WHAT ABOUT THE B-LISTERS? AN OFFENDER ANALYSIS OF THE HIGH POINT
DRUG MARKET INTERVENTION
By
Alexis Norris

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Criminal Justice - Doctor of Philosophy
2014

ABSTRACT

WHAT ABOUT THE B-LISTERS? AN OFFENDER ANALYSIS OF THE HIGH POINT
DRUG MARKET INTERVENTION
By
Alexis Norris
The current study examines the effect of participation in the High Point Drug Market
Intervention (DMI) on re-offending post intervention. The DMI is a program that attempts to
eliminate overt open-air drug markets through a focused deterrence approach. This approach
involves targeting the individuals driving the crime problem in a targeted area, and enhancing the
threat of criminal sanctions while at the same time offering help from the community and social
service providers. By increasing the perceived likelihood and severity of the sanctions for those
most likely to be involved in committing crimes in the area, it is expected that the offenders’
perceptions of risk will be altered and they will be deterred from offending. Using survival
analysis to compare the recidivism patterns of participants in the program relative to those of an
equivalent comparison group, this dissertation examines the DMI’s deterrent effect on the
offenders targeted to participate in the program. While studies on the DMI have found it to have
a positive crime-control effect in the communities where it is implemented, participant outcomes
at the individual level have not been examined. This study addresses this gap in the literature by
providing a more complete understanding of the DMI’s impact on the offenders targeted to
participate in the program.

The findings suggest that offenders who participated in the DMI were arrested faster and
were at greater risk of arrest for prohibited offenses than those in the comparison group. While
these results are not consistent with what was hypothesized, they may indicate success for the
law enforcement follow-up component of the intervention. The findings also suggest that there
was no difference in time to re-arrest and risk of re-arrest between the DMI and comparison
group for drug offenses and any offense. Overall, the findings indicate that participation in the
DMI did not have a deterrent effect on the targeted offenders. Future research should focus on
components of the intervention that are thought to influence offender behavior such as social
services, family involvement, enforcement consequences, the call-in process, and deterrence
message to determine if they are having the expected affect on the targeted offenders.

Copyright by
ALEXIS NORRIS
2014

I dedicate this dissertation to my parents Janet and Derek, and my brother Sean.

v

ACKNOWLEDGEMENTS

I would like to thank Dr. Edmund McGarrell who served as my thesis and dissertation
committee chair. You have supported me throughout my time in the program and provided me
with a number of opportunities to learn and grow as a scholar; for this I am truly appreciative. I
would like to thank Dr. Natalie Hipple who joined my committee late but has been invaluable in
helping me through the dissertation process. You were always there to answer any question and
help with everything from putting me in contact with High Point to get my data, responding to
the IRB, looking at drafts of my chapters, to just offering words of encouragement whenever I
needed them. I hope you know how grateful I am for everything you have done for me. I would
also like to thank the rest of my committee Drs. Chris Melde and John Besley for helping to
guide me and provide me with feedback and support through this process.
I would especially like to thank Dr. April Zeoli and Dr. Louie Rivers who both served as
mentors throughout my time in the doctoral program. April, you taught me so much about how
to go about doing research, both quantitative and qualitative, and gave me a number of
opportunities to work on different projects that really helped to develop my skills. Having the
opportunity to work with you the first few years in the program is more than I could have asked
for. Louie, you helped me to further develop my qualitative skills and taught me how to relax
and enjoy myself. You were a great source of support and you always knew the best way to
reassure me that things will be okay; what’s the worst that could happen.
I would like to thank the High Point Police department which supported this research and
made the data available to me. I would specifically like to acknowledge Judy Brenner, the crime
vi

analyst at HPD, who provided me with whatever data I needed and took the time to answer any
questions I might have had. I would also like to acknowledge and thank Dhruv Sharma and
Steven Pierce from the Center for Statistical Training and Consulting (CSTAT) at MSU who
helped me a great deal when I was having trouble with some of the analyses for my dissertation.
To those who have impacted my life and made my time here more enjoyable than it
otherwise would have been, I will miss you. To my friends Julie Yingling, Kimberly Bender,
John Hakola, Tamara Dempsey, Jason Rydberg, Rebecca Stone, Mike Cassidy, Lev Fejes,
Derrick Franke, Carole Gibbs, Mike Suttmoeller, Sarah Fitzgerald, Yi Ting, Gio Circo, Byung
Lee, Roy Fenoff, and Jaemia Pratt, I enjoyed getting to know all of you and will miss the
tailgates, football games, trivia, and general hanging out. Although it took some of you a little
while to come around, I’m glad that we were able to make the most out of being here together.
To Mary Lee VanderMoere, Melissa Christle, Peggy Donahue, Denise Davenport, and Terri
Bulock, I truly appreciate all the help and support you have provided through the years. I
especially appreciate that you always let me know when there was food or treats upstairs, I loved
it. No matter what I needed you all always had an answer and I will miss being able to come
upstairs and talk to you all every week.
I would also like to express my gratitude to Carole Zimmerman who was my mentor
during my masters program and convinced me to apply to the doctoral program. Nick Corsaro,
my first officemate, who always provided me with good advice, made me laugh, provided a since
of support, and encouraged me to go on to the doctoral program. I always appreciate your
support and friendship. I also would like to thank Carmen Gear who has been a sounding board

vii

for me over the past 2 years. I appreciate all you have done for me; you have helped me make
my last couple of years in the program more enjoyable.
Finally, and most importantly, I owe my biggest thanks to my family. My parents Janet
and Derek have always been a constant source of love and support. Mom and Dad thank you for
supporting me in whatever I do and always being there for me. To Sean, my brother, thank you
for all your support and encouragement throughout this process. And to the rest of my family
and friends, the thing I always looked forward to the most was coming back home and spending
time with you. I always tell everyone how lