 .057
Nonwhite % -22.99 9.36 -2.46 .014
Employment Rate * .353 .1364 -2.59 .010
Dependent Var. lag 1 .6209 .0241 25.73 .000
Dependent Var. lag 2 -.0909 .0232 -3.92 .000
R-square change (ofc., arrests) .0229

R-square .4595

N=9,362 * multiplied by 1,000 to reduce leading zeros.

Vehicle Theft

Increases in the number of ofﬁcers per 10,000 citizens were associated with

reductions in vehicle theft as indicated by the t-value of -2.60 in Table 25 (below).

Additionally, increases in drug arrests were also marginally signiﬁcant (-1.80) and

inversely associated with vehicle theft. However, changes in jurisdictional disorder

arrests were not associated with this offense. The effect sizes of changes in ofﬁcer rates
and drug arrests pertaining to larceny will be discussed in a subsequent section. With
regard to the inﬂuence of control variables, only the proportion of population that was
young (aged 15 to 24) was associated with vehicle theft at the .05 level. These data
indicate that as the proportion of youth population increased, reports of vehicle theft also
increased. It was also noted that per capita income was marginally and inversely related
to reports of vehicle theft.

Table 25. Final Granger Regression for Vehicle Theft (Full Model)

146

 

 

Vehicle theft Coef. Robust S.E. t Probability
Ofﬁcer rate (lagged) -.4536 .1742 -2.60 .009
Drug Arrests (lagged) -2.303 1.281 -1.80 .072
Disorder Arrests (lagged) -l .133 .7046 -l .61 .108
Per Capita Income * -6.839 3.823 -1 .79 .074
Age 15-24 % 19.192 9.334 2.06 .040
Nonwhite % 3.561 4.901 0.73 .468
Employment Rate * -.1397 .0929 -1.50 .133
Dependent Var. lag l .7711 .0472 16.33 .000
Dependent Var. lag 2 -.1506 .0300 -5.01 .000

R-square change (ofc., arrests) .0157
R-square .6221

N=9,362

Final Model Effect Sizes

* multiplied by 1,000 to reduce leading zeros.

As previously discussed, effect sizes pertaining Granger analyses must be

adjusted for the inﬂuence of the lagged dependent variables. Consequently, effect sizes

were calculated using the same formula indicated in Tables 16-18. These results are

shown under the columns labeled “Total” in Tables 26-28 (below). However, to provide

some additional context or perspective with regard to the relative inﬂuence of officers

and minor arrests, the following column indicates the change that would be required to

reduce crime by 1 unit per 100,000 population. For example, in Table 26 below, the total

ofﬁcer effect on larceny was calculated to be -5.36. That is, each additional officer per

10,000 citizens was associated with a reduction of 5.36 larcenies in a jurisdiction of size

100,000. To convert this ﬁgure into individual units (i.e. 1 larceny), the formula l/total

was calculated (e. g. l/-5.36), resulting in a ﬁgure of .19 ofﬁcers per 10,000 citizens. This

equates to 1.9 ofﬁcers per 100,000 population.

According to the Granger regression procedure and these data as shown in Table

26, the number of ofﬁcers required to reduce one larceny per 100,000 citizens was 1.9.

The number of ofﬁcers required to reduce one burglary per 100,000 citizens was 13.
Additionally, 8.3 and 9.3 ofﬁcers were required to result in one less vehicle theft and
aggravated assault.

Table 26. Effect Size Calculations for Final Models of Ofﬁcers and Crime

 

 

 

Officers Lag DVl Lag 2 Total To reduce by 1
unit per 100,000

Murder -- .l 175 .0553 n.s. --

Rape -- .1917 .0244 n.s. --
Robbery -- .93 78 -.1791 n.s. --
Assault -.4922 .6328 -.0892 -l .08 9.3 ofﬁcers
Burglary -.3611 .6041 -.0662 -0.78 13 ofﬁcers
Larceny -2.520 .6209 -.0909 -5.36 1.9 ofﬁcers
Vehicle theft -.4536 .7711 -.1506 -l.20 8.3 ofﬁcers
N=9,362

Table 27 below shows the calculated effect sizes pertaining to minor drug arrests.
The same procedure followed with regard to ofﬁcer rates in the previous table was
utilized to provide additional context pertaining to the relative inﬂuence of drug arrests.
The right hand column indicates the change that would be required to reduce crime by 1
unit per 100,000 population. For example, in Table 27 below, the total effect of minor
drug arrests on burglary was calculated to be -8.37. That is, each additional arrest per
1,000 citizens was associated with a reduction of 8.37 burglaries in a city of size 100,000.

According to the Granger regression procedure and these data as reported in Table
27, the number of arrests per 1,000 required to reduce one burglary per 100,000 citizens
was .119 (or 11.9 annual arrests in a jurisdiction of 100,000 citizens. To reduce crime by
l robbery per 100,000 citizens, 26 minor drug arrests were required. The impact of drug
arrests on murder was very small such that 2,000 additional arrests were associated the

reduction of one murder per 100,000 population.

148

Table 27. Effect Size Calculations for Final Models of Drug Arrests and Crime

 

 

Drug Lag DVl Lag 2 Total To reduce by 1
unit per 100,000

Murder -.0394 .1 175 .0553 -0.05 2000 arrests
Rape -- .1917 .0244 n.s. --
Robbery -.9269 .9378 -. 1791 -3.84 26 arrests
Assault -.7923 .6328 -.0892 -1.74 57.5 arrests
Burglary -3.868 .6041 -.0662 -8.37 11.9 arrests
Larceny -- .6209 -.0909 n.s. --
Vehicle theft -- .771 l -.1506 n.s. --

 

N=9,362

Table 28 below shows the calculated effect sizes pertaining to minor disorder
arrests. The same procedure followed with regard to minor drug arrest rates in the
previous table was utilized to provide context or perspective with regard to the relative
inﬂuence of disorder arrests. According to the Granger regression procedure and these
data as reported in Table 28, the number of arrests per 1,000 required to reduce one
robbery per 100,000 citizens was .416 (or 41.6 annual arrests in a jurisdiction of 100,000
citizens. No offense types other than robbery were signiﬁcantly reduced by additional
disorder arrests when ofﬁcer rates and drug arrests rates were simultaneously entered into

the regression model.

Table 28. Effect Size Calculations for Final Models of Disorder Arrests and Crime

 

Disorder Lag DVl Lag 2 Total To reduce by 1
unit per 100,000

 

149

Murder -- .1 175 .0553 n.s. --

 

Rape -- .1917 .0244 n.s. --
Robbery -.5783 .9378 -.1791 -2.40 41.6 arrests
Assault -- .6328 -.0892 n.s. ~-
Burglary -- .6041 -.0662 n.s. --
Larceny -- .6209 -.0909 n.s. --
Vehicle theft -- .7711 -.1506 n.s. --
N=9,362

Summary of Results - Officers, Drug and Disorder Arrests - Hypotheses 1 and 2

Hypothesis 1, which asserted that increases in the number of ofﬁcers per capita
contributed to reductions in crime during the 19908 was supported for all offense
categories other than murder, rape and robbery. However, effect sizes were very small.
The largest officer effect was for the offense of larceny and indicated that nearly 2
additional officers were required to reduce one larceny per 100,000 citizens nationwide.

Hypothesis 2, which asserted that increases in arrests for minor drug and disorder
offenses contributed to reductions in crime during the 19908 was supported only for
robbery, aggravated assault and burglary with regard to drug arrests. The strongest drug
arrest effect was substantial for the offense of burglary and indicated that only 11.9
additional misdemeanor drug arrests would be needed to reduce one burglary per 100,000
citizens nationwide. However, hypothesis 2 was not supported for disorder arrests with
the exception of the offense of robbery. A total of 41 .6 additional disorder arrests were
associated with the elimination of l robbery per 100,000 citizens nationwide.

The total amount of variation in crime that was explained by changes in ofﬁcers
and minor drug and disorder arrests varied by offense category. These speciﬁc policing
changes demonstrated no effect on the offenses of murder and rape. However, their total

combined effect as indicated by the change in the value of the R-square statistic

150

suggested that slightly more than 1% of the change in robbery rates nationwide could be
attributed to changes in ofﬁcers and minor arrests. Additionally, these policing variables
were associated with 2.2% of the reduction in aggravated assault, as well as 1.9% of
burglary, 1.6% of vehicle theft, and 2.3% of the reduction in larceny rates nationwide
during the 19903.

Results Section II - Policing Practices

Analyses of changes in policing practices (i.e. tactics and strategies) other than
arrest and police force strength was conducted in two parts. The ﬁrst portion assessed
changes among broad categories of practices through the creation of summative indexes
of individual survey items pertaining to similar types of tactics. The second part assessed
the inﬂuence of changes in the reported implementation of speciﬁc individual tactics.
The National Survey of Community Policing Strategies was administered in two waves
(1993 and 1997). Both waves consisted of 80 items pertaining to the current
implementation of speciﬁc tactics and strategies by each police organization.

Reported changes in the breadth of each of the 10 policing practice indexes were
ﬁrst included in regressions of crime individually. For example, problem solving was
included as the key independent variable for murder, robbery, rape, aggravated assault,
burglary, larceny and vehicle theft. Next, after assessing the crime effects of each
summative index individually, the indexes were entered into regressions simultaneously
to determine which of them explained unique variation in crime. All regression models
included the control variables that were utilized in previous analyses as well as lagged
values of the level and changes of the dependent variable as described in the

methodology and illustrated in Figure 44.

151

Initial Models - Policing Practice Indexes

Table 29 (below) shows the beta coefﬁcients and standard errors from these
regressions. Numbers that are shown in bold print reﬂect statistically signiﬁcant
relationships with crime. These regression results suggest that robbery, burglary, larceny
and vehicle theft were each associated with broader implementation of at least one
policing practice index. However, none of the policing practice indexes signiﬁcantly
explained changes in murder, rape, or aggravated assault. Vehicle theft was inversely
associated with increases in reported citizen involvement (-16.12), collaboration efforts
by the police agency (-17.64), place or “hot spot” oriented approaches (-14.51), and
problem solving techniques that were implemented at both the organization (-22.55) and
line (~17.12) levels. Larceny and burglary also decreased as citizen involvement and
organizational level problem solving increased. Additionally, robbery decreased as the
number of reported problem and place oriented activities increased. The inﬂuence of

control variables and relative effect sizes are discussed in the following sections.

152

.EE Eon E 850% 2a 2: .v3 $86508 Hﬁocammm
.885:an E m.m “253* .wcwuz

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

938 93.: E 9 93.8 90.8 3.9 E .8
.3- 2.. _- $2.. 8. _- m2- 2 .o 8.- 356qu
:28 53: Ag 8 am 8 A88 8&8 f8
2.:- eo._~- an 8- 8 N Se- 28 8.8 3:28 .28 528;
and 84.5 85.8 $8 and 8.. 8 22.8
8.2. 3.:1 MS:- RN 2.....- emo- 2 .o- 358 .28 528;
33.8 33.8 SN 8 and God :28 A2 .8
5.:- Ri- we w- 32. ane- S o 2 .o- 8380 8e:
33: $28 as. m: 2:8 3 2 so 8 Q8
2.2- 2.3. e. 2- a o 3;- eod- mod 2.235. 8.5
33: 33.8 G” N: 892 :38 3e 8 $8
2.3 8.8 8. o- z o- 3:. So- end 3%; ,6
35.8 338 so 8 2: 8 3w 8 24.8 8N8
Se 28. a Y z e- 2.? X o Rd- 2828 355 82582.8
35.8 :28 ca 8 and a2 8 93.8 $8
8.:- SQ- 2. :- cm _- met 2 o 2 S- 358 98$ 8.528236
23.28 £38 a; o: 5.8 A8 0 pm 8 5.8
2.2- 2.2- K _- 2 m 8e- ad ome- Eeaeeeem 88
$58 838 3.3.8 are :28 A88 9.8
2.2- who..- 2.2- EN :3. S o 2 .0. 3063.25 586
~82: o_oEo> as am 2:83, bagel»— omalm SUB—2 28$-

 

 

mcxowE motes-r mEozom 28 2:10 .58 ESE Envy—Sm 98 356580 :ommmohwox .om 038-

153

Full Model - Policing Practice Indexes

After assessing the crime effects of each summative index individually, these
items were entered into a regression simultaneously to determine which indexes
explained unique variation in crime. For example, to assess effects on vehicle theft,
citizen involvement, collaboration efforts, place oriented and problem solving indexes
were entered into a regression model along with all control variables. Indexes that did
not explain a signiﬁcant amount of variation in vehicle theft were removed from ﬁirther
regressions, one at a time, beginning with the least inﬂuential index. This procedure was
followed until all remaining indexes explained variation at the .10 level or lower. The
ﬁnal results of the data reduction process pertaining to the policing practice indexes are
shown in Tables 30-33 below.

Vehicle Theft

When the relevant indexes (i.e. those that were signiﬁcant in Table 29 above)
were entered together in into the regression of vehicle theft, the two problem solving
indexes were all that remained in terms of statistical signiﬁcance. Among the 868
jurisdictions, each additional problem solving tactic that was implemented at the
organization level was associated with a decrease of nearly 17 vehicle thefts (-16.91) per
100,000 citizens (Table 30). Additionally, each new problem solving tactic that was
implemented at the line-level was associated with a decrease of 10.59 vehicle thefts.
With regard to the inﬂuence of control variables, this analysis indicated that the
percentages of the population that were nonwhite and young (i.e. aged 15 to 24) were
positively associated with reported vehicle theft. The inﬂuence of youth in the

population was quite strong, indicating that a one percentage increase in population aged

154

15 to 24 was associated with an increase of approximately 87 vehicle thefts per 100,000
citizens nationwide.

The last line of the table labeled “R-square change” reﬂects that total variation in
vehicle theft that was associated with the addition of both problem solving indexes to the
regression. These data indicate that 3.1% of the decline in vehicle thefts from 1993 to
1997 was associated with increases in the breadth of problem solving tactics by police

agencies nationwide.

Table 30. Regression Results of Vehicle Theft and Policing Practice Indexes

 

 

Vehicle Theft Coef. Robust S.E. t Probability
Problem Solving (Org.) -16.9141 6.6167 -2.556 .0107
Problem Solving (Line) -10.5966 5.6230 -1.885 .0597
Employment Rate .7752 .6150 1.260 .2078
Nonwhite % 32.3039 14.3254 2.255 .0243
Age 15-24 % 87.0767 32.6334 2.668 .0077
Per Capita Income -.0042 .0087 -.487 .6263
Vehicle 1990 Level -.3180 .0300 -10.609 .0000
Vehicle change 1990-1993 -.3839 .0985 -3.896 .0001
R-square .508

R-square change .031

 

N=868 Signiﬁcant items (p<.10) in bold print.

Larceny

The data reduction process described above was also followed with regard to
larceny. When the relevant indexes (i.e. those that were signiﬁcant in Table 29 above)
were entered together in into the regression, only the citizen involvement index remained
statistically signiﬁcant. Among the 868 jurisdictions, each newly reported citizen
involvement was associated with a decrease of nearly 50 larcenies (-49.7) per 100,000
citizens. With regard to the inﬂuence of control variables, this analysis indicated that the

rate of employment was positively associated with reported larcenies. A one unit

155

increase in per capita employment was associated with approximately four fewer
larcenies (3.99). The R-square change value also indicated that 2.5% of the decline in
larcenies from 1993 to 1997 was associated with increases in the breadth of citizen
involvement in policing. The percentages of the population that were nonwhite and
young (i.e. aged 15 to 24) were also weakly and positively associated with reported

larcenies.

Table 31. Regression Results of Larceny and Policing Practice Indexes

 

 

Larceny Coef. Robust S.E. t Probability
Citizen Involvement -49.7781 22.4911 -2.213 .0271
Employment Rate 3.9999 1.5891 2.517 .0120
Nonwhite % 51.0735 29.0574 1.758 .0791
Age 15-24 % 131.3268 75.4158 1.741 .0819
Per Capita Income -.0357 .0236 -1.512 .1308
Larceny 1990 Level -.1303 .0266 -4.893 .0000
Larceny change 1990-1993 -.1280 .0763 -1.678 .0936
R-square .134

R-square change .025

 

N=868 Signiﬁcant items (p<.10) in bold print.

Burglary

With regard to burglary, when the relevant indexes were entered together in into
the regression, only problem solving at the organization level remained statistically
signiﬁcant. Among the 868 jurisdictions, each newly reported problem solving technique
at the organization level was associated with a decrease of over 18 burglaries (-18.2) per
100,000 citizens. The R-square change value also indicated that 1.6% of the decline in
burglaries from 1993 to 1997 were associated with increases in the breadth of
organization level problem solving.

With regard to the inﬂuence of control variables, this analysis indicates that the

rate of employment was positively associated with reported burglary. A one unit increase

156

in per capita employment was associated with approximately 1.6 additional burglaries per
100,000 citizens. Additionally, the inﬂuence of youth in the population was quite strong,
indicating that a one percentage increase in population aged 15 to 24 was associated with
an increase of approximately 71 burglaries. Lastly, as per capita income increased,
reported rates of burglary decreased (-2.16).

Table 32. Regression Results of Burglary and Policing Practice Indexes

 

 

Burglary Coef. Robust S.E. t Probability
Problem Solving (Org) -18.2256 7.6982 -2.368 .0181
Employment Rate 1.5961 .6111 2.612 .0091
Nonwhite % 7.8941 9.1681 .861 .3894
Age 15-24 % 71.1785 28.2359 2.521 .0118
Per Capita Income -.0203 .0093 -2.167 .0304
Burglary 1990 Level -.2937 .0208 -14.088 .0000
Burglary change 1990-1993 -.3103 .0473 -6.565 .0000
R-square .379

R-square change .016

 

N=868 Signiﬁcant items (p<.10) in bold print.

Robbery

For robbery, when the relevant indexes were entered together in into the
regression, the two problem solving indexes were all that remained statistically
signiﬁcant. Among the 868 jurisdictions, each additional problem solving tactic that was
implemented at the organization level was associated with a decrease of nearly 7
robberies (-6.8) per 100,000 citizens (Table 33). Additionally each new problem solving
tactic that was implemented at the line-level was associated with a decrease of 4.2
robberies. The R-square change value also indicated that 2.1% of the decline in
robberies from 1993 to 1997 was associated with increases in the breadth of problem

solving at the organization and line levels.

157

With regard to the inﬂuence of control variables, this analysis indicated that the

percentages of the population that were nonwhite and young (i.e. aged 15 to 24) were

positively associated with reported robbery. A one percentage increase in population

aged 15 to 24 was associated with an increase of approximately 16 robberies. Lastly, a

one percentage increase in population that was nonwhite was associated with an increase

of approximately 21 robberies per 100,000 citizens nationwide.

Table 33. Regression Results of Robbery and Policing Practice Indexes

 

 

Robbery Coef. Robust S.E. t Probability
Problem Solving (Org) -6.8340 3.1436 -2.l74 .0299
Problem Solving (Line) -4.2292 2.2405 -l.888 .0593
Employment Rate .0268 .2565 .105 .9167
Nonwhite % 21.0976 7.1713 2.942 .0033
Age 15-24 % 16.0653 7.9073 2.032 .0424
Per Capita Income -.0049 .0050 -.983 .3260
Robbery 1990 Level -.3411 .0255 -13.388 .0000
Robbery change 1990-1993 -.4070 .0889 -4.580 .0000
R-square .757

R-square change .021

 

N=868 Signiﬁcant items Q)<.10) in bold print.

Summary of Results - Policing Indexes - Hypothesis 3

Hypothesis 3 asserted that increases in the implementation of broad categories of

Collaboration Effort, Problem Solving, Citizen Involvement, Code Enforcement,

Community Policing Policies, Crime Analysis, Place Oriented Strategies and Efﬁciency

Efforts contributed to reductions in crime during the 19903. This hypothesis was

supported only for problem solving and citizen involvement. Each additional policing

practice pertaining to problem solving reduced crime by more than 10 robberies, 18

burglaries and 27 vehicle thefts per 100,000 citizens. Additionally, each new citizen

involvement was associated with a reduction of nearly 50 larcenies per 100,000

population nationwide. None of the broad policing indexes explained changes in murder,
rape or aggravated assault.

The total amount of variation in crime that was explained by increases in problem
solving and citizen involvement varied by offense category. These broad indexes of
policing practices demonstrated no effect on the offenses of murder, rape or aggravated
assault. However, their total combined effect as indicated by the change in the value of
the R-square statistic suggested that slightly more than 2% of the change in robbery rates
nationwide could be attributed to increases in generic problem solving. Additionally,
policing practice indexes were associated with 1.6% of the reduction in burglary, as well
as 3.1% of vehicle theft, and 2.5% of the reduction in larceny rates nationwide from 1993
to 1997.

Initial Models - Individual Policing Practices

The following portion of the analyses of changes in policing practices are oriented
toward speciﬁc, individual tactics and strategies. As a ﬁrst step, each of the crime types
were regressed against each of the 80 individual survey items separately along with
control variables. The following steps involved a data reduction procedure, similar to the
one implemented with the summative indexes. Each of the items that signiﬁcantly
explained crime in the individual regressions were simultaneously included in a
regression model and the weakest non-signiﬁcant items were removed one at a time until
only signiﬁcant policing practices remained.

Table 34 (below) lists all of the individual items that were signiﬁcantly associated
with reductions in one or more of the crime rates. Of the 80 original individual policing

practices, 39 were signiﬁcantly associated with reduction of at least one crime type as

159

indicated by the value “X”. Each variable is numbered and can be cross referenced to the
complete list of tactics shown in Appendix B to obtain more speciﬁc information about
the survey question and topic. However, to facilitate an overview of the relationships,
items are listed by policing practice index category. For example, the ﬁrst ﬁve items in
the table below (items 68-73) all pertain to citizen involvement.

The table may be read from top to bottom to determine which groups of tactics
were most commonly associated with the reduction of a speciﬁc type of crime.
Additionally, the table may be read from left to right to determine which tactics were
associated with the greatest number of crime types. Reading from top to bottom in the
vehicle theft column for example, it is apparent that 9 of the 16 problem solving items
were associated with reductions that were signiﬁcant at the .10 level or lower. Moving
one column to the left, the data indicated that ﬁve of the citizen involvement items were
associated with reductions in larceny rates. Additionally, burglary and robbery were
associated with the implementation of a variety of tactics. However, murder, rape and
aggravated assault were associated with very few of the 80 tactics.

Reading from left to right, the data indicated that two tactics (#56 and #79) were
associated with reductions in the ﬁve of the seven crime types. Variable #56 pertains to
civil and regulatory code enforcement by most patrol officers in the department (see
appendix B). Variable #79 indicates that citizens prepare agreements specifying work to
be done the police pertaining to neighborhood problems. These two policing practices
were associated with reductions of most of the crime categories. It is also noteworthy
that implementation of a crime analysis function (item #43) was associated reductions in

four of the seven crime categories nationwide from 1993 to 1997.

160

Table 34. Signiﬁcant Individual Policing Practices from Initial Crime Regression Models

Index Var Murder Assault Vehicle
Citizen involv 68
Citizen involv 69
Citizen involv 70
Citizen involv 73
Civil codes 56
Collaboration 7
Collaboration 1 1
Collaboration 1 3
Collaboration 25
Collaboration 26
Collaboration 46
Collaboration 5 1
Collaboration 74
CP ' 1
Crime anal ' 43
' 8
34
E ' 36
' 53
Place 9
Place 10
Place 29
Place 30
Place 32
Place 38
Place 39
Place 40
Problem ' 14
Problem ' 15
Problem ' 22
Problem 35
Problem ' 47
Problem ' 50
Problem ' 52
Problem So 57
Problem 60
Problem ' 61
Problem 62
Problem ' 63
Problem ' 66
Problem ' 79

N=868

><><><><><><><><

 

161

Final Models - Individual Policing Practices

Table 35 (below) lists the variables that were signiﬁcantly associated with crime
declines after the data reduction process described above. Vehicle theft reduction was
associated with civil code enforcement by patrol officers (#56), foot patrol (#11),
implementation of a crime analysis function (#43) and implementation of a program
where citizens prepared work agreements for the police pertaining to neighborhood
problems (#79). Additionally, the implementation of a citizen police academy was
associated with reductions in both burglary and larceny. After the data reduction process,
the two most broadly inﬂuential variables were civil code enforcement (#56) and the
implementation of a crime analysis unit (#43). More detailed analyses pertaining to the

full models of speciﬁc offenses and crime can be found in the following sections.

162

 

ﬂows—2 coamohmod 35m - 9:638; couosvom Sun 228.4... 303an wEo=om Enouimﬁ

.mm 2an

163

Although the previous two tables are useful for obtaining an overview of
regression results, more detailed information regarding the speciﬁc effects of each
individual variable after data reduction can be found in Tables 3642 (below). Each of
these regressions included all control variables that were utilized in the previous
regressions. Additionally, consistent with the regressions of the summative indexes,
information pertaining to levels and changes in crime that occurred just prior to the ﬁrst
survey wave (i.e. prior to 1993) were also included.

Vehicle Theft

Results from Table 36 indicate that citizen prepared (problem solving) work
agreements explained the largest portion of unique variation in vehicle theft when
combined with other inﬂuential tactics. Agencies that began a citizen work agreement
procedure after 1993 experienced 83 fewer vehicle thefts per 100,000 population across
jurisdictions nationwide. Implementation of foot patrol as a speciﬁc assignment for
officers was associated with a reduction of over 61 vehicle thefts per 100,000 population
in addition to the reductions from crime analysis (-30.7) and civil code enforcement (-
43.8). Finally, with regard to the inﬂuence of control variables it is noteworthy that an
increase in youth population (individuals aged 15-24) was associated with increases in
vehicle theft. Similar to previous regressions, each additional percentage increase in
population aged 15 to 24 was associated with an increase of approximately 75 vehicle
thefts per 100,000 citizens nationwide.

The last line of the table labeled “R-square change” reﬂects the total variation in
vehicle theft that was associated with the addition of the individual policing practices to

the regression. These data indicate that 6.5% of the decline in vehicle thefts from 1993 to

164

1997 was associated with the implementation of citizen work agreements, foot patrol,

crime analysis and civil code enforcement among surveyed police agencies.

Table 36. Regression Results of Vehicle Theft and Individual Policing Practices - Final

 

 

Model

Vehicle Theft Coef Robust S.E. t Prob
V11. Foot Patrol -61.8012 26.8520 -2.302 .0215
V43. Crime Analysis Unit -30.6792 16.1624 -l.898 .0579
V56. Ofc. Enforce Civil Codes -43.8294 24.8125 -1.766 .0776
V79. Citizens contract ofﬁcer work -83.2962 28.2191 -2.952 .0032
Employment Rate .6402 .5932 1.079 .2807
Nonwhite % 23.7746 14.0371 1.694 .0906
Age 15-24 % 75.4629 29.2946 2.576 .0101
Per Capita Income -.0051 .0088 -.585 .5590
Vehicle Theft (Rate 1990) -.3171 .0278 -1 1.423 .0000
Vehicle Change (90-93) -.3979 .0940 -4.231 .0000
R-Square .542

R-Square Change .065

 

=868 Signiﬁcant items (p<.10) in bold print.

Larceny

Although civil code enforcement explained the largest portion of unique variation

in larceny rates (-204.5) when this variable was combined with other inﬂuential tactics,

the results listed in Table 37 below clearly indicated that citizen involvement was a

central factor pertaining to the reduction of larceny rates. After the data reduction, three

of the ﬁve remaining tactics fell into this category. These data suggest that agencies that

began allowing citizens to participate in the ofﬁcer hiring process, implemented a citizens

police academy and initiated citizen review of ofﬁcer complaints experienced over 425

fewer larcenies per 100,000 population on average. Implementation of a drug task force

was also associated with reduction in larcenies (-151.1). Overall, these variables

explained a substantial 7.7% of the reduction in larceny rates between 1993 and 1997.

165

With regard to the inﬂuence of control variables, this analysis indicated that the
rate of employment was positively associated with reported larcenies. Similar to the
previous regression of the policing indexes, a one unit increase in per capita employment
was associated with approximately four fewer larcenies (3.97). Lastly, each percentage
increase in nonwhite population was associated with an increase of more than 52

larcenies.

Table 37. Regression Results of Larceny and Individual Policing Practices - Final Model.

 

 

 

Larceny Coef Robust S.E. t Prob
V36. Drug Task Force ~151.l711 79.8267 -1.894 .0585
V56. Ofc. Enforce Civil Codes -204.4458 53.4279 -3.827 .0001
V69. Citz. Review Complaints -l44.3570 64.0883 -2.253 .0245
V70. Citizens Hire Officers -149.7331 73.8271 -2.028 .0428
V73. Citizen Police Academy -132.3151 58.3981 -2.266 .0236
Employment Rate 3.9730 1.5608 2.546 .0110
Nonwhite % 52.5942 25.3104 2.078 .0379
Age 15-24 % 77.8352 69.5491 1.119 .2633
Per Capita Income -.0428 .0226 -1.898 .0580
Larceny (Rate 1990) -.1213 .0255 -4.753 .0000
Larceny Change (90-93) -.1195 .0736 -1.624 .1046
R-Square .186

R-Square Change .077

N=868 Signiﬁcant items (p<.10) in bold print.

Lam

Citizen prepared (problem solving) work agreements and assignment of ofﬁcers
to ﬁxed patrol areas explained substantial portions of unique variation with regard to
burglary rates when these variables were combined with other inﬂuential tactics (Table
38). Agencies that began a citizen work agreement procedure after 1993 experienced
48.7 fewer burglaries per 100,000 population across jurisdictions nationwide.
Implementation of a crime analysis unit was associated with a reduction of approximately

46 fewer burglaries per 100,000 population in addition to the reductions ﬁ'om drug task

166

force implementation (-51.2) and civil code enforcement (-35.7). Additionally,
implementation of a citizens academy (-43.2) and ﬁxed patrol assignment (-46.0) were
also associated with burglary reduction. Overall, these variables explained 4.2% of the
reduction in larceny rates between 1993 and 1997.

With regard to the inﬂuence of control variables it is noteworthy that an increase
in youth population (individuals aged 15-24) and the rate of employment were associated
with increases in burglary theft. These positive relationships may lend support to routine
activities theory, which asserts that the number of motivated offenders and suitable
targets that are away from home are both increased when employment rates and youth
population are high.

Table 38. Regression Results of Burglary and Individual Policing Practices - Final
Model.

 

 

Burglary Coef Robust S.E. t Prob
V9. Fixed Patrol Assign. -46.0085 27.4239 -1.678 .0937
V36. Drug Task Force -51.2124 30.5258 -1.678 .0937
V43. Crime Analysis Unit -43.9837 23.1268 -1.902 .0574
V56. Ofc. Enforce Civil Codes -35.6881 18.4091 -1.939 .0528
V73. Citizen Police Academy -43.2390 23.8905 -1.810 .0706
V79. Citizens contract ofﬁcer work -48.7178 22.8665 -2.131 .0333
Employment Rate 1.5777 .5892 2.678 .0075
Nonwhite % 6.8774 8.3344 .825 .4094
Age 15-24 % 63.5226 25.6280 2.479 .0133
Per Capita Income -.0229 .0086 -2.656 .0080
Burglary (Rate 1990) -.2961 .0209 -14.160 .0000
Burglary Change (90-93) -.2902 .0472 -6. 147 .0000
R-Square .406

R-Square Change .042

 

N=868 Signiﬁcant items (p<.10) in bold print.

Aggravated Assault

Only three of the 80 tactics demonstrated any signiﬁcant relationship with regard

to reductions in rates of aggravated assault among jurisdictions across the nation after the

167

data reduction procedure. Agencies where most ofﬁcers began making efforts to assist

communities in organizing experienced fewer aggravated assaults (~48.5) as did agencies

that implemented mobile, neighborhood based ofﬁces or stations (~39.5). In this regard,

it is noteworthy that implementation of mobile ~~ but not permanent ~~ neighborhood

based stations was associated with reduction in aggravated assaults. Finally, jurisdictions

that implemented a drug or crime tip telephone line also experienced fewer aggravated

assaults (~51.3). Overall, these variables explained a substantial 2.3% of the reduction in

larceny rates between 1993 and 1997.

Table 39. Regression Results of Agg. Assault and Individual Policing Practices ~ Final

 

 

Model.

Aggravated Assualt Coef Robust S.E. t Prob
V8. Drug/Crime Tip Line ~51.2594 28.1025 ~1.824 .0684
V29. Neighborhood Stations ~39.5115 16.8125 ~2.350 .0189
V51. Assist Comm. Organization ~48.5381 23.1879 ~2.093 .0365
Employment Rate ~.4470 .4463 ~1 .002 .3167
Nonwhite % ~1.2983 6.6888 ~.194 .8461
Age 15-24 % 27.8880 13.5814 2.053 .0402
Per Capita Income .0047 .0063 .745 .4565
Agg. Assault (Rate 1990) ~.2632 .0308 8545 .0000
Agg. Assault Change (90-93) ~.5047 .0995 ~5.074 .0000
R-Square .412

R-Square Change .023

 

=868 Signiﬁcant items (p<.10) in bold print.

Robbeﬂ

Six of the 80 variables were associated with signiﬁcant reductions in robbery rates

between 1993 and 1997 after the data reduction procedure. Among these variables, the

crime reduction effects were fairly evenly distributed. However, it is noteworthy that one

half of these signiﬁcant policing practices pertained to problem-solving, even though

these items were highly correlated. Citizen prepared problem-solving work agreements

168

were associated with 31 fewer robberies (~31.3). Additionally, implementation of
interagency problem-solving (~17.5) and the ability of middle managers to redesign the
organization to support problem-solving efforts (~12.0) were also associated with
substantial robbery rate reductions. The impact of civil code enforcement by patrol

officers (~23.0) and implementation of a crime analysis unit (~16.4) were also signiﬁcant.

Overall, these variables explained a substantial 5.7% of the reduction in robbery rates

between 1993 and 1997.

With regard to control variables, only the portion of nonwhite population was

signiﬁcantly associated with rates of robbery. A one percent increase in nonwhite

population was associated with an increase in approximately 19 robberies per 100,000

citizens nationwide. Employment and youth population demographics were not

associated with robbery rates during the 1993 to 1997 period among the 868 sampled

agencies.

Table 40. Regression Results of Robbery and Individual Policing Practices - Final Model.

 

 

Robbery Coef Robust S.E. t Prob
V22. Interagency Problem Solving ~17.4729 8.1307 ~2.149 .0318
V43. Crime Analysis Unit ~16.3900 7.9373 ~2.065 .0391
V56. Ofc. Enforce Civil Codes ~23.0228 8.9490 ~2.573 .0102
V57. Redesign Organization ~12.0327 6.5065 ~1.850 .0646
V70. Citizens Hire Officers ~22.8167 11.9502 ~1.909 .0565
V79. Citizens contract ofﬁcer work ~31.2793 10.5010 ~2.979 .0030
Employment Rate .0860 .1961 .438 .6612
Nonwhite % 19.3200 5.3043 3.642 .0003
Age 15-24 % 8.6368 6.9083 1.250 .2115
Per Capita Income ~.0059 .0043 ~1.364 .1729
Robbery (Rate 1990) ~.3285 .0200 ~16.401 .0000
Robbery Change (90-93) ~.3541 .0844 ~4.196 .0000
R-Square .794

R-Square Change .057

 

N=868 Signiﬁcant items (p<.10) in bold print.

169

Rag

Interestingly, all three of the policing practices that demonstrated a signiﬁcant
relationship with regard to reported declines in rates of rape (Table 41) were oriented
toward coordination with external organizations and individuals. Implementation of
multidisciplinary teams to deal with special problems was associated with 2.6 fewer
reported rapes per 100,000 citizens. Additionally, implementation of interagency
problem-solving (~2.50) and working with other agencies (~2.6l) were also associated
with signiﬁcant rape reduction. These relationships are particularly noteworthy given
that rates of reported rape did not decline among most agencies and actually increased
among smaller jurisdictions during the late 19905.

With regard to the inﬂuence of control variables, this analysis indicated that the
rate of employment was positively associated with reported rape. A one unit increase in
per capita employment was associated with approximately 0.1 additional rapes per
100,000 citizens. Lastly, as per capita income increased, reported rates of rape decreased
(~0.001). Overall, the policing variables explained 1.5% of the reduction in rape rates

between 1993 and 1997.

170

Table 41. Regression Results of Rape and Individual Policing Practices ~ Final Model.

 

 

Rape Coef Robust S.E. t Prob
V22. Interagency Problem Solving ~2.4959 1.2593 ~l.982 .0477
V35. Multidisc. Problem Teams ~2.6124 1.0911 ~2.394 .0168
V47. Work With Other Agencies ~1.9455 .9595 ~2.028 .0428
Employment Rate .0907 .0325 2.794 .0053
Nonwhite % .7303 .7094 1.029 .3035
Age 15-24 % .1196 1.4808 .081 .9357
Per Capita Income ~.0016 .0006 ~2.709 .0068
Rape (Rate 1990) ~.3095 .0402 -7.699 .0000
Rape Change (90-93) ~.6755 .0656 ~10.293 .0000
R-Square .464

R-Square Change .015

 

N=868 Signiﬁcant items (p<.10) in bold print.

mm

Although none of the summative policing indexes were associated with
signiﬁcant reductions in murder rates among jurisdictions across the nation, two of the 80
individual policing practice variables were associated with small but signiﬁcant
reductions in this offense. Implementation of mobile, neighborhood-based police stations
was associated with a reduction of about one murder annually per 100,000 population
(~.95). In this regard it is noteworthy that implementation of mobile ~~ but not permanent
~~ neighborhood based stations was associated with murder reduction. The
implementation of mobile neighborhood based stations was also associated with
reduction in aggravated assault, a closely related criminal offense that often leads to
murder. Additionally implementation of a crime analysis unit also demonstrated a small
but signiﬁcant statistical effect (~.80).

With regard to the inﬂuence of control variables, only youth population was
signiﬁcantly related to rates of murder. A one percentage increase in population aged 15

to 24 was associated with a approximately 1 additional murder per 100,000 citizens.

171

Overall, the policing variables explained 1.0% of the reduction in murder rates between

1993 and 1997.

Table 42. Regression Results of Murder and Individual Policing Practices ~ Final Model.

 

 

Murder Coef Robust S.E. t Prob
V29. Neighborhood Stations ~.9475 .4631 ~2.046 .0410
V43. Crime Analysis Unit ~.7850 .4353 ~1.803 .0716
Employment Rate ~.0106 .01 10 ~.959 .3380
Nonwhite % .1913 .2308 .829 .4073
Age 15-24 % 1.0059 .4691 2.144 .0322
Per Capita Income -.0001 .0002 ~.305 .7606
Murder (Rate 1990) ~.3122 .0359 8707 .0000
Murder Change (90—93) ~.5824 .0509 ~1 1.454 .0000
R-Square .602

R-StLuare Change .010

 

N=868 Signiﬁcant items (p<.10) in bold print.

Summag of Results - Individual Policing Tactics ~ Hymthesis 4

Hypothesis 4 asserts that increases in the implementation of speciﬁc tactics and
strategies pertaining to Collaboration Efforts, Problem Solving, Citizen Involvement,
Code Enforcement, Community Policing Policies, Crime Analysis, Place Oriented
Strategies and Efﬁciency Efforts contributed to reductions in crime. This hypothesis was
fully supported. At least one individual policing tactic within each of the broader
categories was associated with reductions in all seven of the offense types.

With regard to the size of the relative effects that were associated with speciﬁc
policing practices, the inﬂuences of individual items were typically equal to or greater
than any of the summative indexes or changes in officer rates. For example, the
implementation of citizen prepared work agreements was associated with a decline of
more than 31 robberies and 48 burglaries. In contrast, however, the combined effect of

both problem solving indexes was associated with only 11 fewer robberies and 18 fewer

172

burglaries. Further, in order to obtain an equivalent reduction in robbery, 806 minor drug
arrests, or 1,426 minor disorder arrests would be required. In order to obtain an
equivalent reduction in burglary, 576 minor drug arrests, or additional 624 officers would
be required, based on Granger analyses.

These variations in crimes that were associated with speciﬁc, individual policing
practices also suggested that a variety of policing tactics resulted in equivalent crime
effects from a statistical standpoint. For example, implementation of a crime analysis
function was associated with 44 fewer burglaries and more than 16 fewer robberies.
Additionally, enforcement of civil and regulatory code violations by most patrol ofﬁcers
was associated with 44 fewer burglaries and 23 fewer robberies. These effect sizes
reﬂect crime changes over a four year period and are not inordinately high in terms of
credibility or believability. Furthermore, the relative effects of individual policing
practices were not substantially altered by the inclusion of variables pertaining to ofﬁcer
and minor arrest rates (Tables 43-46). They were also robust to the removal of large and
small agencies from regression models (Tables 45-46).

The total amount of variation in crime that was explained by speciﬁc policing
practices varied by offense category. In contrast to the summative indexes, these speciﬁc
policing practices were associated with reductions of all seven offenses, including
murder, rape and aggravated assault. The total combined effect of individual items after
the data reduction procedure as indicated by the change in the value of the R-square
statistic suggested that 5.7% of the change in robbery rates nationwide was associated
with their implementation. Additionally, these individual policing practices were

associated with 4.2% of the reduction in burglary, 6.5% of vehicle theft, 7.7% of larceny,

173

2.3% of aggravated assault, 1.5% of rape and 1.0% of murder rates nationwide from 1993
to 1997.

Perhaps equally important, however, these analyses of individual tactics
facilitated further assessment of the nature or types practices that were most associated
with each offense. For example, Table 34 clearly suggests that the vast majority of
problem solving items were associated with reduction in vehicle theft. Nevertheless,
although using the index categories to analyze policing practices is useful, it relies upon
a priori categorization of tactics and strategies. Analyses of patterns or trends among
policing practices that were associated with speciﬁc offenses may also include more
complicated constructs, such as collective efﬁcacy or fear reduction techniques.
Consequently, the discussion that follows pertaining to individual policing practices and
crime will include observations regarding the nature of practices that were associated
with each type of offense without reference to the a priori index categories.

Robustness Checks

Previous analyses indicated that changes in the number of ofﬁcers per capita, as
well as drug and disorder arrests rates were not signiﬁcantly correlated with changes in
problem-solving, citizen involvement, or community collaboration index measures.
Thus, their inclusion in regression models should not substantially alter relationships
pertaining to policing practices and crime. However, to test this assertion, changes in
officer rates, along with drug and disorder arrests rates were included in regressions of
both the summative policing indexes and individual policing practice items.

Table 43 below indicates that the number of ofﬁcers per capita, drug arrest and

disorder arrests rates were not signiﬁcantly associated with robbery reduction in the two-

174

wave National Survey of Community Policing Strategies sample. More importantly
however, as expected, the inclusion of these variables did not substantially alter the
inﬂuence of the problem-solving index. Thus, the problem-solving index was robust to
the inclusion of these additional policing variables.

Table 43. Robustness Check - Adding Ofﬁcers, Arrests to Policing Index Regression
of Robbery

 

 

Robbery Coef Robust S.E. t Prob
Ofﬁcer Rate .7446 .9691 .768 .4424
Disorder Arrest Rate 1.3523 2.6339 .513 .6077
Drug Arrest Rate .4265 1.0914 .391 .6960
Problem Solving ~9.0777 3.3421 ~2.716 .0067
Employment Rate ~.0168 .2512 ~.067 .9467
Nonwhite % 19.7172 6.9494 2.837 .0046
Age 15-24 % 17.7498 7.5466 2.352 .0188
Per Capita Income ~.0078 .0043 ~l.837 .0664
Robbery Rate (1990) ~.3854 .0268 ~14.400 .0000
Robbery Change (90-93) ~.4679 .0754 -6.209 .0000

 

N=868 Signiﬁcant items (p<.10) in bold print.

As indicated in Table 44 below, the number of ofﬁcers per capita, drug arrest and
disorder arrests rates were also added to a regression of individual policing practices. The
number of ofﬁcers per capita, drug arrest and disorder arrests rates were not signiﬁcantly
associated with larceny reduction in the two-wave National Survey of Community
Policing Strategies sample. Additionally, as expected, the inclusion of these variables did
not substantially alter the inﬂuence of the individual policing practice items. The
implementation of a drug task force, civil code enforcement by officers, and citizen
involvement variables remained signiﬁcant and robust to the inclusion of these additional

policing variables.

175

Table 44. Robustness Check ~ Adding Officers, Arrests to Individual Policing Practice
Regression of Larceny

 

 

Larceny Coef Robust S.E. t Prob
Ofﬁcer Rate 14.9596 8.9333 1.675 .0942
Disorder Arrest Rate ~7.4842 7.2922 ~1.026 .3049
Drug Arrest Rate 10.3822 6.9901 1.485 .1377
V36. Drug Task Force ~210.5028 90.8077 ~2.318 .0206
V56. Ofc. Enforce Civil Codes ~211.3682 57.5699 ~3.672 .0002
V70. Citizens Hire Ofﬁcers ~153.5360 85.1993 ~l.802 .0717
V73. Citizen Police Academy ~129.5309 63.0498 ~2.054 .0401
Employment Rate 3.9916 1 .6746 2.384 .0173
Nonwhite % 49.8792 26.2889 1.897 .0580
Age 15-24 % 141.1548 73.0338 1.933 .0534
Per Capita Income ~.0315 .0230 ~l .373 .1697
Larceny Rate (1990) -. 1036 .0223 ~4.640 .0000
Larceny Change (90-93) ~.1 103 .0731 ~1.509 .1314

 

N=868 Signiﬁcant items (p<.10) in bold print.

To ensure that the regressions of policing practices were not driven solely by
large or small agencies (or sample size), additional robustness checks shown in Tables 45
and 46 below were conducted by removing all responding jurisdictions from the largest
and smallest population categories. That is, all jurisdictions with over 250,000
population were removed and all jurisdictions with populations of smaller than 25,000
were removed from the analysis. Regression of the problem-solving and citizen
involvement of indexes were again conducted to determine whether the relationship
between these variables and crime would substantially change. Tables 45 and 46 below
clearly indicate that both problem-solving and citizen involvement indexes were robust to

the removal of large and small agencies as well as the resulting decline in sample size.

176

Table 45. Robustness Check ~ Removing Agencies With Over 250,000 and Less

Than 25,000 Population ~ Regression of Robbery and Problem Solving Index

 

 

Robbery Coef Robust S.E. t Prob
Problem Solving ~18.4168 5.6658 ~3.251 .0015
Employment Rate ~.3173 .9541 ~.333 .7400
Nonwhite % 19.7009 11.7821 1.672 .0970
Age 15-24 % 69.2044 21.9159 3.158 .0020
Per Capita Income .0003 .0101 .027 .9784
Robbery Rate (1990) ~.3373 .0347 ~9.712 .0000
Robbery Change (90-93) -.3758 .141 l ~2.664 .0087

 

N=567 Signiﬁcant items (p<.10) in bold print.

Table 46. Robustness Check ~ Removing Agencies With Over 250,000 and Less
Than 25,000 Population ~ Regression of Larceny and Citizen Involvement Index

 

 

Larceny Coef Robust S.E. t Prob
Citizen lnvolvernent ~75.0950 32.7389 ~2.294 .0235
Employment Rate 8.1626 5.2101 1.567 .1197
Nonwhite % 55.2938 53.5163 1.033 .3035
Age 15-24 % 358.7265 199.8768 1.795 .0751
Per Capita Income -.0674 .0619 ~1.089 .2783
Larceny Rate (1990) ~.1735 .0574 ~3.021 .0030
Larceny Change (90-93) ~.0857 .1679 ~.510 .6107

 

N=567 Signiﬁcant items (p<.10) in bold print.

177

CHAPTER 6
DISCUSSION
Officers and Crime ~ Connections to Prior Empirical Findings

After an extensive review of studies conducted prior to 1995, Eck and Maguire
(2000) concluded that there was little evidence showing reductions in crime rates due to
marginal increases in officers. However, other researchers have noted that most prior
analyses failed to utilize methodologies to address the reciprocal nature of police force
strength and crime rates (Kovandzic & Sloan, 2002; Marvel & Moody, 1996). Recent
research suggests that the number of ofﬁcers per capita increased nationwide during the
19903 (Roth et al., 2000) and three recent analyses have found support for the
relationship between the number of ofﬁcers within jurisdictions and crime rates
(Kovandzic & Sloan, 2002; Marvell & Moody, 1996; Levitt, 1997). Nevertheless, prior
analyses of the effects of ofﬁcers on crime included only large jurisdictions or regional
subsamples. The analyses in this study include jurisdictions of all sizes from all regions
of the nation and indicate that the nationwide average ofﬁcer rate increased by more than
16 percent from 1990 to 2000. Moreover, these data are generally supportive of
assertions that increases in the number of police ofﬁcers per capita were associated with
crime reduction.

However, when jurisdictions of all sizes were included in analyses, the "ofﬁcer
effects" '2 for the nation as a whole were somewhat smaller than in previous studies. For
example, Marvell and Moody (1996) analyzed data from cities with over 250,000

population, ﬁnding that the addition each new ofﬁcer resulted in a reduction of 5.3

 

'2 The term “effect” as utilized here refers to the strength of the beta coefﬁcient and is not intended to
suggest that a cause-effect relationship has been established.

178

burglaries, 12.5 larcenies and 4.5 vehicle thefts. In contrast, these more comprehensive
data indicated that each additional officer per 10,000 citizens resulted in a reduction of
2.5 burglaries, 8.1 larcenies, and 2.9 vehicle thefts for each 100,000 citizens nationwide
when no other explanatory variables were included in regression models.

The types of crime offenses that were found to be signiﬁcantly impacted by
increases in officers in previous research have varied substantially. Marvell and Moody
(1996) found signiﬁcant effects for all index offenses other than rape. Levitt (1997)
primarily focused on generic violent and property indexes but included robustness checks
indicating that vehicle theft, larceny, aggravated assault, and rape were not consistently
reduced. Finally, in a county level analysis of Florida jurisdictions, Kovandzic and Sloan
(2002) found signiﬁcant ofﬁcer effects for burglary, larceny, and robbery but not murder,
rape or aggravated assault. Thus, with the exception of rape, which has not demonstrated
statistical signiﬁcance in any of these studies, ﬁndings have been more consistent with
regard to property crime than violent offenses. These nationwide data support previous
ﬁndings regarding the ofﬁcer related crime reducing effects of each of the property
offense categories. Additionally, consistent with previous ﬁndings no ofﬁcer effect was
found with regard to rape.

A noteworthy critique of these recent studies, however, is that none of them
included variables pertaining to rival explanations concerning what it is that the police
actually do or the tactics that are employed (Skogan & Frydl, 2004). An objective of this
study, therefore, was to address this concern by entering minor drug and disorder arrests
as well as other policing practices into models along with changes in ofﬁcer force

strength. When information pertaining to minor drug and disorder arrests was included,

179

the officer effect was further reduced for all categories and the apparent relationship
between ofﬁcer increases and robbery rates was eliminated. More speciﬁcally, the
decline in robbery rates nationwide was more closely associated with minor drug and
disorder arrest rates than changes in ofﬁcers. This ﬁnding is of particular interest given
that all three recent studies have indicated a signiﬁcant relationship with regard to
robbery. This ﬁnding also lends support to assertions by criminal justice researchers
(N agin, 1998; Skogan & Frydl, 2004) that apparent crime reducing effects may in fact
depend on the actions or assertiveness of ofﬁcers.

Final Granger regression models that included minor drug and disorder arrests
along with control variables indicated that each additional ofﬁcer per 10,000 citizens
resulted in a reduction of less than one burglary (.78), 5.4 larcenies, 1.2 vehicle thefts,
and 1.1 aggravated assaults per 100,000 citizens. Stated another way, in order to reduce
one burglary per 100,000 citizens, 13 additional ofﬁcers were required. A reduction of
one larceny required nearly two additional ofﬁcers, and the reduction of one vehicle theft
or aggravated assault required more than eight additional ofﬁcers. Consequently, the
amount of crime reduction nationwide due to increases in ofﬁcers during the 1990s was
small. Nevertheless, these data also suggest that a portion of the “ofﬁcer effect”
originated through increases in minor drug and disorder arrests that coincided with
increases in ofﬁcers.

Increases in the number of ofﬁcers per capita within jurisdictions were correlated
with increases minor drug and disorder arrests. The correlation between officer rate and
minor drug arrests was quite high (.60). This ﬁnding provides some support for

assertions that sustainable anti-drug crackdowns may require increases force strength

180

(Scott, 2004; Sherman & Weisburd, 1995; Weisburd & Green, 1995). However, it is also
a noteworthy that the correlation between ofﬁcer rates and disorder arrests was much
weaker (.19), suggesting that increased enforcement of minor disorder offenses may not
necessarily require a substantial increase in ofﬁcer strength.

Drug Arrests ~ Connections to Prior Empirical Findings

One recent econometric analysis of policing data from the state of Florida during
the 19808 suggested that increased police efforts toward drug enforcement resulted in a
counterbalancing increase in property offending as a result of the drain on police
resources (Benson, Sebastian & Rasmussen, 2001). However, these data from agencies
across the nation suggest that increases in minor drug arrests were not signiﬁcantly
associated with larceny or vehicle theft and were inversely related to burglary, robbery,
and aggravated assault. That is to say, increases in arrests for minor drug offenses were
associated with decreases in burglary, robbery and aggravated assault. Thus, these data
do not support the assertion that increased attention to minor drug offenses necessitate an
increase in property offending. It is also important to note, however, that substantial
federal funding for additional ofﬁcers occurred during the 1990s and this may have offset
the potential resource drain that was asserted during the 1980s.

Among the 9,362 reporting agencies during the 19905, the mean number of minor
drug arrests per 1,000 citizens increased by more than 106 percent. Based on the
nationwide average, Granger regression analyses in this study suggested that in a city of
size 100,000 population, a reduction of one burglary was associated with an increase of
only 12 minor drug arrests and that a reduction of one robbery was associated with an

increase of 26 minor drug arrests. These effect sizes are substantial relative to the annual

181

arrest count of approximately 485 to 1,000 in jurisdictions of this size during the 19905.
Stated another way, an increase of a single drug arrest per month jurisdiction-wide was
associated with the elimination of one burglary and an increase of slightly more than two
minor drug arrests per month was associated with the reduction of one robbery.
Additionally, minor drug arrests were weakly associated with aggravated assault such
that increases of around ﬁve arrests per month were associated with the reduction of one
offense.

This ﬁnding pertaining to sustained (i.e. annual) increases in arrests for minor
drug offenses stands in stark contrast to a different approach ~~ drug raids—~ which involve
intensive short-term policing efforts that have recently been listed in the “what doesn't
wor ” category of a prominent report to the National Institute of Justice (Sherman et al.,
1998). Targeted, short-term raids on drug areas have typically failed to produce
sustainable reductions in violent crime or disorder (Sherman et al., 1998).

Nevertheless, continuous or longer-term enforcement of drug offenses has been
associated with the reduction of violent crime and disorder. An analysis of the rise and
decline of drugs and violence in New York City over a forty-year period found that
strong law enforcement and increasing intolerance of drug offenses were associated with
signiﬁcant reductions in violence (Golub, Johonson & Dunlap, 2000). Additionally, in
Oakland California the implementation of a Specialized Multi-Agency Response Team
(SMART) that coordinated with other city agencies to inspect and identify drug/problem
properties resulted in reduced calls regarding drug offenses. Finally, a randomized
experiment that involved a drug law enforcement strategy was conducted in 56 "hot

spots" of drug activity in Jersey City, NJ and a seven-month follow-up indicated that calls

182

pertaining to disorder and emergency police services were reduced (Wiesburd & Green,
1995). Researchers from both of these studies also reported a diffusion of beneﬁts effect
to neighboring areas that were not speciﬁcally targeted (Green, 1996; Wiesburd & Green,
1995).

Disorder Arrests ~ Connections to Prior Empirical Findings

Recent literature suggests that the popularity of community policing increased
substantially during the 1990s (Roth et al., 2000). Further, given that an integral part of
the community policing concept includes an increased focus on issues of “quality of life”
and minor but routine disorder, one would expect to ﬁnd a corresponding increase in
arrests for minor misdemeanor offenses among responding agencies. However, contrary
to this expectation, these data clearly indicate that arrests for minor disorder declined
substantially, nationwide. A summative index of minor disorder arrests was created by
combining curfew, disorderly conduct, drunkenness, prostitution, runaway and vagrancy
arrests. Nationwide, among the 9,362 jurisdictions that reported data, a 22 percent
overall decline in the disorder arrest index occurred between 1990 and 2000.

Despite this decline in enforcement activity, Granger regression analyses
indicated an inverse relationship between minor disorder arrests and burglary, vehicle
theft and robbery when modeled without covariates. Nationwide, jurisdictions that
decreased their rates of arrests for minor disorder offenses experienced higher levels of
burglary, vehicle theft and robbery when compared with jurisdictions where disorder
enforcement remained stable or increased. Although this ﬁnding lends some support to
advocates of the "broken windows perspective," the relationship between disorder arrests

and burglary and vehicle theft was weak. When the disorder arrest variable was entered

183

into models along with changes in drug arrest and ofﬁcer rates, only the relationship with
robbery retained signiﬁcance.

These data indicate that, on average, an increase of 3.5 minor disorder arrests per
month (or 41.6 arrests per year) was associated with a decline in one robbery per 100,000
citizens. Given that the average jurisdiction of 100,000 population reported 1,041
disorder arrests in 1990, the effects of an additional 41.6 annual arrests are fairly
substantial. These ﬁndings are consistent with those from a panel data assessment among
75 large precincts in New York City during the 19903, which found a highly signiﬁcant
relationship between misdemeanor arrests and violent crime (Kelling & Sousa, 2001).
Moreover, with regard to the offenses of robbery and burglary, the ﬁndings from this
analysis are strikingly similar to those in two other studies.

Wilson and Boland (1978), found that cities with a high ratio of trafﬁc citations
per officer (a proxy measure of a legalistic or aggressive policing approach) had lower
levels of robbery. Approximately, ten years later, Sampson and Cohen (1988) replicated
and extended this work in a study of 171 cities with population greater than 100,000.
Their study utilized the number of arrests per ofﬁcer for disorderly conduct and drunk
driving as a proxy for aggressive policing. This measure of aggressive policing was
employed as an instrumental variable to identify the crime function and statistical results
supported their assertion of an aggregate-level effect with regard to robbery.
Additionally, consistent with the Granger analyses in this current study, Sampson and
Cohen found weak effects with regard to burglary.

In recent years an intense debate among criminal justice researchers has emerged

with regard to the ethics and utility of increased enforcement of minor disorder

184

infractions. In this regard, it has been asserted that during the 19903 a number of large
police agencies implemented so-called “zero-tolerance” strategies along with community
policing practices (Green, 1999; Sherman et al., 1998). New York City Police, for
example, targeted grafﬁti artists, public drunkenness and other disorderly activities such
as panhandling and public urination (Green, 1999; Kelling & Sousa, 2001). After an in-
depth analysis of New York’s order maintenance tactics during the 19903, some
researchers concluded that this policing approach had a direct effect on crime reduction
(Kelling & Sousa, 2001; Sousa, 2003). However, Green (1999) argued that the number
of citizen complaints against police officers in New York increased substantially during
this period and that the problem oriented policing approach was better for citizens and
more efﬁcient than aggressive policing of disorder. .

Nevertheless, given the substantial nationwide decline in minor disorder arrest
rates during the 19905, concerns that "zero-tolerance" policies may supplant community
policing practices appear to be unfounded. However, it is also noteworthy that the
nationwide decline in minor disorder arrests was accompanied by a substantial increase in
minor drug arrests. At present, it is unknown whether these trends are indicative of a
substitution effect (from enforcement of general disorder toward drug offenses) or these
changes are being driven by external factors. It was noted, however, that despite the
different trend trajectories, both levels and changes of minor drug and disorder arrest
rates were positively correlated during the 19908. Thus, if these trends were the result of
a substitution effect, a negative or inverse relationship might have been expected.

These data indicate that the combined or net total of arrests for minor offenses

nationwide increased. The comparative efﬁcacy of arrests for minor offenses versus

185

problem-solving, community collaboration, or other community policing practices will be
discussed in the following sections pertaining to policing practices such as problem
solving and crime analysis. However, to inform that comparison, the theory and
mechanisms by which officers and minor arrests may reduce crime are discussed below.

Theory and Mechanisms of the Effects of Officers and Minor Arrests on Crime
Enforcement Based Persgctives (Deterrence and Incapacitation)

What are the mechanisms through which ofﬁcers and minor arrests affect more
serious felony offenses? The primary mechanism of crime control under professional
model of American policing was to seek the arrest of more serious offenders after a crime
was committed (Peterson et al., 1982; Skogan & Frydl, 2004; Visher, 2000). Thus,
objectives under this approach included incapacitation of offenders and deterrence of
future acts among observers who became aware of responses by the justice system. In
this study, ﬁndings of an inverse relationship between officer rates and aggravated
assault, burglary, larceny, and vehicle theft lend support to the deterrence and
incapacitation perspectives.

Previous studies that assessed the crime effects of relatively small increases in
ofﬁcer rates during the traditional era, when randomized patrol was the standard, often
found nonsigniﬁcant results (Eck & Maguire, 2000). Thus, it has been asserted that
jurisdiction-wide increases in officer presence are generally not effective with regard to
crime (Eck & Maguire, 2000). However, increased police presence in high crime areas
has been associated with crime reduction in more recent studies (Sherman, 1998;
Sherman & Rogan, 1995) and these ﬁndings are suggestive of a deterrent effect at

neighborhood, “hot spot” or other aggregate levels. Temporary increases in police

186

presence, however, have typically resulted in only temporary reductions in crime
(Sherman, 1998). Nevertheless, one study found that intermittent police presence of only
10 minutes in high crime areas resulted in sustained crime reduction. Thus, continuous
police efforts may be required in order to result in a long-term deterrent effect.

A recent panel data analysis indicated that the addition of ofﬁcers to force
strength also typically leads to increases in arrest (Zhao, Scheider & Thurman, 2003) and
the ﬁnding of a positive correlation between ofﬁcer rates and minor drug and disorder
arrest rates in the current study is consistent with this previous research. Sustained
increases in arrests for minor drug and disorder offenses may have a deterrent effect at
aggregate levels by signaling to potential offenders that criminal behavior will not be
tolerated.

Nevertheless, potential incapacitation effects of sustained increases in minor
arrests have been overlooked by many policing researchers. Continuous police efforts to
address minor drug and disorder offenses will likely lead to repeated arrests of frequent
offenders. Consequently, judicial action against frequent offenders in the community
may be more likely to result in increased jail time. For example, a typical sentence for
disorderly conduct may be one week in jail. However, individuals with multiple recent
disorderly conduct convictions may be sentenced to substantially longer time in jail (e. g.
6 to 12 months) by community judges that grow weary of their repeated offenses. Thus,
sustained increases in minor arrests may result in a cumulative incapacitation effect that
is substantially greater than the average penalty for any single offense.

Findings from recent studies also suggest that the combined effect of increased

officers, drug and disorder arrests may assist police in locating more serious offenders.

187

Recent analyses of ADAM (Arrestee Drug Abuse Monitoring) data, for example,
indicated that heavy drug users were typically frequent offenders and that individuals
with expensive habits were often persistent offenders (Kleiman, 1997). Another study
indicated that most frequent serious offenders also deal drugs in high volume (Chaiken &
Chaiken, 1982). These frequent and persistent offenders may be more likely to be
detected by increased officer presence and enforcement of minor drug offenses.
Additionally, recent data indicate that a substantial portion of high frequency, persistent
offenders commit robbery and burglary at rates of two to four per week, 20 to 40 times as
frequently as the median offender (Visher, 2000). Thus, the location and incapacitation
of these more serious offenders may also be enhanced through increased ofﬁcer presence
and enforcement in high crime areas.

Given previous analyses, which suggest that frequent, persistent offenders often
commit serious offenses an exceptionally high rate, the ﬁnding in the current study that
12 additional drug arrests per year (jurisdiction-wide) were associated with elimination of
one burglary is noteworthy. When viewed in the most narrow perspective, this ﬁnding
might suggest that 1 in 12 individuals arrested for a minor drug offense were also likely
to commit a burglary. However, a similar result would also possible if the pool of
individuals arrested for minor drug offenses included a small number of persons that
were frequent burglary offenders. Additionally, a broader perspective of this ﬁnding
might incorporate analyses of the effects of mediating or moderating factors, such as
perceptions of social disorder. A discussion of broader theoretical perspectives can be

found in the following section.

188

Communiry Crime Theory and Ecological Perspectives

The traditional or professional model of policing has been criticized as
fundamentally reactive in nature and tends to provide a narrow, enforcement oriented
view of the ofﬁcer function (Trojanowicz et al., 1998). However, ecological and social
disorganization perspectives suggest that the role of ofﬁcers in addressing fear of
victimization may also be important (Field, 2002; Shaw & McKay, 1942; Skogan &
Frydl, 2004). Models of social disorganization suggest that fear leads individuals to
physically and psychologically withdraw from community life (Shaw & McKay, 1942).
This withdrawal may lead to a weakening of the informal social control processes that
inhibit crime and disorder, resulting in further deterioration of the neighborhood. In this
regard, an increased ofﬁcer presence may serve to reduce citizen fear of victimization,
particularly if the increase in this combined with community policing efforts.
Assessments of the impact of foot patrols in both New Jersey and Michigan, for example,
suggest that citizen fear was reduced (Braga, 2002).

Another noteworthy ﬁnding with regard to the relationship between fear and
disorder is that community fear may arise from observations of physical signs or cues
such as abandoned cars, trash or grafﬁti (Perkins & Taylor, 1996). During the 19903, a
central tenet of the community policing movement involved the expansion of the
functional role of police to include efforts to address “quality of life” issues in
communities (Trojanowicz et al., 1998). Much of the motivation behind this broadened
policing role emanated from the “broken windows” thesis, which asserted that signs of
disorder and decay in the visible environment (e. g. litter, grafﬁti, abandoned automobiles,

boarded-up buildings, and other signs of physical disorder) serve as cues to residents and

189

passers by that residents in the neighborhood do not care about disorder, cleanliness and
upkeep (Wilson & Kelling, 1982). An extension of this thesis further suggests that
“behavioral incivilities” such as aggressive panhandling, public urination and defecation,
prostitution, loitering, public intoxication and other signs of social disorder increase
perceptions that personal victimization may be likely, causing law-abiding citizens to fear
for their safety and thus, attempt to avoid the area (Field, 2002; Kelling & Coles 1996;
Skogan, 1990).

The arrest indexes pertaining to minor drug and disorder offenses in the current
study were designed to provide a test of “broken windows” and “behavioral incivilities”
assertions. In this regard, the ﬁnding of an effect of a broad array of disorder arrests with
regard to robbery in the current study is interesting. Ofﬁcer rates were included in
regression models along with disorder arrests, reducing the likelihood of inﬂuences due
to changes in police force size. Additionally, the inclusion of drug arrest rates in the
model reduced the inﬂuence of changes in the level of police assertiveness as well as
drug offending. Thus, results indicated that increases in minor disorder arrests were
associated with reductions in robbery while controlling for changes in ofﬁcer levels and
drug arrest rates. Moreover, it is noteworthy that the nationwide average with regard to
disorder arrests declined substantially during the 19903, reducing the possibility that
signiﬁcant effects would be found due to general upward trends in the independent
variable and downward trends in the dependent variable.

Does this ﬁnding support the broken windows thesis? Unfortunately, the answer
to this question is somewhat unclear. The minor disorder arrest index was not associated

with declines in any offense category other than robbery when covariates and control

190

variables were included. Based on arguments pertaining to disorder and behavioral
incivilities, one would expect that other felony offenses would be affected as well. It was
noted, however, that burglary and vehicle theft were weakly related to minor disorder
arrests rates when ofﬁcer and drug arrest rates were not included in the model.
Additionally, it is possible that increases in arrests for minor disorder offenses were
associated with reductions in robbery because ﬁ'equent, persistent offenders are more
likely to commit serious offenses and were detected through the increased scrutiny
(Visher, 2000). Nevertheless, the ﬁnding of a strong relationship between disorder arrest
rates and only one of the seven offense types raises questions about the speciﬁc
mechanisms by which minor disorder arrests may effect robbery rates. Further research
may be helpful to determine which speciﬁc offenses (e. g. prostitution, curfew, runaway)
in the disorder index were most closely associated with crime reduction.
Changes in Ofﬁcers and Minor Arrests in the Two Wave Panel

The second portion of this discussion pertains to changes in policing practices that
occurred between 1993 and 1997 as reported to the National Survey of Community
Policing Strategies. It is noteworthy that neither officer rates nor minor drug and disorder
arrest rates demonstrated a signiﬁcant relationship with crime in regressions that involved
policing survey data. However, the Granger methodology that was previously discussed
involving 12 annual waves of data was not feasible with only two waves of data.
Additionally, although the reporting agencies in the policing survey represented nearly
one half of all reported crime during 1993 and 1997, the sample size was substantially

smaller. Consequently, these differences may account for the lack of signiﬁcant ﬁndings

191

with regard to ofﬁcers and minor arrests in the two wave panel. Further discussion of
this issue can be found in the limitations section.

Is also important to note, however, that changes in ofﬁcer rates, minor drug and
disorder arrests were generally not correlated with changes in policing tactics and
strategies as indicated by the eight policing practice indexes. Further, when variables
pertaining to ofﬁcer rates, minor drug and disorder arrests were entered into regression
models, the policing practice items continued to demonstrate signiﬁcant relationships
with regard to crime reduction. The inclusion of data pertaining to changes in ofﬁcers
and minor arrests did not reduce or signiﬁcantly alter the relationships between policing
practices and crime. Consequently, the variation in crime that is explained by policing
practices in the two wave panel is not the same variation that is explained by changes in
ofﬁcer rates and minor arrests. Thus, statistical effects of changes in policing practices
may be thought of as separate from and supplementary to (i.e. in addition to) the effects
of changes in ofﬁcers and minor arrests with regard to crime reduction. '3

Policing Practice Indexes ~ Connections to Prior Empirical Findings

Recent literature pertaining to policing practices during the 1990s suggests a
general trend toward policies that support community oriented policing as well as
increases in collaborative efforts by police organizations and reported citizen
involvements in daily activities (Roth et al., 2000). Data from the National Survey of
Community Policing Strategies, implemented in 1993 and 1997 are supportive of these

assertions and indicate that policing practices in each of these areas increased

substantially.

 

'3 See Table 47 on page 194 for estimates of strength of the policing-crime relationship.

192

Broadening Communiry Policing Policies

A summative index that assessed the breadth or increase in number of speciﬁc
types of policing practices indicated that community policing policies increased
approximately 77 percent by 1997 as compared with 1993 among responding agencies.
In the policing survey, question items that were included in this index pertained to
whether the agency was currently implementing community policing, had published
community policing goals, and whether new policies or ordinances had been connected to
support these objectives. Most prior research concerning the effectiveness of adopting a
community policing approach or implementation of generic community policing policies
has not found a signiﬁcant relationship with crime (Weisburd & Eek, 2004).
Additionally, a recent systematic review of policing practices listed community policing
efforts with no clear crime focus under the “what doesn't work” category (Sherman et al.,
1998). Results from regression analyses in this study are consistent with these previous
ﬁndings and indicate that increases in community policing policies were not related to
larceny, burglary, vehicle theft, robbery, aggravated assault, rape or murder.
Broadening Communiry Collaboration Practices

A nationwide survey of over 1,800 agencies, conducted in 1998 indicated that
efforts by police ofﬁcials to collaborate with citizens were less likely to be fully
implemented than other components of community policing ideology such as problem-
solving or changes in policy (Mastrofski, Parks & Wilson, 2003). These data are
consistent with this previous ﬁnding and although the index of collaboration efforts
indicated an increase of 16.7 percent, this increase was substantially lower than indexes

pertaining to problem solving, citizen involvement, or community policing policies. The

193

collaboration efforts index was comprised of questionnaire items pertaining to efforts by
police ofﬁcials to increase routine citizen interaction or feedback.

Although the collaboration attribute has not previously been empirically tested
with regard to crime reduction, criminal justice researchers have suggested that increased
feedback from citizens may assist police in identifying community concerns (Carter,

1995; Trojanowicz et al., 1998). In this study, the number of reported collaborative

—-—

efforts by police organizations was associated with signiﬁcant reductions in vehicle theft.
However, when problem-solving indexes were included in regression models, the
collaboration efforts variable became nonsigniﬁcant. This result is most likely due to the
substantial correlation between collaboration efforts and problem-solving (.40).
Nevertheless, from a theoretical standpoint, collaboration efforts and problem-solving
should be correlated since a primary objective of increasing citizen feedback is to address
neighborhood problems.
Broadening Citizen Involvement Practices

The index of citizen involvement differs ﬁom the collaboration efforts index in
two important respects. First, questionnaire items in the citizen involvement category
reﬂect actual implementation concerning police activities, whereas the collaboration
index is simply oriented toward organizational efforts or objectives. Second, citizen
involvement index items need not be initiated by police ofﬁcials. Indeed, citizen review
boards and neighborhood advisory councils may result from dissatisfaction or lack of
collaborative efforts. Consistent with ﬁndings in previous literature (Roth et al., 2000;

Trojanowicz et al., 1998), the index of citizen involvement showed a substantial increase

194

of more than 36 percent with regard to the number of policing procedures in which
citizens were actively involved.

Criminal justice researchers have asserted that increased citizen involvement
should lead to reductions in crime that result from efforts by individuals to assert
informal social controls and increase information to police regarding criminal activities
(Trojanowicz et al., 1998). Regression results from the current study are supportive of
this assertion and indicate that increases in citizen involvement were associated with
signiﬁcant declines in burglary, larceny and vehicle theft. Similar to the collaboration
efforts index, however, when entered into a regression model with problem-solving, the
relationship with burglary and vehicle theft became non-signiﬁcant. Nevertheless,
although citizen involvement and problem-solving were correlated, citizen involvement
demonstrated a stronger relationship than problem solving with regard to larceny. This
ﬁnding may indicate that citizens are in a better position than police to address this
relatively high frequency and widely dispersed type of offense.

Broadening Problem Solving Practices

Problem-solving is a general class of policing activities that involves systematic
efforts to address intermediate recurring mechanisms that encourage crime. Criminal
justice literature indicates that problem-solving was among the more popular innovations
across police agencies nationwide during the 19905 (Roth et al., 2000). Data from this
study are supportive of this assertion as indicated by the 29 percent increase in problem-
solving activities from 1993 to 1997.

A recent econometric analysis of problem-solving among 164 large cities did not

ﬁnd a relationship with regard to homicide or robbery (MacDonald, 2002). Nevertheless,

195

this study included serious methodological weaknesses, such as the failure to adjust for
heteroskedasticity and an assumption that problem-solving efforts were not in place
among large jurisdictions in 1993. Additionally, other studies have found that problem-
solving can reduce violent crime. A quasi experiment in 24 high crime locations in New
Jersey found signiﬁcant reductions in calls for police service regarding robbery and
property crime (Braga et al., 1999). Problem-solving approaches were also utilized to
reduce gun related homicide by gang members in Boston (Kennedy, 1998) and another
quasi-experiment resulted in decreased citizen complaints about crime and disorder in
Santa Ana, CA (Jesilow, Meyer, Parsons, et al., 1998).

Results from regression analyses in this study are supportive of a relationship
between increases in problem-solving approaches and crime declines. Increases in
problem-solving activities were associated with signiﬁcant reductions in robbery,
burglary, larceny and vehicle theft rates. Additionally, when other policing indexes were
included in regression models, problem-solving activities remained the strongest
predictor of declines in each of these offense types with the exception of larceny. As
previously mentioned, declines in larceny were best explained by increases in citizen
involvement. It is also noteworthy that problem-solving activities at both line (i.e. patrol
ofﬁcers and sergeants) and organization levels demonstrated effectiveness. Nevertheless,
actions at the organization level, such as efforts to redesign the organization to deal with
speciﬁc problems, were most strongly associated with declines robbery, burglary, larceny

and vehicle theft.

196

Broadening Crime Analysis Technigues

Crime analysis has been utilized to support problem-solving efforts by identifying
locations where frequent calls for police service occur. Additionally, due to increases in
the availability of microcomputers during the 19903, it was expected that an increase in
crime analysis techniques would be detected. However, the crime analysis index
indicated only a 6 percent increase among responding agencies, the smallest overall
increase among any of the policing indexes. It is possible that this small change in the
broadening of crime analysis practices indicates that current organizational needs were
being met through existing practices.

In terms of prior research, there is anecdotal evidence regarding the effectiveness
of crime analysis in New York City and New Jersey (Braga et al., 1999; Seaskate, 1998;
Sousa, 2001). However, no rigorous national evaluation has previously been conducted.
Regression analyses in this study did not detect a signiﬁcant relationship between
broadened crime analysis techniques and any of the seven UCR crime categories.
Coefficients were, however, in the expected direction (i.e. negative) and the values
pertaining to robbery were nearly signiﬁcant. Nevertheless, the adoption of a crime
analysis function among agencies where none previously existed was associated with
crime reduction. Further discussion of this ﬁnding can be found in the analyses of
individual policing techniques that follows.

Broadening Place Oriented Practices

Place oriented policing practices are a speciﬁc type of problem-solving technique

that involves efforts to increase police presence, enforcement, and oversight in areas with

frequent criminal activity. Criminal justice literature suggests that police awareness of

197

the importance of the immediate environment in problematic places increased
substantially during the 19905. In 1995, researcher John Eck advanced a “general model
of illicit retail marketplaces,” noting that routine activities theory suggests that sellers of
illegal goods and services will strive to limit the distance that they move in order to
maximize proﬁts. Data fi'om this study are supportive of assertions that place oriented
tactics increased during the 19905 as indicated by the 40 percent overall increase in the
place oriented index.

Recent studies suggest that place oriented tactics can be effective in reducing
crime. Dallas police department, for example, dramatically increased patrols in areas
with high gang violence in 1996 and results from a one-year quasi-experiment indicated
64 to 73 percent reductions in gang-related offenses among targeted areas (F ritsch, Caeti
& Taylor, 2003). A similar approach was also implemented by Detroit police department
between 1995 and 1998, yielding promising but less spectacular results (Bynum &
Varano, 2003). Consistent with these previous ﬁndings, regression results from this
study suggest that the broadened or increased number of place oriented practices among
responding jurisdictions was associated with reductions in robbery and vehicle theft.
However, when these policing practices were entered into a combined model with other
indexes, regressions indicated that place oriented practices did not explain a signiﬁcant
amount of unique variation in crime beyond that of problem-solving and citizen
involvement.

Broadening Civil Code Enforcement Practices
Criminal justice literature suggests that the use of civil and regulatory code

enforcement increased substantially during the 19903 among police jurisdictions

198

(Mazerolle & Roehl, 1998). Civil and regulatory codes have been utilized as a means of
decreasing social disorganization and dealing with recurring crime problems. They have
also been utilized in conjunction with place oriented strategies as a means to persuade or
coerce law abiding individuals and non-offending third parties to take responsibility and
action to prevent or end criminal or nuisance behavior (Mazerolle & Roehl, 1998).
Consistent with these assertions in criminal justice literature, data in the current study

indicate that the number of civil code enforcement techniques utilized by responding

f ‘_ 'm—El

jurisdictions increased approximately 30 percent.

A number of recent studies have suggested that the use of civil and regulatory
code enforcement can be effective in combating crime. One such study, involved the
creation of a special team known as SMART or Specialized Multi-Ageney Response
team which coordinated with other city agencies to inspect and identify properties that
failed to comply with civil ordinances and where frequent calls for police service
occurred (Green, 1995). The team utilized regulatory codes and drug abatement laws,
resulting in a decline of observable disorder and drug offenses (Green, 1995).

Civil and regulatory code enforcement tactics have also been utilized in
conjunction with training for landlords and property managers, ﬁnes for non-compliance,
and property forfeiture to deal with residences with a long history of criminal activities.
In a recent study of 121 rental properties that had been recent targets of drug
enforcement, landlords were contacted by a narcotics detective and a city code
enforcement ofﬁcial who inspected the property and worked with the property owner to
devise a plan to remove or reduce the drug problem (Eek & Wartell, 1998). A 30 month

review indicated that crime declined by approximately 60% in places where landlords

199

were met by city ofﬁcials while crime in places that had received a letter of warning
declined by 13% when compared with a control group.

Civil remedies have also expanded to speciﬁcally target offending parties such as
gangs, domestic batterers, and delinquent youth. In Southern California, police and legal
ofﬁcials sought gang injunctions that would allow civil ﬁnes or arrest of certain
individuals that were identiﬁed as gang members (Maxson, Hennigan & Sloan, 2003).
Two cities that utilized this approach reported substantial decreases in gang related crime
(Maxson, Hennigan & Sloan, 2003).

In the current study, regression results pertaining to broadening or increasing the
number of civil code enforcement techniques were not associated with reductions in any
of the criminal offense categories. However, jurisdictions that reported that civil code
enforcement was not utilized by most patrol ofﬁcers in 1993 and later began utilizing this
approach reported signiﬁcant reductions in murder, robbery, burglary, larceny and
vehicle theft. The enforcement of civil and regulatory codes by most police officers
within jurisdictions was among the strongest and most broadly effective crime reduction
strategies in the policing survey. Further discussion of results pertaining to civil code
enforcement as an individual tactic may be found in a following section.

Broadening Efficiency Oriented Practices

During the past 30 years, law enforcement ofﬁcials made efforts to shorten
response time to criminal incidents and implement single-ofﬁcer vehicular patrols with
the goal of increasing efﬁciency (Sparrow, Moore & Kennedy, 1990). Additionally, in
recent years, a number of technological advances occurred among electronic devices that

supported this objective. However, community policing ideology, which also increased

200

during the 19903, emphasizes relations with citizens and quality police services to a
greater extent than efficiency (Trojanowicz et al., 1998). Thus, criminal justice literature
offered little guidance with regard to expectations pertaining to organizational efficiency.
Nevertheless, data from this study indicate a modest 12 percent increase in organizational
efficiency related efforts. 1

Questionnaire items that comprise the efﬁciency index include implementation of
a drug tip or crime hotline, prioritization of calls to increase ofﬁcer time, providing
agency access to other city or county databases, and a variety of other techniques. Very
little empirical work has been conducted with regard to efﬁciency and technology.
However, the implementation of in-car computers and mobile data terminals was the
subject of one study, which indicated that even among the smallest agencies (i.e. those
with fewer than 20 sworn ofﬁcers), nearly 1 in 5 reported utilizing mobile data terminals
by the year 2000 (Eastern Kentucky University, 2003). Another study in Forth Worth,
Texas during the early 1990s analysis indicated that vehicle recovery rates (but not arrest
rates) were improved after the implementation of mobile data terminals (Nunn, 2001).
Nevertheless, efforts to improve officer response time have generally not been found to
be effective with regard to crime reduction (Sparrow, Moore & Kennedy, 1990).

Consistent with assertions by criminal justice researchers and community policing
advocates, regression analyses in the current study did not detect any signiﬁcant
relationships pertaining to the number of efﬁciency related efforts and crime. However, a
number of individual items that fell under this classiﬁcation were associated with
reductions in crime. For example, agency implementation of a specialized crime

prevention unit and participation in an interagency drug task force were associated with

201

declines in robbery, burglary and larceny. Additionally, implementation of a drug or
crime tip hotline was associated with reductions and aggravated assault. These individual
policing practices will be further discussed in a following section.
Individual Policing Practices ~ Connections to Crime Theory
Although analyses of policing practice indexes can provide information regarding

the breadth or expansion of broad categories of policing tactics, this approach offers little

n;
'1

information regarding the potential crime reduction effects of speciﬁc tactics and

1F< '

strategies. As previously noted, for example, police organizations that expanded existing
crime analysis techniques did not experience signiﬁcant declines in crime. However,
agencies that ﬁrst began conducting crime analysis after 1993 did report substantial
crime reductions. '4 Consequently, analyses of each of the 80 individual survey items
pertaining to tactics and strategies that comprise the indexes were also conducted. These
individual analyses provided additional detail regarding the speciﬁc types of problem-
solving practices that were most closely associated with crime reductions as well as the
effectiveness of speciﬁc types of collaboration, place oriented, and other tactics and
strategies.

Another advantage of analyses of individual items is that they provide further
information regarding the types of policing practices that tended to affect speciﬁc
offenses such as larceny when compared with practices that affected burglary, robbery
and each of the other offense categories. A qualitative assessment of similarities and
patterns among policing practices may suggest that the majority of effective practices
involved increasing police presence in high crime locations, for example. The primary

objectives of these analyses were to provide a starting point for further research by

202

identifying the types of policing practices that were associated with declines in crime as
well as to identify potential mechanisms through which crime effects may have occurred.
Larceny and Individual Policing Practices

Overall, 11 of the 80 individual policing practice items were associated with
reductions in larceny rates (Appendix C, Table 1). In this regard, the majority of
signiﬁcant items described citizen actions or police efforts to gain the assistance of
citizens. These items included implementation of citizen patrols, citizen prepared work
agreements specifying work to be done by police, and implementation of a citizens police
academy. The signiﬁcance of these items may emanate from the fact that larceny is a
fairly pervasive and frequent offense such that police are heavily reliant upon the
observation and intervention of individuals in order to reduce occurrences. Additionally,
this ﬁnding is consistent with assertions by Hunter (1985) who argued that informal and
formal social mechanisms can be classiﬁed into three levels of control (private, parochial,
public). Private control is the most basic level and involves interactions among intimate
groups (e. g. family and close friends). However, parochialism comprises the second
level of social control via interpersonal networks and connections with local institutions
such as churches, stores and schools. Thus, police efforts to gain the assistance of
citizens may also result in increased informal social controls.

Survey items describing citizen input regarding the performance and selection of
ofﬁcers as well as citizen review of complaints against police comprised another theme
with regard to larceny. These types of citizen actions likely increase scrutiny of police
services, potentially leading to increased efforts on the part of officers to pursue

investigative leads and recover property. However, community policing advocates have

 

'4 See Appendix D for details regarding the changes of each of the 80 policing practices.

203

also argued that increased citizen input regarding police activities leads to improved
relationships between police and citizens. There is evidence, for example, that citizen
trust and perceptions of legitimacy increased in Chicago after police implemented a
community policing approach (Skogan & Hartnett, 1997). Moreover, ﬁndings from Carr
(2003) indicated that many Chicago area residents changed how they responded to youths
whose families were unwilling or unable to exert sufficient private control. Under this
“new parochialism,” many residents in this ﬁve-year ethnographic study contacted
community police ofﬁcers who intervened on their behalf, rather than contacting the
families of disorderly youths directly.

A ﬁnal theme with regard to individual policing practices that were associated
with larceny reduction involved organizational adaptability and enforcement activities.
In this regard, a signiﬁcant portion of jurisdictions that experienced a decline in larceny
had recently begun aggressive civil and regulatory code enforcement as well as
participation in an anti drug task force. However, data indicate that these enforcement
activities were often supplemented by new authorities for middle-management to
redesign the organization to support problem-solving efforts as well as implementation of
mobile, neighborhood based police stations. The statistical signiﬁcance of these policing
practices pertaining to enforcement with adaptability is supportive of recent assertions by
criminal justice researchers that targeted and focused enforcement activities are more
likely to result in crime reduction than broadly applied or undirected enforcement actions

(Skogan & Frydl, 2004).

204

Bu :1 and Individual Poliein Practices

A total of 15 policing practices were associated with reductions in burglary rates
and a number of these were also associated with reductions in larceny (Appendix C,
Table 2). The offenses of burglary and larceny are similar in nature and it is not
surprising that similar policing practices would affect both types of offenses.
Nevertheless, the majority of policing practices that were associated with reductions in
burglary were not associated with larceny and thus the patterns of effective policing
practices also differed somewhat. Although citizen involvements with regard to
preparing work agreements and attending a citizens police academy were signiﬁcant, the
primary pattern among policing practices pertaining to burglary involved the long term
assignment of police ofﬁcers to speciﬁc neighborhoods and efforts to assist citizens in
community organization. In this regard, the ﬁxed assignment of patrol ofﬁcers to
speciﬁc areas and designation of patrol routes that coincide with neighborhood
boundaries were both inﬂuential. Nevertheless, jurisdictions that reported conducting
citizen surveys and where officers worked directly with citizens in their patrol areas to
identify and resolve problems also experienced declines in burglary along with agencies
that provided speciﬁc problem-solving training to citizens.

The ﬁnding that neighborhood based community building efforts were associated
with reductions in burglary is consistent with recent studies pertaining to the importance
of collective efficacy with regard to community social control. In this regard, one study
implemented a systematic social observation of urban neighborhoods ﬁnding that
collective efﬁcacy was linked to lower rates of violent crime (Sampson & Raudenbush,

1999). Additionally, data from a survey of 2,482 Illinois residents suggested an

205

interaction-effect between fear and social ties (Ross & J ang, 2000). The authors of this
study concluded that interpersonal social connections decreased fear of victimization and
that participation in neighborhood organizations also provided a “buffering effect” with
regard to fear and perceptions of disorder (Ross & J ang, 2000). Consequently, police
efforts to contact and train citizens with regard to problem-solving may also result in a
decrease in fear among citizens.

A ﬁnal trend among policing practices that were associated with reductions in

—. _____i

burglary concerns enforcement efforts in high crime areas. Jurisdictions that began
aggressively enforcing civil and regulatory codes, initiated a crime prevention function,
and participated in multi-jurisdictional drug task forces experienced signiﬁcant
reductions in burglary. The combination of these items suggests that enforcement
activities in these jurisdictions may have been targeted to speciﬁc crime problem areas.
These ﬁndings are consistent with assertions by criminal justice researchers that place or
“hot spot” oriented enforcement activities may result in crime decreases (Eek, 1995). It
is also noteworthy that participation in a drug task force was associated with reduction in
burglary because previous analyses in this study indicated that increases in minor drug
arrests were also associated with reductions in burglary. Thus, both data sources (i.e.
arrest and policing practice data) in this study suggest that there is a relationship between
drug activity and rates of burglary.
Vehicle Theft and Individual Policing Practices

Overall 21 individual policing practices were associated with reductions in
vehicle theft (Appendix C, Table 3). In this regard, the primary pattern with regard to

policing practices was clearly related to problem-solving activities and new authorities

206

for middle managers. Eight of the signiﬁcant policing practices involved problem~
solving efforts. Additionally, ﬁve of these survey items indicated active participation by
middle managers with regard to problem-solving, such as seeking the input of patrol
ofﬁcers, determining how resources should be utilized and structuring organizational
procedures to deal with problems. These ﬁndings provide support for assertions by
criminal justice researchers that increases in the level of input and decision making of
personnel at all levels of the organization may improve quality of policing (Trojanowicz
et al., 1998). They also provide support for assertions by problem-solving advocates that
the dedication and focus of agency resources may increase chances for success with
regard to crime (Braga, 1999).

Another trend with regard to signiﬁcant policing practices pertaining to vehicle
theft involved effort to increased feedback and support from citizens. In this regard,
conducting citizen surveys, implementing regular community meetings, allowing citizens
to prepare work agreements pertaining to neighborhood problems, and providing
problem-solving training to citizens were important. However, unlike the citizen related
involvements pertaining to larceny and burglary, which tended to be oriented toward
community self-help, signiﬁcant policing practices with regard to vehicle theft were
directed more toward enabling police ofﬁcials. In this regard, citizen involvement
practices were supplemented with the implementation of foot patrol, ﬁxed assignment of
officers and the implementation of crime analysis, specialized crime prevention units and
increased civil code enforcement. These ﬁndings support assertions that improved
police-community relations may be necessary to enable increased police enforcement

action in high crime areas (Trojanowicz et al., 1998).

207

 

Robbery and Individual Policing Practices

Overall 14 policing practices were signiﬁcantly associated with reductions in
robbery (Appendix C, Table 4). However, practices pertaining to citizen involvement
played a much smaller role with regard to this offense. Not a single tactic or strategy
from the collaboration efforts index demonstrated statistical signiﬁcance in relation to
robbery. Additionally, it was noted that the distribution of signiﬁcant policing practices
was much less concentrated for robbery than for the other offense types (Table 34),
suggesting that fewer of the practices found in the community policing survey were
effective. There was a small cluster of problem solving activities that demonstrated
signiﬁcance. However, effective policing practices with regard to robbery were most
often related to neighborhood oriented enforcement activities.

Neighborhood oriented enforcement activities included implementation of
neighborhood or community-based decision making, ﬁxed assignment of patrol ofﬁcers
to speciﬁc beats or areas, mobile neighborhood-based stations, implementation of crime
analysis and participation in a drug task force. Recall also that previous ﬁndings in this
study pertaining to minor drug and disorder arrests also demonstrated strong relationships
with robbery rates. Consequently, these ﬁndings are consistent with prior studies that
indicated that aggressive enforcement activity is associated with declines in rates of
robbery (Sampson & Cohen, 1988; Wilson and Boland, 1978).

Rape and Individual Policing Practices

The patterns among policing practices pertaining to the offense of rape were very

different than for any of the other offense categories (Appendix C, Table 5). First, only

six of the 80 policing practices demonstrated statistical signiﬁcance when assessed

208

individually. Policing practices that were weakly related to reported rates of rape include
implementation of a drug education program in schools, physical decentralization of
investigations and citizen cooperation with police regarding neighborhood problems.
However, when all six items were entered into the regression together, three policing
practices demonstrated stronger relationships with rates of rape. These items include the
implementation of multidisciplinary teams to deal with special problems, increased
efforts by patrol ofﬁcers to work with other city agencies, and interagency involvement in
problem identiﬁcation and resolution.

It is noteworthy that all three of these policing practices involve coordination with
other agencies. The theoretical basis for these statistical relationships is unknown and
should be explored more closely. However, given that reported rates of rape did not
substantially decline in most jurisdiction during the 19903, the likelihood that these
practices, which are similar in nature are spuriously related to rape is reduced. It was also
noted that during the 1990s, the Ofﬁce on Violence Against Women provided substantial
grant funding and training to local police agencies. Thus, it is possible that the formation
of multidisciplinary teams and increased coordination with other agencies resulted from
these efforts.

Aggravated Assault and Individual Policing Practices

Six policing practices were associated with reductions in aggravated assault when
items were assessed individually (Appendix C, Table 6). In this regard, patterns among
signiﬁcant policing practices suggest that police efforts to aid citizens in regaining
control of their communities were important. For example, signiﬁcant items pertaining to

implementation of a drug or crime tip hotline, as well as mobile, neighborhood-based

209

stations and drug-free zones around schools parks or churches each provide mechanisms
for citizens to report suspicious activities. Additionally, reported increases in efforts by
patrol ofﬁcers to help communities organize and increased efforts to teach residents how
to address community problems may have result improved the ability of the community
to help itself recover from crime and disorder problems. Thus, these ﬁndings are
supportive of work by Greenberg et a1. (1982) who identiﬁed three forums that are
dedicated to community self-regulation including informal surveillance, movement-
goveming rules and direct intervention or questioning of strangers and residents in the
neighborhood about suspicious activities.
Murder and Individual Policing Practices

Overall ﬁve policing practices were associated with reductions in murder
(Appendix C, Table 7). When all ﬁve policing practices were entered together into a
regression, the two that were most strongly related to murder rates were implementation
of a crime analysis function and implementation of mobile neighborhood-based ofﬁces or
stations. Given that a substantial portion of murder occurs as a result of gang and drug
activity, the combination of crime analysis and mobile police stations suggest that
agencies may have strategically increased their presence wherever crime problems
emerged. However, it is also noteworthy that increases in efforts by patrol ofﬁcers to
make door-to-door contacts with citizens in their patrol area, increased civil code
enforcement and citizen prepared work agreements regarding neighborhood problems
were also weakly associated with murder rates. Taken together, these practices are
indicative of police efforts to become a part of the neighborhood social fabric in high

crime areas. Consequently, these ﬁndings are consistent with the assertions by criminal

210

 

justice researchers regarding the importance in social networks (Bursik & Grasmick,
1997) and the emergence of a new form of parochialism (Carr, 2003).

Discussion Summary and Conclusions
Policing Practices

Changes in policing tactics and strategies during the 19908 were associated with
reductions in each of the seven felony index offenses. However, although the number or
breadth of reported problem solving techniques and areas of citizen involvement were
associated with relatively small declines in crime ranging from 1.6 to 3.1 percent,
implementation of speciﬁc individual policing practices were generally associated with
more substantial crime reductions. For example, the combined effects of policing
practices pertaining to citizen involvement, civil code enforcement and participation in a
drug task force were associated with a 7.7 percent decline in larcenies from 1993 to 1997.
Additionally, the combined effects of similar policing practices were associated with
declines in robbery, burglary and vehicle theft of 5.7%, 4.2% and 6.5% respectively.
However, it was also noted that individual policing practices were associated with
smaller declines in the offenses of murder (1%) and rape (1.5%).

An assessment of a broad spectrum of 80 individual policing tactics and strategies
indicated that increased enforcement of civil and regulatory code violations and
implementation of a crime analysis function were the most broadly inﬂuential practices in
terms of the number of crime declines with which they were associated. However,
policing practices pertaining to increased citizen involvement were most important with

regard to burglary and larceny. Additionally, various individual problem solving tactics

211

were associated with substantial reductions in every offense category with the exception
of larceny.

The types of speciﬁc policing practices that were associated with crime reduction
varied for each offense category. For example, reductions in aggravated assault were
associated with police efforts to aid citizens in regaining control of their communities,
such as implementation of a drug or crime tip hotline and mobile, neighborhood-based
stations. In contrast, reductions in rates of robbery were more closely associated with
targeted enforcement activities. Finally, jurisdictions that initiated multidisciplinary
teams to deal with special problems and that worked closely with other agencies reported
greater decreases in rates of rape.

Changes in Ofﬁcer and Minor Arrest Rates

During the 1990s, changes in the number of ofﬁcers per capita and minor arrests
for drug and disorder offenses were also associated with declines in most offenses
categories with the exception of murder and rape. However, in contrast to changes in
policing practices, the amount of crime declines that were explained by ofﬁcers and
minor arrests were small, ranging from zero to 2.3 percent. Nationwide rates of drug
arrests increased substantially during the 1990s and these increases were associated with
reductions in robbery, burglary, and aggravated assault. However, although the
nationwide average number of arrests for minor disorder declined, jurisdictions that did
not follow this national trend and that increased minor disorder arrest rates reported
signiﬁcant reductions in rates of robbery. Finally, the number of ofﬁcers per capita
nationwide also increased substantially this decade and these increases were associated

with decreases in larceny, burglary, vehicle theft and aggravated assault.

212

The Relationship Between Policing Changes and Crime During the 1990s

Prior to the 1990s, a number of criminal justice researchers asserted that police
actions could not affect jurisdiction-wide crime rates (Bayley, 1994; Walker, 1984).
There were both theoretical and empirical reasons supporting this assertion. From a
theoretical standpoint, criminologists had argued that a substantial portion of crimes were
committed as consequences of extreme poverty (Cullen & Agnew, 2003). Consequently,
since police could do little to address this causal factor, it was assumed that little could be
done to reduce crime rates. Additionally, studies regarding incarceration suggested that
"nothing worked" with regard to recidivism. Finally, in the policing arena, studies
regarding randomized patrol and shorter response time to criminal incidents failed to
demonstrate effectiveness (Kelling et al., 1974; Sparrow, Moore & Kennedy, 1990).

During the late 19805 and early 19903, however, research ﬁndings increasingly
suggested that speciﬁc types of police activities, such as directed patrol of high crime
areas could reduce the number of criminal incidents (Sherman & Rogan, 1995).
Additionally, the emergence of community policing and problem solving approaches
increasingly suggested that efforts to identify and reduce aggravating intermediate causal
mechanisms in neighborhoods and small geographic places could be successful in
reducing crime under certain conditions.

Despite the advancement in knowledge resulting ﬁom recent studies which
suggested that a narrow range policing efforts could impact crime, little prior research
had been conducted to determine the speciﬁc types of practices that may be effective in
reducing speciﬁc types of crime. Additionally, prior to the implementation of this study

very little was known regarding the effects of policing tactics and strategies among

213

jurisdictions of all sizes and from all regions of the nation. Nevertheless, this assessment
of data from two large national samples clearly suggests that a wide array of policing
practices may result in crime reduction and were, in fact, associated with substantial
reductions in crime during the 19905.

In terms of the broader overall policing effect, Table 47 (next page) contains
estimates of the nationwide crime declines that were associated with policing changes
during the 19905. The data utilized in this study indicate that approximately 10 percent
of the reduction in larceny was associated with policing changes. The vast majority of
this relationship between policing and reported rates of larceny was associated with
changes in tactics (7.7%), rather than increases in ofﬁcers or minor arrests. Estimated
policing changes associated with declines in robbery, aggravated assault, burglary, and
vehicle theft ranged from 4.5 to 8.1 percent nationwide as well. However, corresponding
declines in rape (1.5%) and murder (1%) were very small and were associated with a only

a few types of policing tactics (see Tables 34 and 35).

214

.A Sam-o8: 8.8 8:82 3:.ch 20 .53 8:88.82 38 mo: no 8.8 N

.33 can 32 5 25.8 8388.. 3528.5 .5 gov 5:888»: 8688 80 833 5:53: a 89¢ 88 co 88m r

 

 

 

 

 

o._ m4 m6 wd ..w mo o.o_ Hootm w50=om Bumﬁzmm :38.
.2: .2: Nd 2 m: 0; Wm N 382.5. .852 can 8850
3 2 2 5 we 2. E _ 888m 93 883.
X. X. ..\e .x. ..\e ..\e e\e
8 .52 and “58¢. be cm 0 2 o> he wczm 58:3 8mg 0 was. on
8 ~ an _ .n _ n . .—

 

8wS£U 56:8 53, 8588.22. 8580 25.5 3:582 80 88588; 8353m— 54 038.

215

Limitations

The analyses of policing practices in this study emphasized change rather than
levels. As such, an objective of this methodology was to identify potential policing
practices that may have been associated with crime reduction while minimizing the
impact of omitted variable bias. However, given that the implementation of some
policing practices did not substantially change or increase during the 19905, this approach
cannot be utilized to determine that a speciﬁc policing tactic or strategy is ineffective
with regard to crime reduction. For example, if implementation of black and white patrol
cars was a proven crime reduction technique such that all police departments utilized this
approach, the resulting change in implementation during the 1990s would be zero.
Although statistical analyses of the implementation of black and white patrol cars would
result in nonsigniﬁcant ﬁndings, this lack of statistical signiﬁcance (due to the lack of
change in the population) would not mean that the tactic was ineffective. Consequently,
although the primary research topic in this study involved an assessment of whether
policing changes during the 19903 were associated with crime reductions, nonsigniﬁcant
ﬁndings for speciﬁc individual policing practices are not necessarily indicative of
ineffectiveness.

The primary statistical analyses pertaining to changes in per capita ofﬁcers and
minor arrests were conducted via a Granger causality model and demonstrated signiﬁcant
relationships with regard to a variety of reported offenses. However, it was noted that
officers and minor arrests did not demonstrate statistical significance with regard to crime
in the smaller, two-wave survey subsample. As previously discussed, these

nonsigniﬁcant ﬁndings may have resulted from the fact that a Granger approach cannot

216

be utilized with a two-wave panel. Nevertheless, these nonsigniﬁcant ﬁndings may also
indicate that relationships between officers, minor arrests, and crime are weak or not very
robust to changes in sample size. However, as previously mentioned, no less than three
prior studies have also found relationships between changes in per capita ofﬁcers and
crime (Kovandzic & Sloan, 2002; Levitt, 1997; Marvell & Moody, 1996). Thus, there is
reason for some measure of conﬁdence with regard to this relationship.

It was noted in the methodology section that 278 agencies that responded to the
National Survey of Community Policing Strategies did not report complete crime data in
both 1993 and 1997. Consequently, the fact that nearly three quarters of these agencies
police populations of less than 50,000 citizens indicates that the policing survey data in
this study underrepresented smaller jurisdictions. However, given that the variable of
interest in this study is crime and that the vast majority of crime occurs in larger
jurisdictions this underrepresentation is not highly problematic. Jurisdictions that were
included in analyses pertaining to the policing survey subsample represented nearly one
half (49%) of all reported crime in 1993 and 1997.

It was also noted in the methodology that although the policing survey subsample
was generally consistent with the crime demographics of the UCR population, the
proportion of total crime in the Northeastern region was approximately 6% below that of
the UCR population. This underrepresentation of crime from the Northeast is a minor
data limitation. The Northeastern region represents the smallest geographic area and
reports the smallest portion of crime nationwide. Thus, although inferences to the
Northeastern region are not as strong as for other regions of the nation, there is little

reason to believe that the effectiveness of policing practices in the Northeast is

217

substantially different than in other regions. Consequently, the overall representation of
crime among the policing survey subsample is sufficiently similar to allow reasoned
inferences to the broader UCR population.

Policy Recommendations

Given that ﬁndings from this study suggest that policing activities have been
associated with reductions in crime, a logical question arises as to which speciﬁc
practices are most advisable. In this regard, readers should be cautioned that analyses
pertaining to individual policing practices (e. g. implementation of a citizens academy)
were largely exploratory in nature. However, data in this study clearly indicate that a
variety of policing practices were associated with crime reductions across a wide array of
places, environments and organizational contexts and thus merit further scrutiny.
Nevertheless, with regard to current policing policies, potential costs and beneﬁts must be
considered.

Although formal cost-beneﬁt analyses are beyond the scope of this study, three
classiﬁcations were utilized as a framework for recommendations. The ﬁrst classiﬁcation
pertains to practices that were associated with crime reduction and that may be
implemented with relative ease and at low cost. The second classiﬁcation involves
practices that were associated with reductions of multiple (i.e. 3 or more) offense
categories. And, the ﬁnal classiﬁcation pertains to practices that were strongly associated
with the reduction of a speciﬁc offense when entered into regression models along with

an array of other policing tactics and strategies.

218

Practices That Are Low Cost and Easy to Implement

Broadened implementation of problem solving was associated with reductions in
nearly every offense category and may be among the least expensive policing strategies.
Among the individual policing survey items pertaining to problem solving, virtually all
were associated with reductions of at least one type of crime. Thus, the speciﬁc method
of problem solving that is utilized may not be as important as increased organizational
effort to address problems in a more general sense.

Nevertheless, analyses in this study indicated that three speciﬁc problem solving
practices were associated with substantial reductions in multiple offenses. Police efforts
to encourage citizens to become involved in neighborhood problems, coordination with
other organizations,” and increased authority of middle managers to restructure
organizational processes were associated with substantial reductions in at least three
types of offenses. Although these speciﬁc problem solving strategies may require a fair
amount of time to initiate, their implementation is not costly ﬁom a monetary
perspective. Additionally, similar arguments are applicable with regard to broadening
efforts to encourage citizen involvement in other aspects of police activities. Police
administrators should consider increasing organizational efforts to address problems and
to improve citizen involvement as potentially effective and low cost generalized
approaches to crime reduction.

Analyses in this study also suggested that increased enforcement of civil and
regulatory codes by patrol ofﬁcers was among the strongest and most broadly applicable

techniques that were associated with crime reduction. Jurisdictions that reported that

 

'5 Including both law enforcement and non-law enforcement agencies.

219

 

most officers enforced civil and regulatory codes in an effort to reduce crime reported
signiﬁcant reductions in murder, robbery, burglary, larceny and vehicle theft. However,
it should be noted that the code enforcement questionnaire item does not provide details
regarding the extent of enforcement by individual ofﬁcers. Rather, data indicate that a
change with regard to the number of patrol ofﬁcers that utilize this technique was
associated with crime reduction. That is to say, agencies that reported in 1993 that few or
none of their patrol ofﬁcers enforced regulatory codes and then reported that most did so
in 1997, experienced substantial crime decreases. This policing practice may also be
combined with problem solving and citizen involvement techniques at little monetary
cost since incarceration or physical apprehension of the offender is generally not required
with enforcement of civil and regulatory codes.

The ﬁxed assignment of patrol ofﬁcers to speciﬁc neighborhoods was another
relatively low cost and easily implemented practice that was associated with substantial
reductions in robbery, burglary and vehicle theft. Community policing advocates have
argued that when ofﬁcers become familiar with individuals and routines in their districts
they may become more effective in their patrol efforts (Trojanowicz et al., 1998). Given
the serious nature of the robbery and home invasion as well as the substantial monetary
losses associated with vehicle theft, police managers should seriously consider the ﬁxed
assignment of patrol ofﬁcers to geographic areas.

A ﬁnal low cost strategy involves the implementation of foot patrol as a speciﬁc
assignment for some patrol ofﬁcers. Although only one offense category, vehicle theft,
was associated with the implementation of foot patrol, the ease of implementation of this

approach makes it an attractive recommendation. However, it is also noteworthy that a

220

substantial number of jurisdictions implemented foot patrol along with crime analysis,
which suggests that the timing and location of these efforts must be carefully considered.
Nevertheless, this policing practice was associated with reductions in vehicle theft in
another recent study (Di Tella & Schargrodsky, 2004). Therefore, officials in
jurisdictions that experience episodic problems with vehicle theft in speciﬁc areas should

consider the implementation of foot patrol during times of peak occurrence.

Practices that were Associated with Reductions of Multiple (3 or more) Offense

Categories

Increases in the number of ofﬁcers per capita among jurisdictions nationwide
were associated with declines in rates of aggravated assault, burglary, larceny and vehicle
theft. However, the addition of patrol officers may be among the most costly of policing
practices that were analyzed in this study. Findings indicate that, based on nationwide
averages during the 19903, the addition of two police ofﬁcers was required to eliminate
one larceny per 100,000 citizens. Furthermore, an increase of more than eight ofﬁcers
was associated with the reduction of one vehicle theft or aggravated assault.
Consequently, increases to police force size may be most appropriate for jurisdictions
that are understaffed such that beneﬁts will result in other aspects of the organization in
addition to the potential for crime reduction.

It is also noteworthy that increases in arrests for minor drug offenses were
associated with declines in robbery, burglary and aggravated assault. When compared
with the monetary costs associated with the addition of police personnel, increased
enforcement of drug use or possession may be an attractive alternative. Analyses in this

study suggest that an increase of only 12 arrests per year was associated with the

221

elimination of one burglary per 100,000 citizens. However, increased drug enforcement
may be time-consuming if ofﬁcers are required to take physical custody of offenders,
rather than issuing a citation. Additionally, laboratory testing of controlled substances or
trace blood levels may be required, increasing the monetary cost of enforcement. Finally,
this approach may not be feasible in jurisdictions where police community relations are
strained unless efforts to increase citizen support are also implemented.

Implementation of a process whereby citizens prepare work agreements or
contracts specify work to be done by police concerning neighborhood problems was
associated with reductions in murder, robbery, burglary, larceny and vehicle theft. Thus,
this approach offers a less costly and potentially effective alternative to increased ofﬁcer
and drug arrest rates. The initial steps toward citizen prepared work agreements may
require substantial administrative time to ensure that procedures are appropriate and
useful. Nevertheless, if implemented properly, this approach may be among the most
broadly effective practices with regard to crime reduction.

The ﬁnal policing practice that was associated with reductions in multiple offense
categories pertains to crime analysis. The implementation of a crime analysis unit or
ﬁmction was associated with reductions in murder, robbery, burglary and vehicle theft. It
should be noted, however, that the current study does not provide a clear indication as to
whether the national average of the statistical crime analysis effects apply equally to
smaller jurisdictions where crime rates are low. Therefore, the potential monetary cost of
this approach should be carefully considered. Nevertheless, crime analysis may be a
linchpin with regard to the proper application of problem solving and place or “hot spot”

oriented strategies in large jurisdictions.

222

 

Practices that were Strongly Associated with the Reduction of a Specific Offense

Although the national trend with regard to minor disorder arrests was downward,
jurisdictions that increased their enforcement of prostitution, curfew violation, disorderly
conduct and similar offenses experienced substantial declines in rates of robbery. Based
on the nationwide average, increases of approximately 3.5 minor disorder arrests per
month were associated with the elimination of one robbery per 100,000 citizens.
However, as with drug arrests, police-community relations should be carefully considered
if this practice is implemented. Additionally, police organizations may experience
increased costs pertaining to enforcement if subjects resist police intervention for minor
infractions and physical custody of offenders is required (rather than simple issuance of a
ticket or summons). Nevertheless, if implemented properly, increased enforcement of
disorder related offenses may result in declines in robbery in some jurisdictions.

Citizen input into police hiring was strongly associated with reductions in robbery
and larceny. Additionally, citizen review of complaints against police was strongly
associated with reductions in larceny. Additional research is needed to determine the
mechanisms by which increased citizen involvement in police personnel processes were
associated with crime reduction. It is possible, for example, that increased citizen
involvement results in the hiring of ofﬁcers that are more attentive to community needs.
Alternatively, increased review of police conduct may result in improved attention to
community problems or minor infractions. Nevertheless, increased citizen input into
these routine police processes may be appropriate for some jurisdictions and should be

considered by police administrators.

223

 

A ﬁnal policing practice that was strongly associated with crime reduction was
the implementation of a citizens police academy. This practice was strongly associated
with reductions in burglary and larceny. However, this approach may be somewhat
costly and time-consuming with regard to administrative requirements. Additionally, the
precise mechanisms of property crime reduction associated with this practice are
unknown. It is possible that academies increase the motivation of individuals to act as
place managers and to report disorderly conduct in problematic locations. Nevertheless,
citizens academies offer a unique method for police officials to have extensive one-on-
one contact with persons who are concerned with neighborhood problems and wish
greater community involvement. Additionally, this approach may provide an opportunity
for police organizations to build a foundation of support within the community.
Therefore, it is recommended that police administrators consider the implementation of a
citizens police academy, especially where community-relations have been strained.

Concluding Remarks

Although the preceding policy recommendations detail 12 types of policing
practices that may lead to crime reduction in police jurisdictions, the most effective
implementation of any of these methods may result from a process that begins with a
comprehensive assessment of organizational strengths and community needs. At present,
the extent to which these policing practices are interdependent is currently unknown.
Consequently, police administrators should make an effort to select practices that
complement one another. For example, aggressive enforcement approaches may best be
accompanied by efforts to increase citizen involvement to ensure that community~

relations are not damaged. Research pertaining to the situations and contexts in which

224

speciﬁc practices may be most effective is just beginning. However, analyses in the
current study clearly suggest that the future role of the police in crime reduction is

promising.

225

 

APPENDIX A

National Crime Trends (Summed Rates)

Figure l. Murder/Manslaughter

 

    

1®1Q1Q41%1% 2000

 

 

Figure 2. Robbery

 

 

02188888

 

19901a721994196m21m

 

226

 

Figure 3. Aggravated Assault

 

 

 

 

.0388???

 

 

 

 

 

Figure 4. Rape

 

 

 

E3858

 

 

 

O

 

l T

1%1Q1Q41%1% ZID

 

 

227

Figure 5. Burglary

 

 

 

 

 

 

 

 

 

1911921941%1% 21D

 

Figure 6. Auto Theft

 

 

08888888

 

19)1@1941%1% KID

 

228

 

 

APPENDIX B

Complete List of 80 Policing Tactics from the National
Survey of Community Policing Strategies

229

 

5:055:00
5:850:00
5:055:00
5:055:00
5:055:00
5:055:00
5:850:00
:0:0:0:0:00
5:850:00
5:850:00
5:850:00
5:850:00
5:850:00

8:00 :>:0
8:00 :>:0
8:00 :>:0
8:00 :50

>55: 00050

305: 00050

>:0>5 005:0

202:: 005:0

>:0>5 005:0

>:0>5 005:0

>:0>5 005:0

>:0>5 005:0

202:: 005:0

302:: 005:0
205: 005:0

808011.85

0:005 b50500 ::3 3:0:0w0: :005 0:00:00 :52

5:005:50 w5500w5 5 :0:000 800:”:0 :52

00:0 5 08:00: 55500 ::3 02:05:50.: 558: 80050 :002
50503000 .:0 00:0 5 0:050 50:50 080:":0 :52

0:555:30: 5 0:00:08 :00:.0:.:00: 00:00: 080:”:0 :002

:00 05:20: 350500 :0N:0:00:0 0 00: >000w<

050:: 500:5:0: 0:00:000 500:0.

00:>.:00 00:0: 0:00:0>0 0: 0:050 5050 0:00:08 500:0.

0:00: 5:005:50 055:0: 0: 0:050 005:0 0:00:50 5:090.
0:005 b55000 ::B 85:00:: 35:58 3:0:0w0: 0:5: 5:03:
0:00:n:0 5.: 5:050:00 5:05: 0 00 500: :00.: 0000 >000w<
50055000 058:0 0 00 5:0: :00: :000 50090.

0:00:00 5 50.50:: 00:00::0 we: 0:00:50 >000m<

000:0 50:: 5 05:0:0:> 0:00 :00 :50 005.50 0:00:.:0 :002
505005.50 0:00 5:08:05: 80: 5:05:55

0:55 :00 0w0:: :0:::00 0: 8:00 0:50:05: 800 5:090.

25:0 0550: ::0: 0: 505005.50 0:00 w5::0: 0000 5:09..
0:05:05 5:0w0 8:5 :0 .550 5.: :08» 283:0 5 0>:00 055:0
5050:: 5:03 500 5 0:050:00: 055:0

:m:>m.: m9>¢E0 :0 0:00:00 583:0 005:0 5 0>:00 055:0
:m>m.: DOOIMOmEOEZ :0 0::00000 b00:>:0 005:0 5 0>:00 055:0
50:30 00:0: 005:0 ::0::0 055:0

5:05 00:0: 0:: 55:5 0:00:00:0> 00 0>:00 000050

00000:: 05:25:: 5 0:05.05: 055:0

50050 30: :0.: 0800:: 5:00:00 5 00555: 000050

00:0: :050w0 0:505:50 30:8: ::0: 055:0

00:05:05: :00:.:0 0:00:08 ::0: 055:0

85:0: 55:0: 558: ::0: 055:0

 

.mm
.:m
.9.
.m:
0:
.mm
.:m
.:N
.mm
.m:
.N:
.3

K.

.:m
Sm
.om
.0:
.w\.
.5:
0:.
.m:
.2.
in:
.E.
.21
.0:
.w:
8:

 

5:: 5:800 >0>:0m :0055m Q: .:0>

 

8:005:86 :5: 00:00:. 555: cm .:0 :05 329000 005285 w50:0: 350500 ,:0 >05m :05:0 Z .: 030:.

230

 

000:0
000:0

000:0

000:0

000:0

00:2
8:0:05m:
85:05:
85:05:
85:05:
85:05.":
85:05.":
85:05:
85:05.":
85:05:
85:05m
85:05.”:
85:05.":
0:00:08 05:0
0:00:08 0550
0:00:08 05::0
0:00:08 05:0
0:00:08 05::0
5:8 5
5:2: 5
5:2: .5
5:8 :0
5:2: .5
5:808:00
5:808:00
5:058:00

05:80 :0 00050 :000:-:00::0::w:0: .0::02

05:80 :0 00050 :08:-:00::0::w:0: .:585.:00

008:0:505 :0::0 5.: 08% 0:0 w550:: :0w085\::0:::0: 8: 85w<
:00:.:0::w:0:= :0 3:505:50: 8 80050 0500 .:0 00:08:08:
80:0 :0 0:00: 050000 0: 80050 5000 .:0 5050508 :050
55.5: 0: ..0:000.. :0 080:0 558:0: :0 5:0: :0:0w0: 800 85m<
05:05:50: 858.55: :0::0:00 :.: 5:00

00:0 5 880 :0 005505: :53 58:08: :53 80050 :002
3:008 8:: 5:0 0:05 0: 558:0: 05:0 :00: 800 85w<
008: 0:0: 5:000 :0 5:0 5:5 w5000000 .:0 0:005 0 8: 8:0w<
005.: 0:00: we: 8:08:05: 8 5 8:005:80 85m<

:00 5:585 055 5585000 0 8: 859::

8% 0:0 0080508 5:05 0 8: 859:.

30.00 5:58: 0500:: 05:00:08 ::3 :0::0:w0:5 00:09:
08580 05:00:50 5505500 :55 00:88:: 8:09:
08000: 05:58: ..wd: 0:80 :0: 50:55 0000000: 05:05::<
80:0 800005 05:0 :0 0:: :0: 0: men:

05: :0050 08055 0: 0::00 .:0 5:85:55 :8 :0::00::08:0
5:50:58: .:0 00:0 0::00:wo0w 5.: 0:00:08 05:5 0885 5 :0 :00
:5808 .:0 00:0 5.: 0:00:08 055 50:50 80050 :002
00:000.: 5:: 0:00:08 055 558.0500: 0 8: >05w<
5:055:00 0:00:08 055 55:05:00 0 8: 85w<

80050 0: 28:08 0:05 0:00:08 055 :08: 5805005000
w50:00 5505500 w5::0::0:05: 5:555 0: >05w<

w50:00 5505500 5.: 0:00w 5:550 8:080

85:05.:0 w50:00 5505500 :50000 0 0. ”8085::0 307:
00005 55:0 .0555 :55 50:00:05: 00:00 ”00:00 >87:
00:0:0w0 ::05::0>0w 5:5 :55 0005805: 00:00 ”8:00 >87:
05:80 50:5 8:05:500 85w<

80:0 5:03 :00::0::w:0: 0 8: >05w<

80:00: 5:05:50 5:3 50500 855: 58505 5 :0 :00

 

231

 

w8>3m 80305
w8>3m 8030.5
w8>3m 80305
w8>3m 80305
w8>3m 8030.5
w8>3m 80305
w8>3m 80305
w8>3m 80305
w8>3m 8030.5
w8>3m 8030.5
w8>3m 8030.5
w8>3m 80305
w8>_0m 8030.5
w8>30m 80305
w8>3m 8030.5
w8>_0m 8030.5

0005

0005

0005

0005

0005

080308 00 0000 03 00 0003 M05380 000080080 000N008 000005
08030.8 00030033800 02000.. 00 00008 303 0003 000005

308 00 00000—00 00030 0000800000050 800 .088 0000 5 .00 80
80308 0200 00 000000000 >000w0 00030 088000 3000 0x08 5 ..0 80
080308 .8800 0008 20003 00 303 00030 008000 3000 00.08 5 .00 80
000000000 03 :05 080308 3033 00030 0808000 .000 0x08 3 00 8O
M8300 8030.8 008800 00 000008080 0900000 000 5 .00 80
080308 38000000 0000000 00 303 00000000 30000 0.000000 0002
080308 00.00 020000 000 @0008 00 000080 383 0003 0.000000 0002
080308 33800 0200 00 000000w0 >000 .050 303 0003 0.000000 0002
000 80300 80308 000080080 0 003 >000w<

080308 80080 383 .000 00 08000 00000080000032

0008000.. 000 0000000008 80308 8 808030008 0000800008
0000—0000 00 000000000000 80308 8 0000000 0800 >000w<

w8>30 8030.8 00.0 8800 0000080 000308 >000w< .

080308 0008033800 303 .000 00 000000008 00030.. >000w<
00000w000>8 .00 0000000000000 0080005

0003.000 200 .00 0000000000000 300085

000000003 00030033800 383 080800 000000003 .0005
000030033800 00 000 00 5030008000 88080800000 .00 0008800
00300030 .00 .0008 .030300 000000 00000 000080 0000 >000w<

.00
.00
.8
.N0
. 3
do
Sm
.Nm
.om
.50
.30
.mm
.NN
.m—
.3

.ov
.3“
.mm
.mm
.om

232

 

 

APPENDIX C

Individual Policing Survey Items - Signiﬁcantly Related to Crime Reduction.

Table l. Larceny

 

 

 

 

 

Var. ID Summarized Survey Question Item
68. Citizens help evaluate ofﬁcer performance
69. Citizens review complaints against police
70. Citizens participate in selection process
73. Citizens attend citizens police academy
74. Agency coordinates citizen patrols
79. Citizens prepare work agreements
56. Most officers enforce civil and regulatory codes
57. Managers can redesign organization to support problem-solving
29. Mobile, neighborhood based ofﬁces or stations
34. Agency has specialized crime prevention unit
36. Agency participates in interagency drug task force
Table 2. Burglary
Var. ID Summarized Surveruestion Item
68. Citizens help evaluate ofﬁcer performance
73. Citizens attend citizens police academy
38. Patrol boundaries coincide with neighborhood boundaries
9. Fixed assignment of patrol ofﬁcers to speciﬁc beats or areas
26. Agency conducts citizen surveys to evaluate police service
14. Agency provides speciﬁc training for problem-solving
15. Agency trains and citizens in problem identiﬁcation and resolution
50. Most ofﬁcers work with citizens to identify and resolve area problems
57. Managers can redesign organization to support problem-solving
66. Citizens work with police to resolve neighborhood problems
79. Citizens prepare agreements specifying work to be done on problems
56. Most ofﬁcers enforce civil and regulatory codes
34. Agency has Specialized crime prevention unit
36. Agency participates in interagency drug task force
43. Centralized crime analysis function

 

233

Table 3. Vehicle Theft

 

 

 

 

 

Var. ID Summarized Survey Question Item
56. Most ofﬁcers enforce civil and code violations in their areas
43. Agency has a centralized crime analysis function/function
34. Agency has a specialized crime prevention unit
53. Most ofﬁcers work regularly with detectives on cases in area
9. Fixed assignment of patrol ofﬁcers to speciﬁc beats or areas
10. Designation of some ofﬁcers as "community" or "neighborhood
38. Patrol boundaries coincide with neighborhood boundaries
39. Physical decentralization of ﬁeld services
1 1. Agency used foot patrol as a speciﬁc assignment
13. Agency holds regularly scheduled meetings with community groups
25. Agency conducts citizen surveys to determine community needs
51. Most ofﬁcers assist in organizing community
52. Most ofﬁcers teach residents how to address community problems
22. Interagency involvement in problem identiﬁcation and resolution
57. Cpt or Lt can redesign organization to support problem solving
60. Cpt or Lt make ﬁnal decision about which problems will be addressed
61. Cpt or Lt make ﬁnal decision about how to handle most comm. problems
62. Cpt or Lt make ﬁnal decision about agency resources to solve problem
63. Cpt or Lt elicit input from ofﬁcers/deputies about solutions to prob
66. Citizens work with police to resolve neighborhood problems
79. Citizens prepare agreements specifying work to be done on problems
Table 4. Robbery
Var. ID Summarized Survey Question Item
70. Citizens participate in selection process
22. Interagency involvement in problem identiﬁcation and resolution
57. Cpt or Lt can redesign organization to support problem solving
60. Cpt or Lt make ﬁnal decision about which problems will be addressed
61. Cpt or Lt make ﬁnal decision about how to handle most comm. problems
79. Citizens prepare agreements specifying work to be done on problems
56. Most ofﬁcers enforce civil and regulatory codes
34. Agency has specialized crime prevention unit
36. Agency participates in interagency drug task force
43. Agency has a centralized crime analysis function/function
9. Fixed assignment of patrol ofﬁcers to speciﬁc beats or areas
29. Mobile, neighborhood based ofﬁces or stations
1. New policy: Police interactions with other government agencies
32. Command or decision-making responsibility is tied to neighborhoods

 

234

 

Table 5. Rape

 

 

 

 

 

 

 

 

Var. ID Summarized Survey Question Item
7. Agency conducts drug education program in schools
40. Physical decentralization of investigations
66. Citizens work with police to resolve neighborhood problems
22. Interagency involvement in problem identiﬁcation and resolution
35. Multidisciplinary teams to deal with special problems
47. Most ofﬁcers work with other city agencies to solve neighb. problems
Table 6. Aggravated Assault
Var. ID Summarized Survey Question Item
51. Most ofﬁcers assist in organizing community
8. Drug tip hot line or Crime Stoppers program
53. Most ofﬁcers work regularly with detectives on cases in area
29. Mobile, neighborhood-based ofﬁces or stations
30. Agency uses drug-free zones around schools, parks, or churches
52. Most ofﬁcers teach residents how to address community problems
Table 7. Murder
Var. ID Summarized Survey Question Item
56. Most ofﬁcers enforce civil and code violations in their areas
46. Most ofﬁcers make door-to-door contacts in neighborhoods
43. Agency has a centralized crime analysis function/ﬁnnetion
29. Mobile, neighborhood-based offices or stations
79. Citizens prepare agreements specifying work to be done on problems

 

235

 

APPENDIX D

Reported Changes in Polcing Tactics from 1993 to 1997

236

Table 1. Changes in Each of the 80 Policing Practices

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Value Frequency Percent Total Tot Pet
1 Policy - Interactions with other agencies
-1 144 16.59 144 16.59
0 528 60.83 672 77.42
1 196 22.58 868 100
2 Policy - Interactions with citizens
-I 146 16.82 146 16.82
0 511 58.87 657 75.69
1 211 24.31 868 100
3 Policy — Procedures to deal with neighborhood problems
-1 148 17.05 148 17.05
0 517 59.56 665 76.61
1 203 23.39 868 100
4 Policy - To support community policing goals
-1 50 5.76 50 5.76
0 569 65.55 619 71.31
1 249 28.69 868 100
5 Policy - Community policing goals published
-1 73 8.41 73 8.41
O 532 61.29 605 69.7
1 263 30.3 868 100
6 Calls prioritized to increase ofﬁcer time
-1 117 13.48 117 13.48
0 526 60.6 643 74.08
1 225 25.92 868 100
7 Drug education program in schools
-1 23 2.65 23 2.65
0 795 91.59 818 94.24
1 50 5.76 868 100
8 Drug tip hotline or crime stoppers program
-1 81 9.33 81 9.33
0 668 76.96 749 86.29
1 119 13.71 868 100
9 Fixed assignment of patrol ofﬁcers
-1 76 8.76 76 8.76
0 621 71.54 697 80.3
1 171 19.7 868 100
10 Assign some ofﬁcers as “communigy” ofﬁcers
-1 71 8.18 71 8.18
0 480 55.3 551 63.48
1 317 36.52 868 100

 

237

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Value Frequency Percent Total Tot Pct
11 Foot patrol as a speciﬁc assignment
-1 112 12.9 112 12.9
0 631 72.7 743 85.6
1 125 14.4 868 100
12 Foot patrol as periodic expectation for patrol ofﬁcers
-1 112 12.9 112 12.9
0 549 63.25 661 76.15
1 207 23.85 868 100
13 Regularly scheduled meetings with community groups
-1 77 8.87 77 8.87
0 583 67.17 660 76.04
1 208 23.96 868 100
14 Function - Speciﬁc training for problem solving
-1 68 7.83 68 7 .83
0 451 51.96 519 59.79
1 349 40.21 868 100
15 Function - Problem Solving training for citizens
-1 89 10.25 89 10.25
0 467 53.8 556 64.06
1 312 35.94 868 100
16 Regular radio or television “spots” to inform
-l 150 17.28 150 17.28
0 535 61.64 685 78.92
1 183 21.08 868 100
17 Alternative response methods for calls
-1 92 10.6 92 10.6
0 588 67.74 680 78.34
1 188 21.66 868 100
18 Function - Landlord/M gr. training for order maint.
-1 66 7.6 66 7.6
0 606 69.82 672 77.42
1 196 22.58 868 100
19 Function - Building code enforcement to remove crime
-1 76 8.76 76 8.76
0 559 64.4 635 73.16
1 233 26.84 868 100
20 Function - Use of regulatory codes to combat crime
-1 108 12.44 108 12.44
0 530 61.06 638 73.5
1 230 26.5 868 100

 

238

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Value Frequency Percent Total Tot Pet
21 Function - Geographically based crime analysis
-1 148 17.05 148 17.05
0 534 61.52 682 78.57
1 186 21.43 868 100
22 Interagency involvement in problem resolution
-1 111 12.79 111 12.79
0 517 59.56 628 72.35
1 240 27.65 868 100
23 Function - Integration with community corrections
-1 115 13.25 115 13.25
0 561 64.63 676 77.88
1 192 22.12 868 100
24 Function - Integration with alternative dispute resolution
-1 84 9.68 84 9.68
O 636 73.27 720 82.95
1 148 17.05 868 100
25 Citizen surveys to determine needs and priorities
-1 59 6.8 59 6.8
O 530 61.06 589 67.86
1 279 32.14 868 100
26 Citizen surveys to evaluate police services
-1 56 6.45 56 6.45
0 577 66.47 633 72.93
1 235 27.07 868 100
27 Victim assistance program
-1 95 10.94 95 10.94
0 576 66.36 671 77.3
1 197 22.7 868 100
28 Permanent neighborhood based ofﬁces or stations
-1 45 5.18 45 5.18
O 592 68.2 637 73.39
1 231 26.61 868 100
29 Mobile, neighborhood based ofﬁces or stations
-1 45 5.18 45 5.18
0 701 80.76 746 85.94
1 122 14.06 868 100
30 Drug free zones around schools, parks, or churches
-1 71 8.18 71 8.18
O 610 70.28 681 78.46
1 187 21.54 868 100

 

239

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Value Frequency Percent Total Tot Pct
31 Police/youth programs (PAL, School liaison etc.)
-1 73 8.41 73 8.41
0 616 70.97 689 79.38
1 179 20.62 868 100
32 Command or decision making tied to neighborhoods
-1 135 15.55 135 15.55
0 508 58.53 643 74.08
1 225 25.92 868 100
33 Specialized community relations unit
-1 116 13.36 116 13.36
0 560 64.52 676 77.88
1 192 22.12 868 100
34 Specialized crime prevention unit
-1 107 12.33 107 12.33
0 637 73.39 744 85.71
1 124 14.29 868 100
35 Multidisciplinary teams to deal with special problems
-1 135 15.55 135 15.55
0 531 61.18 666 76.73
1 202 23.27 868 100
36 Interagency drug task force
-1 87 10.02 87 10.02
0 696 80.18 783 90.21
1 85 9.79 868 100
37 Function - Interagency code enforcement
-1 114 13.13 114 13.13
0 525 60.48 639 73.62
1 229 26.38 868 100
38 Patrol boundaries coincide with neighborhoods
-1 128 14.75 128 14.75
0 523 60.25 651 75
1 217 25 868 100
39 Physical decentralization of ﬁeld services
-1 67 7.72 67 7.72
O 622 71.66 689 79.38
1 179 20.62 868 100
40 Physical decentralization of investigations
-1 61 7.03 61 7.03
0 710 81.8 771 88.82
1 97 11.18 868 100

 

240

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Value Frequency Percent Total Tot Pet
41 Fixed shifts (changing no more often than annually)
-1 163 18.78 163 18.78
0 476 54.84 639 73.62
1 229 26.38 868 100
42 Access to other city or county databases
-1 70 8.06 70 8.06
O 654 75.35 724 83.41
1 144 16.59 868 100
43 Centralized crime analysis unit/function
-1 140 16.13 140 16.13
0 573 66.01 713 82.14
1 155 17.86 868 100
44 Decentralized crime analysis unit/function
-1 45 5.18 45 5.18
O 765 88.13 810 93.32
1 58 6.68 868 100
45 Specialized problem solving unit
-1 104 11.98 104 11.98
0 541 62.33 645 74.31
1 223 25.69 868 100
46 Patrol - Make door-to-door contacts in neighborhoods
-1 161 18.55 161 18.55
0 573 66.01 734 84.56
1 134 15.44 868 100
47 Patrol - Work with other city agencies ogroblems
-l 163 18.78 163 18.78
0 549 63.25 712 82.03
1 156 17.97 868 100
48 Patrol - Conduct surveys in area of assignment
-1 68 7.83 68 7.83
0 750 86.41 818 94.24
1 50 5.76 868 100
49 Patrol - Develop familiarity with communipl leaders
-1 175 20.16 175 20.16
0 535 61.64 710 81.8
1 158 18.2 868 100
50 Patrol - Work with citizens to resolve problems
-1 165 19.01 165 19.01
0 525 60.48 690 79.49

 

1 178 20.51 868 100

241

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Value Frequency Percent Total Tot Pet
51 Patrol - Assist in organizing the community
-1 83 9.56 83 9.56
0 704 81.11 ~ 787 90.67
1 81 9.33 868 100
52 Patrol - Teach residents problem solving
-1 99 11.41 99 11.41
0 666 76.73 765 88.13
1 103 1 1.87 868 100
53 Patrol - Work regularly with detectives in neighborhood
-1 201 23.16 201 23.16
0 495 57.03 696 80.18
1 172 19.82 868 100
54 Patrol - Conduct crime analysis for area of assignment
-1 87 10.02 87 10.02
0 722 83.18 809 93.2
1 59 6.8 868 100
55 Patrol - Meet regularly with community groups
-1 81 9.33 81 9.33
O 711 81.91 792 91.24
1 76 8.76 868 100
56 Patrol - Enforce civil and code violations in area
-1 175 20.16 175 20.16
0 521 60.02 696 80.18
1 172 19.82 868 100
57 Managers - Redesign organization for problem solving
-1 96 11.06 96 11.06
0 464 53.46 560 64.52
1 308 35.48 868 100
58 Managers - Regular contact with community leaders
-1 127 14.63 127 14.63
0 534 61.52 661 76.15
1 207 23.85 868 100
59 Managers - Establish interagencyrelationships
-1 108 12.44 108 12.44
0 592 68.2 700 80.65
1 168 19.35 868 100
60 Managers - Final decision, problems to be addressed
-1 136 15.67 136 15.67
0 495 57.03 631 72.7

 

1 237 27.3 868 100

242

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Value Frequency Percent Total Tot Pet
61 Managers - Final decision, how to handle problems
-1 131 15.09 131 15.09
0 491 56.57 622 71.66
1 246 28.34 868 100
62 Managers - Final decision, agency resources, problems
-1 109 12.56 109 12.56
0 463 53.34 572 65.9
1 296 34.1 868 100
63 Managers - Elicit input from deputies, problem solving
-1 189 21.77 189 21.77
0 550 63.36 739 85.14
1 129 14.86 868 100
64 Managers - Crime analysis for area of assignment
-1 192 22.12 192 22.12
0 478 55.07 670 77.19
1 198 22.81 868 100
65 Citizens - Participate in neighborhood watch program
-1 38 4.38 38 4.38
0 758 87.33 796 91.71
1 72 8.29 868 100
66 Citizens - Work to resolve neighborhood problems
-1 59 6.8 59 6.8
O 574 66.13 633 72.93
1 235 27.07 868 100
67 Citizens - Help develop policing policies
-1 85 9.79 85 9.79
0 653 75.23 738 85.02
1 130 14.98 868 100
68 Citizens - Help evaluate ofﬁcer performance
-1 59 6.8 59 6.8
0 721 83.06 780 89.86
1 88 10.14 868 100
69 Citizens - Help review complaints against police
-1 57 6.57 57 6.57
O 741 85.37 798 91.94
1 70 8.06 868 100
70 Citizens - Participate in selection process for ofﬁcers
-1 71 8.18 71 8.18
0 688 79.26 759 87.44
1 109 12.56 868 100

 

243

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Value Frequency Percent Total Tot Pet
71 Citizens - Participate in promotion process for ofﬁcers
-1 79 9.1 79 9.1
0 690 79.49 769 88.59
1 99 11.41 868 100
72 Citizens - Serve as volunteers within the agency
-1 100 11.52 100 11.52
0 582 67.05 682 78.57
1 186 21.43 868 100
73 Citizens - Attend citizens police academy
-1 35 4.03 35 4.03
0 523 60.25 558 64.29
1 310 35.71 868 100
74 Citizens - Serve in coordinated citizens patrols
-1 75 8.64 75 8.64
0 598 68.89 673 77.53
1 195 22.47 868 100
75 Citizens - Advisory councils at neighborhood level
-1 149 17.17 149 17.17
0 505 58.18 654 75.35
1 214 24.65 868 100
76 Citizens - Advisory councils at city-wide level
-1 117 13.48 117 13.48
0 516 59.45 633 72.93
1 235 27.07 868 100
77 Citizens - Participate in court watch program
-1 72 8.29 72 8.29
O 731 84.22 803 92.51
1 65 7.49 868 100
78 Citizens - Advisory group for chief or other managers
-1 121 13.94 121 13.94
0 538 61.98 659 75.92
1 209 24.08 868 100
79 Citizens - Prepare work agreements, police problem solv
-1 94 10.83 94 10.83
0 655 75.46 749 86.29
1 119 13.71 868 100
80 Policy - Currently implementing community policing
-1 44 5.07 44 5.07
O 446 51.38 490 56.45
1 378 43.55 868 100

 

 

244

REFERENCES

Allen, R. (1996). Socioeconomic conditions and property crime: a comprehensive review
and test of the professional literature. American Journal of Economics and
Sociology, 55, 293-308.

Allison, P. (2005). Fixed Effects Regression Methods for Longitudinal Data Using SAS.
Cary: SAS Institute.

Bayley, D. H. (1994). Police for the Future. New York: Oxford University Press.

Becker, G. (1968). Crime and Punishment: an Economic Approach. The Journal of
Political Economy, 76, 169-217.

Benson, B., Kim-Iljoong & Rasmussen, D. (1994). Estimating deterrence effects: a
public choice perspective on the economics of crime literature. Southern-
Economic-Journal, 61, 161-168.

Benson, B. & Rasmussen, D. (1998). Deterrence and public policy: Trade-offs in the
allocation of police resources. International Review of Law and Economics, 18,
7 7-100.

Benson, 8., Sebastian, I. & Rasmussen, D. (2001). The impact of drug enforcement on
crime: an investigation of the opportunity cost of police resources. Journal of
Drug Issues, 31, 989-1006.

Bernard, T. (1999). Juvenile Crime and the Transformation of Juvenile Justice: Is There
a Juvenile Crime Wave? Justice Quarterly, 16, 337-356.

Blumstein, A. & Rosenfeld, R. (1998). Explaining recent trends in US homicide rates.
Journal of Criminal Law and Criminology, 88, 1175-1216.

Blumstein, A. & Wallman, J. (eds.) (2000). The Crime Drop in America. Cambridge
University Press: Cambridge, UK.

Blumstein, A., Steinberg, L., Bell, C. & Berger, M. (March, 2002). Perspectives on
Crime and Justice: 2000-2001 Lecture Series. National Institute of Justice:
Washington, DC.

Bonczar, T. (August, 2003) Special Report: Prevalence of Imprisonment in the US.
Population, 1974-2001. NCJ 197976 US. Department of Justice Ofﬁce of Justice

Programs.

Bowers, K., Johnson, S. & Pease, K. (2004). Prospective hot-spotting - The future of
crime mapping? British Journal OfCriminology, 44, 641-658.

245

Braga, A. (1997). Solving Violent Crime Problems: An Evaluation of the Jersey City
Police Department's Pilot Program to Control Violent Places. Dissertation.
Rutgers The State University of New Jersey - Newark, 1997, 200 Pages.

Braga, A. (2002). Problem Oriented Policing and Crime Prevention. Criminal Justice
Press: Monsey, NY.

Braga, A., Weisburd, D., Waring, E. Mazerolle, L. (1999). Problem Oriented Policing in
Violent Crime Places: a Randomized Controlled Experiment. Criminology, 37,
541-580.

Braga, A. (November, 2001). Effects of Hot Spots Policing on Crime. Annals of the
American Association of Political and Social Sciences, 5 78, 104-125.

Bratton, W. & Knobler, P. (1998). Turnaround: How America 's T op Cop Reversed the
Crime Epidemic. Random House: New York.

Breci, M., & Erickson, T. (1998). Community policing: The process of transitional
change. FBI Law Enforcement Bulletin, 67, 16-21.

Bureau of Justice Administration. (2002). Program Brief. FY 2002 Local Law
Enforcement Block Grants Program. Washington, DC: Bureau of Justice
Assistance. Available online:_http://www.ncjrs.gov/pdfﬁles 1 /bja/1 93494.pdf

Bureau of Justice Statistics (2005a). Jail Statistics Data. Washington, DC: US.
Department of Justice. Website. Available online:
http://www.ojp.usdoj.gov/bjs/glance/tables/corthab.htm

Bureau of Justice Statistics (2005b). Prison Statistics Data. Washington, DC: US.
Department of Justice. Website. Available online:
http://www.ojp.usdoj.gov/bjs/prisons.htm

Burgess, E. (1925). The grth of the city. In Park, R., Burgess, E. & McKenzie, D.
(Eds). The City. Chicago: University of Chicago Press.

Bursik, R. & Grasmick, H. (1993). Neighborhoods and Crime: The Dimensions of
Effective Community Control. San Francisco: Jossey-Bass Inc.

Carr, P. (2003). The New Parochialism: The Implications of the Beltway Case for
Arguments Concerning Informal Social Control. American Journal of Sociology,
108, 1249-1291.

Carter, D. L. (1995). Politics and community policing: Variables of change in the
political environment. Public Administration Quarterly, 19, 6-11.

246

   
 
 
  
 
 
  
  
   
  
   
 
 
  
 
 
  
  
 
  
   
  

w . - ~. “1 ~-\" ‘ . 1‘ 1\. 11‘! uniﬁoz _

« 1 - . ' » m'h“. Wriwum
'17....» .. ' ~ 1 :1- 11111-111113 1 '1
v, 1
Vi « * ' 4.1" ' .r"‘."r‘\'1\do‘\q .(-'1_
("r‘ .vsmoM 41

. 1 1'
1 ' 11))I"‘l. ' ' ' ~11 'l/ .‘ ‘.11‘.‘- t .d I Jr‘-
'

'. :1 . ‘ 11‘ .. 530133th 1'-
‘

,1 1:1‘1 ' 1 ' .1' 1 .‘lllllc' ..1 .

vwmzr-rouk ' -

‘ ﬂ ,1 .. .- mu. ‘1 “Morita
.1 - {may :4 '1“nt

'v
1... 1“ n 1 - g .. ~ 1‘ 11 .-,~-vr1.1'.111u:111.l£-

,0 _'.~.1‘1.‘.A '1':x\\8."‘ ~ '
'. ' " ’6

0

111,111. .I w". I 1 ~ . 11 ,1 - rucmmrbA ,,
I ‘ "'UurJIl'l .I' 11'I__1"l. Y I‘ VJ" ll.l'..H; 3:1“18’ " t'- ‘v

..~'~i'~"'1~m' 1 11.1% . ._ 1 13:” nulnv. ~)idalva.

-" I. ‘U 1.1 '1‘: vr'. ':. -" .111- 1 .1- . -..‘ 1,1 “1".) Zblltllﬂa ' ‘ .
.“‘.' "11‘ .W-I‘luW ‘;).1rul.leI--'
MM "112;: 1~" -- 1 - ' “1121.29 =91 M.,-1 {ohm qio. ’

R I )C‘ ,r!11::;|;11f-1,.’." _' 1‘- 1 .40... .01.", 14201111011112.12.
21:111'1' ’:l!'1-.l1':'.1 '11:;rtv'lv' "1.1113900
'rrr.i,.-rrv1.:1" rg1"1I-,«+«_:"b¢u_qtn.

.U rum-Mum .9. 3 330211113 . >1 01111 1 n! m:- 1.11111 191':- mrg NIT .(
«Mn‘l would?!“ 1371311" 1.0309111).va

0mm smn')l1cs abmrtmdrtrnm .1091) .H .10.

: "j “0684000110021an .1112 hum-1" (3111111....

xi sow 011100011011qu 0.11 marlnirlwln‘l will a“
.m‘pm Wk .lmmo) 151308 lsnnoinl gnmmnd)
IQII

$3.54 .

.9};

all numb WV :gn'zurloq \{tulumrmn blur enitiloﬂ ‘ - ‘3‘"
.II-b.Q\ 3:}!me rrn'nmn.1n'rm\1l_-JAM . 1- .1. . . ; '

\'_.V
\ .rIv."

Carter, D. (November, 2004). Law Enforcement Intelligence: A Guide for State, Local
and Tribal Law Enforcement Agencies. Washington, DC: Ofﬁce of Community
Oriented

Policing Services, US. Department of Justice.

Chaiken, M. R. & Chaiken, J. M. (1982). Varieties of criminal behavior. Santa Monica,
CA: Rand Corp.

Challinger, D. (1991) Less telephone vandalism: How did it happen? Security Journal,
2, 111-119.

Cheh, M. (1998). Civil remedies to control crime: Legal issues and constitutional
challenges. Crime Prevention Studies, 9, 45-66.

Clarke, R. (2001). A Situational Crime Prevention Theory: A Foundation for Future
CPTED Practice? 6th International CPTED Conference, Brisbane.

Cliﬁon, W. (1987). Convenience store robberies in Gainesville, Florida: An intervention
strategy by the Gainesville Police Department. Gainesville, FL: Gainesville
Police Department.

Cohen, L. & Felson, M. (1979). Social change and crime rate trends: A routine activity
approach. American Sociological Review, 44, 588-608.

Cohen, L., Felson, M., & Land, K. (1980). Property crime rates in the United States: A
macrodynamic analysis, 1947-1977: With ex ante forecasts for the mid-1980's.
American Journal of Sociology, 86, 90-118.

Cohen, L., Kluegel, J., & Land, K. (1981). Social inequality and predatory criminal
victimization: An exposition and a test of a formal theory. American Sociological
Review, 46, 505-524.

Cohen, L. & Land, K. (1987). Age structure and crime: Symmetry vs asymmetry and
the projection of crime rates through the 19903. American Sociological Review,
52, 170-183.

COPS Ofﬁce (1998). UHP Grant Owners Manual. Ofﬁce of Community Oriented
Policing Services. Washington, DC. Available online:
http://www.cops.usdoj.gov/mime/open.pdf?1tem=1053

COPS Ofﬁce (2000). UHP Fact Sheet. Ofﬁce of Community Oriented Policing
Services. Washington, DC. Available online:
http://www.cops.usdoj.gov/mime/open.pdf'?ltem=330

COPS Ofﬁce (2005). Website: (MORE grants): Ofﬁce of Community Oriented Policing
Services. Available online: http://www.cops.usdoj.gov/default.asp?1tem=55

247

 

Cordner, G. (1988). A problem oriented approach to community policing. In Greene, J .,
& Mastrofski, S. (Eds), Community Policing: Rhetoric or Reality (2"d ed. ). New
York: Praeger Publishers.

Gary C. & Biebel, E. (2005). Problem-oriented policing in practice. Criminology &
Public Policy, 4, 155-181.

Corman, H. & Mocan, N. (2002). Carrots, sticks and broken windows. Working paper,
National Bureau of Economic Research. Cambridge: NBER.

Crank, J. (2003). Institutional theory of police: A review of the state of the art.
Policing, 26, 186-208.

Cullen, F. & Agnew, R. (2003). Criminological Theory: Past to Present. Los Angeles:
Roxbury Publishing.

Cunniff, M. (March 14, 2002). Law Enforcement Federal Grant Experience: Byme to
COPS. Urban Institute Publication. Available online: http://www.urban.org

DeJong, C. (1997). Survival analysis and speciﬁc deterrence: integrating theoretical and
empirical models of recidivism. Criminology, 35, 561-576.

Di Tella, R. & Schargrodsky, E. (2004). Do Police Reduce Crime? Estimates Using the
Allocation of Police Forces after a Terrorist Attack. American Economic Review,
I. 115-133.

Dunworth, T. & Saiger, A. (1997). National Assessment of the Byme Formula Grant
Program, Washington, DC: National Institute of Justice Research Report,
December 1997.

Eastern Kentucky University (2003). National Assessment of Technology and Training
for Small and Rural Law Enforcement Agencies (N ATTS): A Descriptive
Analysis. NIJ: Washington, DC.

Eck, J. (1994). Drug Markets and Drug Places: A Case-Control Study of Spatial
Structure of Illicit Drug Dealing. Dissertation. College Park: University of
Maryland.

Eck, J ., Weisburd, D. (1995). Crime Places in Crime Theory. In Eek & Weisburd, eds.,
Crime and Place. Crime Prevention Studies, 4.

Eck, J. & Maguire, J. (2000). 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.

248

 

 

   
  
  
   
   
   
  
    
   
   
  
    
 
   
   

1"‘IO'H‘1I‘, A .
. 1.1),? '
*‘uumng‘l 1..
1111 .3

K . '
1 ,r , ~

It
.

‘.-

f

D‘
: ll‘r) _, mo" 1
. ‘ ‘ Nev)“ I i .-
V

l

'fmriml .
e

.V" It‘ll
Flt-LP: Q‘

1 .
- 1: p.011. .‘_1 {-02:51 8 ,' .
311.01:th '

O
« .1; 1211: Hum .
1' ruimutrrlnth'

'2;,"1_‘ " ., ‘1
i *- ‘ t ' 0 ﬁrm » ”11.111: war/1112 I“

J U
' "‘ ., .1111~.-.11') 9 1.1x1m01‘21'1.1:1b0m -‘.'_
.1110“ .11.'r:1"11u- ‘ 1.: ' .‘1' '1 1 "L1 l {I Inamba"
\ 1'1-".\'.V“...’\ l‘ifi'n‘. .m‘ H.311" - WY ' ' L 5‘1ﬁ;.'11.‘;]0:tll')q.30 ' '
ﬁx
1.1. rim-'1" '.1.rr-1rn;.’ ,71'51) .r'. .01-5133 ‘5 ’1’

..1'1 . 1'.
I f

.l‘luqvﬂ flower-.131 opium/tr- vii-5.1411 {tux-11:0! J (1 .nutgnrduw

r111;11)1.111rmr11 11mm
.I‘OQI' ,. ,' ,
, o

l ' I

I.

at

amniml 11m: wolomknl '11, 1101312110031 11.1111L1M (€005) (1110:1va .
11'va A 121713110.“ng 111-1110061113 W101 lmuil bu ‘
.10 .norgnirleaw :UW

.. r
§
”MW‘“ 111 «muff :1me hi aual‘l 0"th IBM” .0 -
.b .226th nomwnﬂ wh') .00-N '

(a
Mwmvmapmmdwoovm 100091 ""
mmamwrbm0 110011111121 “1.391111110511101 , '
.wﬂmvmumbhdm 01111151111000 -~ _

\
at: .

M 101111. lemma-501') 1~. 21.-mm and but. awn
Vow zit-1 0.51103 .m'rmmeuﬂ grain-10 9,1110 full“ '

p

u -

Eek, J. (1998). 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, DC: National Institute of Justice.

Eek, J. & Spelman, W. (1987). Problem-Solving Problem-Oriented Policing In Newport
News. Washington DC: National Institute of Justice.

Eek, J. & Wartell, J. (1998). Improving the management of rental properties with drug
problems: A randomized experiment. In Mazerolle, L. & Roehl, J. (Eds). Civil
Remedies and Crime Prevention: An introduction. Monsey, NY: Criminal Justice
Press.

Ehrlich, I. (1973). Participation in Illegitimate Activities: A theoretical and Empirical
Investigation. Journal of Political Economy, 3, 521-565.

Federal Bureau of Investigation. (2002). Crime in the United States: 2002 Uniform
Crime Reports. Washington, DC: US. Department of Justice.

Field, S. (2002). Examining The Incivilities Thesis: A Spatial and Temporal Analysis of
the Relationship Between Public Order Crime And More Serious Crime.
Dissertation. University of Texas at Austin.

Fox, J. (March, 1996). Trends in Juvenile Violence: A Report to the United States
Attorney General on Current and Future Rates of Juvenile Offending. Bureau of
Justice Statistics, Washington, DC.

Fox, J. (2000). Demographics and US. Homicide. In Blumstein, A. & Wallman, J.
(eds.). The Crime Drop in America. Cambridge University Press: Cambridge,
UK.

Frees, E. (2004). Longitudinal and Panel Data: Analysis and Applications in the Social
Sciences. Cambridge University Press: Cambridge UK.

Fritsch, E., Caeti, T. & Taylor, R. (2003). Gang suppression through saturation patrol
and aggressive curfew and truancy enforcement: A quasi-experimental test of the
Dallas Anti-Gang Initiative. In Decker, S. (Ed). Policing Gangs and Youth
Violence. Toronto: Wadsworth.

Government Accountability Ofﬁce (GAO). (2005a). Interim Report on the Effects of
COPS Funds on the Decline in Crime During the 1990’s.” GAO-05-699R
(Washington, DC: June, 2005).

Government Accountability Ofﬁce (GAO). (2005b). Community Policing Grants:

COPS Grants Were a Modest Contributor to Declines in Crime During
the1990’s.” GAO-05-983 (Washington, DC: October, 2005).

249

Goldstein, H. (1979). Improving policing: a problem oriented approach. Crime and
Delinquency, 25, 236-258.

Golub, A., Johnson, B. & Taylor, A. (2002) The validity of arrestees' self-reports:
Variations across questions and persons. Justice Quarterly, 19, 477-502.

Gottfredson, Michael R. and Travis Hirschi. (1990). A general theory of crime. Stanford:
Stanford University Press.

Granger, C. (1969). Investigating causal relations by econometric models and cross-
spectral methods. Econometrics, 39, 199-211.

Green, L. (1995). Cleaning up Drug Hotspots in Oakland, California: the Displacement
and Diffusion Effects. Justice Quarterly, 12, 737-754.

Green, J. (1999). Zero Tolerance: A Case Study of Police Policies and Practices in New
York City. Crime & Delinquency, 45, 171-187.

Greenburg, S., Rohe, W., & Williams, J. (1982). The relationship between informal
social control, neighborhood crime and fear: a synthesis and assessment of the
research. Paper presented at the annual meeting of the American Society of
criminology, Toronto.

Grogger, J. (2000). An Economic Model of Recent Trends in Violence. In Blumstein, A.
& Wallman, J. (eds.). The Crime Drop in America. Cambridge University Press:
Cambridge, UK.

Grogger, J. (2002). The effects of civil gang injunctions on reported violent crime:
Evidence from Los Angeles County. Journal of Law and Economics, XIV, 69-23.

Hannon, L. & Knapp, P. (2003). Reassessing nonlinearity in the urban
disadvantage/violent crime relationship: An example of methodological bias
from log transformation. Criminology, 41, 1427-1448.

Hickman, M. & Reaves, B. (2001). Local Police Departments 1999. US Department of
Justice, Bureau of Justice Statistics. May 2001. NCJ 186478.

Homey, J ., Osgood, D. & Marshall, I. (1995). Criminal careers in the short term: intra-
individual variability in crime and its relation to local life circumstances.
American sociological review, 60, 655-673.

Hunter, A. (1985). Private, Parochial and Public Social Orders: The Problem of Crime
and Incivility in Urban Communities. Pp. 230—42 in Suttles, G. & Zald, M. (Eds).
The Challenge of Social Control. Norwood, N.J.: Ablex.

250

ImrohyoroGlu, A. Merlo, P. Rupert (2004). What accounts for the decline in crime?
International Economic Review, 45, 707-729.

J esilow, P., Meyer, J ., Parsons, D. et a1. (1998). Evaluating problem-oriented policing: a
quasi-experiment . Policing-An-International-Journal-of-Police-Strategies-and-
Management, 21, 449-464.

Johnson, C., & Roth, J. (2003). The COPS Program and the Spread of Community
Policing, 1995-2000. Washington, DC: The Urban Institute.

Kelling, G. & Sousa, W. (December, 2001). Do Police Matter? An Analysis of the Impact
of New York City's Police Reforms. Civic Report #22. Manhattan Institute.

Kelling, G. & Coles, C. (1996). Fixing Broken Windows: Restoring Order and Reducing
Crime in Our Communities. New York: Free Press.

Kelling, G., & Moore, M. (1991). From political to reform to community: The evolving
strategy of police. In Greene, J ., & Mastrofski, S. (Eds), Community Policing:
Rhetoric or Reality (2"" ed. ). New York: Praeger Publishers.

Kelling, George, Anthony Pate, Duane Dickman, and Charles Brown. 1974. The Kansas
City Preventive Patrol Experiment: A Technical Report. Washington, DC: Police
Foundation.

Kennedy, D. (July, 1998). Pulling Levers: Getting Deterrence Right. National Institute
of Justice Journal.

Kleiman, M. (1997). Coerced abstinence: A neopatemalist drug policy initiative. In
Mead (Ed). The New Paternalism: Strategic Approaches to Poverty.
Washington, DC: National Institute of Justice.

Kleiman, M. (2000). Crime Control Policy In California. California Policy Options.
Available Online: http://www.spa.ucla.edu/faculty/kleiman/community.pdf

Koper, C. (1995). Just enough police presence: reducing crime and disorder by
optimizing patrol time and crime hotspots. Justice quarterly, 12, 645-672.

Kovandzic, T. & Sloan, J. (2002). Police levels and crime rates revisited - A county-level
analysis from Florida (1980-1998). Journal Of Criminal Justice, 30, 65-75.

LaFree, G. (19983). Losing Legitimacy: Street Crime and the Decline of Social
Institutions in America. Boulder: Westview Press.

LaFree, G. (1998b). Social institutions and crime bust of the 19905. The Journal of
Criminal Law in Criminology, 88, 1325-1368.

251

Langan, P., Greenfeld, L., Smith, S., Durose, M. & Levin, D. (February, 2001). Contacts
between Police and the Public: Findings from the 1999 National Survey, US.
DOJ: Washington, DC.

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

Levitt, S. (1998). Why do increased arrest rates appear to reduce crime: Deterrence,
incapacitation or measurement error? Economic Inquiry, 36, 353-372.

Levitt, S. (1999). The limited role of changing age structure in explaining aggregate
crime rates. Criminology, 37, 581-597.

Levitt, S. (2002). Using electoral cycles in police hiring to estimate the effect of police
on crime: reply. The American Economic Review, 92, 1244-1250.

Levitt, S. (2002). Using Electoral Cycles in Police Hiring to Estimate the Effect of Police
on Crime: Reply. American Economic Review, 92, 1244-1250.

Levitt, S. (2004). Understanding why crime fell in the 1990s: four factors that explain
the decline in six that do not. Journal of Economic Perspectives, 18, 163-190.

Levitt, S. & Kessler, D. (1999). Using sentence enhancements to distinguish between
deterrence and incapacitation. Journal of Law and Economics, 42. 343-363.

Loftin, C. & McDowall, D. (1982). The police, crime, and econometric theory: An
assessment. Sociological Review, 47, 393-401.

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

Maguire, E. & Katz, C. (2002). Community policing, loose coupling, and sense making
an American police agencies. Justice Quarterly, 19, 503-536.

Maguire, R. & Mastrofski, S. (2000). Patterns of community policing in the United
States. Police Quarterly, 3, 4-45.

Maltz, Michael D. (1999) Bridging Gaps in Police Crime Data. Washington, DC: US
Department of Justice, Ofﬁce of Justice Programs, Bureau of Justice Statistics.
Full report and Executive Summary are available from BJ S website:

http://www.ojp.usdoj.gov/bjs/abstract/bgpcd.htm

Markowitz, F., Bellair, P., Liska, A., & Liu, J. (2001). And wound in Extending Social

Disorganization Theory: Modeling the Relationships between Cohesion, Disorder,
and Feared. In Criminology, 39, 293-319.

252

Martin, S. & Sherman, L. (1986). Selective apprehension: A police strategy for repeat
offenders. Criminology, 24, 155-173.

Martinson, R. (Spring 1974). "What Works? - Questions and Answers About Prison
Reform," The Public Interest, pp. 22-54.

Marvel T. & Moody, C. (1996). Speciﬁcation problems, police levels, and crime rates
Criminology, 34, 609-646.

Marvel T. & Moody, C. (2001). The lethal effect of three strikes laws. The Journal of
Legal Studies, 30, 89-106.

Mastrofski, S. and Uchida, C. (1996), Transforming the police, in Hancock, B. and Sharp,
P. (Eds) Public Policy: Crime and Criminal Justice. Prentice-Hall, Upper
Saddle River, NJ, pp.196-219.

Mastrofski, 8., Parks, R., Wilson, D. (2003). Inﬂuences on the adoption of community
policing in the United States: a ﬁnal report to the National Institute of Justice.
National Institute of Justice: Washington DC.

Matthews, R. (1993). Kerb-crawling. Prostitution and multi-agency Policing. Police
Research Group. Crime Prevention Unit Series Paper 43. London, UK: Home
Ofﬁce.

Maxson, C., Hennigan, K. & Sloane, D. (2003). For the sake of the neighborhood?:
Civil Gang Injunctions as a gang intervention tool in southern California. In
Decker, S. (Ed). Policing Gangs and Youth Violence. Toronto: Wadsworth.

Mazerolle, L. & Roehl, J. (1998). Civil Remedies and Crime Prevention: An
introduction. Crime Prevention Studies, 9, 1-18.

McGarrell, E., Giacomazzi, A. & Thurman, Q. (1999). Reducing Disorder, Fear, and
Crime in Public Housing: A Case Study of Place-Speciﬁc Crime Prevention.
Justice Research and Policy, I, 34-69.

MeKenna, P. (2000). Foundations of Community Policing in Canada. Prentice Hall and
Allyn Bacon Canada: Scarborough, Ontario.

Mullen, K. (1996). The Computerization of Law Enforcement: A Diﬂusion of Innovation
Study. Dissertation. State University of New York at Albany.

Nagin, D. (1998). Criminal deterrence research at the outset of the 21st-century. Crime

and Justice: Crime and Justice: A Review of Research (vol. 23), edited by
Michael Tonry, Chicago, IL: University of Chicago Press: pp. 1-42.

253

National Association of Convenience Stores (1991). Convenience store security Report
and Recommendations: November 1991. Alexandria, VA: National Association
of Convenience Stores.

Newman, 0. (1972). Defensible Space. New York, New York: The MacMillan
Company.

NLECTC (2005). National Law Enforcement and Corrections Technology Center
Website. Available online: http://www.nlectc.org/about/justnet.htm1

Nunn, S. (1993). Computers in the cop car: impact of the mobile digital terminal
technology on motor vehicle theft clearance and recovery rates in a Texas city.
Evaluation Review, 17, 182-203.

Nunn, S. (2001). Police information technology: Assessing the effects of
computerization on Public Administration Review, 61, 221-234.

Nunn, S. & Quinet, K. (2002). Evaluating the effects of information technology on
problem-oriented-policing: If it doesn't ﬁt, must we quit? Evaluation Review, 26,
81-108.

Oliver, W. M. (1998). Community-Oriented Policing: A Systematic Approach. Upper
Saddle River: Prentice Hall.

Ouimet, M. (2002). Explaining the American and Canadian crime drop in the 19905.
Canadian Journal of Criminology, 44, 33-50.

Parks, R., Mastrofski, S., DeJong, C., & Gray, K. (1999). How ofﬁcers spend their time
with the community. Justice Quarterly, 16, 483-518.

Pate, A., Skogan, W., Wycoff, M. & Sherman, L. (1986). Reducing Fear of Crime in
Houston and Newark: a Summary Report. Washington, DC: Police Foundation.

Paternoster, R. & Bushway, S. (2001). Theoretical and Empirical Work on the
Relationship Between Unemployment and Crime. Journal of Quantitative
Criminology, 17, 391-407.

Pease, K. (1991). The Kirkholt Project: Preventing burglary on a British Public Housing
Estate. Security Journal, 2, 73-77.

Perkins, D., & Taylor, R. (1996). Ecological assessments of community disorder: Their
relationship to fear of crime and theoretical implications. American Journal of

Community Psychology, 24, 63-107.

Peterson, M., Chaiken, J ., Ebener, P., & Honig, P. (1982). Survey of Prison and Jail
Inmates: Background and Method. Santa Monica, CA: Rand Corporation.

254

Piliavin, 1., Thornton, C., Gartner, R. & Matsueda, R. (1986). Crime, deterrence, and
rational choice. American Sociological Review, 51, 101-119.

Police Foundation (1981). The Newark Foot Patrol Experiment. Washington, DC:
Police Foundation.

Ralston, R. (1999). Economy and race: interactive determinants of property crime in the
United States, 1958-1995. American Journal of Economics and Sociology, 58,
405-434.

Raphael, S. & Winter-Ebmer, R. (2001). Identifying the effect of unemployment on
crime. Journal of Law and Economics, )ﬂJ V, 259-283.

Reiter, M. S. (1999). Empowerment Policing. FBI Law Enforcement Bulletin, 68, 7-11.

Reisig, M. & Parks R. (2004). Can Community Policing Help The Truly Disadvantaged?
Crime & Delinquency, 50, 139-167.

Reiss, A. (1992). Police Organization in the Twentieth Century. In Michael Tonry and
Norval Morris, eds., Modern Policing. Crime and Justice, Vol. 15. Chicago:
University of Chicago Press.

Roach Anleu, S. (1998). The role of civil sanctions in social control: A socio—legal
examination. Crime Prevention Studies, 9, (pp. 21-43).

Rosenthal, A. & Fridell, L. (2002). 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].

Rosenfeld, R. (2000). Patterns in adult homicide. in Blumstein, A., and J. Wallman, eds.,
The Crime Drop in America. United Kingdom: Cambridge University Press.

Roth, J ., Ryan, J ., Gafﬁgan, S., Koper, S. et a1. (2000). National Evaluation of the COPS
Program—Title I of the 1994 Crime Act, Washington, DC: National Institute of
Justice, August 2000.

Ross, C. & J ang, S. (2000) Neighborhood disorder, fear and mistrust: The buffering role
of social ties with neighbors. American Journal of Community Psychology, 28,
401 -420.

Sampson, R (1986). Crime in Cities: The Effects of Formal and Informal Social Control.

In Reiss, A. & Tonry, M. (Eds). Crime and Justice. Chicago: University of
Chicago Press.

255

 

 

 

Sampson, R. & Cohen, J. (1988). Deterrent effects of the police on crime: a replication
and theoretical extension. Law and Society Review, 22, 163-189.

Sampson, R. & Laub, J. (1993). Crime in the Making: Pathways and Turning Points
Through Life. Cambridge, MA: Harvard University Press.

Sampson, R. & Raudenbush, S. (1999). Systematic social observation of public spaces:
A new look at disorder in urban neighborhoods. American Journal of Sociology.
105, 603-651.

Schuiteman, J. (1999). Cellular phones and pagers for police. Law and Order, 47, 51-53.
NLETC (2005). Online website: http://www.nlectc.org/about/justnet.html

Scott, M. (2004). The beneﬁts and consequences of police crackdowns. USDOJ:
Community Oriented Policing Ofﬁce: Washington, DC [on-line] available:
http://www.cops.usdoj.gov/mime/open.pdf'?ltem=964

 

Scott, Michael S. (2000). Problem-Oriented Policing: Reﬂections on the First 20 Years.
Washington DC: US. Department of Justice, Ofﬁce of Community Oriented
Policing Services.

Seaskate, Inc. (1998). A Technical Report prepared for The National Committee on
Criminal Justice Technology. National Institute of Justice: Washington, DC.

Shaw, C. & H. McKay. (1942). Juvenile Delinquency and Urban Areas. Chicago: Univ.
Press.

Sherman, L. (1995). Hotspots of Crime and Criminal Careers of Places. In Eek &
Weisburd, eds., Crime and Place. Crime Prevention Studies, 4.

Sherman, L., Gottfredson, D., MacKenzie, D., Eek, J. et a1. (1998). Preventing Crime:
What Works, What Doesn't, What's Promising: A Report To The United States
Congress. National Institute of Justice: Washington, DC.

Sherman, L. & Rogan, D. (1995). Effects of gun seizure on gun violence: 'Hot spots'
patrol in Kansas City. Justice Quarterly 12, 673-693.

Sherman, L. & Weisburd, D. (1995). General deterrent effects of police patrol in crime
hotspots: a randomized, controlled trial. Justice Quarterly, 12, 7 55-782.

Skogan, W. G. 1990. Disorder and Decline: Crime and the Spiral of Urban Decay in
American Neighborhoods. New York: Free Press.

Skogan, W. & Frydl, K. (Eds.) (2004). Fairness and Eﬂectiveness in Policing: the

Evidence. National Academies Press: Washington DC. (chapter 6. pp. 217-251).
The effectiveness of police activities in reducing crime, disorder, and fear.

256

Skogan, W. & Hartnett, S. (1997). Community Policing, Chicago Style. Oxford
University Press.

Smith, M. (2001). Police-led crackdowns and cleanups: An evaluation of a crime
control initiative in Richmond, Virginia. Crime & Delinquency, 47, 60-83.

Sousa, W. (2003). Crime reduction in New York City: The impact of the NYPD.
Dissertation. Rutgers State University of New Jersey - Newark.

Sparrow, M., Moore, M. & Kennedy, D. (1990). Beyond 911: A New Era for Policing.
New York: Basic Books, Inc. 1990.

Spelman, W. (1990). Repeat Offender Programs. Washington, DC: Police Executive
Research Forum.

Spelman, W. (1995). Criminal Careers of Public Places. In Eek, J. & Weisburd, D. (Eds).
Crime and Place. Monsey, NY: Criminal Justice Press.

Steffensmeier, D. & Harer, M. (1999). Crime trends, 1980 to 1996/1998: Age
composition affects and other explanations. Journal of Research in Crime and
Delinquency, 36, 235-274.

Stockton, D. (1999). Converting to Laptop computers. Law and Order, 47, 24-26.

Sommers, 1., Fagan, J. & Baskin, D. (1994). Delinquency and acculturation among
Latinos. Justice Quarterly, 11, 207-228.

Taylor, R., Bazemore, G., Boland, B., Clear, T. et a1. (1998). Crime and Place: Plenary
Papers of the 1997 Conference on Criminal Justice Research and Evaluation.
NIJ: Washington, DC.

Travis, J. & Waul, M. (August, 2002). Reﬂections on the Crime Decline. Proceedings
from the Urban Institute Crime Decline Forum. Urban Institute Justice Policy
Center: Washington DC.

Trojanowicz, R. (1982). An Evaluation of the Neighborhood Foot Patrol Program in
Flint, Michigan (East Lansing, Michigan: The National Neighborhood Foot Patrol
Center, Michigan State University.

Trojanowicz, R., Kappeler, V., Gaines, L., & Bucqueroux, B. (1998). Community
Policing: A Contemporary Perspective (2nd ed. ). Cincinnati: Anderson
Publishing Co.

Visher, C. (2000). The Rand inmate survey: A reanalysis. National Academy of
Sciences: Washington DC.

257

Walker, S. (1984). Broken windows' and fractured history: the use and misuse of history
in recent police patrol analysis. Justice-Quarterly, 1, 75-90.

Warr, M. (1998). Life-course transitions and desistance from crime. Criminology, 36,
183-216.

Way, L. (2002). Dirtbag Cars: Proactive Policing, Minorities, and the Poor.
Dissertation. Syracuse University.

Weisburd, D., Bushway, S. Lum, C. & Yang, S. (2004). Trajectories of crime at places: a
longitudinal study of street segments in. Criminology, 42, 283-321.

Weisburd, D. & Eek, J. (November 2004). What Can Police Do to Reduce Crime,
Disorder, and Fear? Annals of the American Association of Political and Social
Sciences, 593, 42-65.

Wiesburd, D. & Green, L. (1995). Policing drug hot spots: the Jersey City Drug Market
Analysis experiment. Justice-Quarterly, 12, 711-735.

Weisburd, D., Mastrofski, S., Greenspan & Willis, J. (April, 2004). The Growth of
Compstat in American Policing. Police Foundation Reports. Available Online:
www.policefoundation.com

Weisburd, D., Mastrofski, S., Mcnally, A. Greenspan & Willis, J. (2003). Reforming to
Preserve: Compstat and Strategic Problem Solving in American Policing.
Criminology and Public Policy, 3, 421-455.

Willis, J ., Mastrofski, S. & Weisburd, D. (2004). Compstat and Bureaucracy: a Case
Study of Challenges and Opportunities for Change. Justice Quarterly, 21 , 463-
496.

Wilson, J. (1990). Gainesville convenience store ordinance: Findings of fact.
conclusions and recommendations. Crime Control Research Corporation.

Wilson, J. & Boland, B. (1978). The effect of the police on crime. Law and Society
Review, 12, 367-390.

Wilson, J. Q. & Kelling, G. (1982). Broken windows: The police and neighborhood
safety, Atlantic Monthly, March, 29-38.

Wyoming Police Department (WY PD). (2000). Annual Report on Police Services. City
of Wyoming, MI. Available Online:
http://www.ci.wyoming.mi.us/pol_annual.htrn

Wyoming Police Tactical Team Website (February, 2006). Available online:
http://www.imageguild.com/tact/

258

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

Zhao, J ., Scheider, M. & Thurman, Q. (2003). A national evaluation of the effect of

COPS grants on police productivity (arrests) 1995—1999. Police Quarterly, 6,
387-409.

259

   

111111111111III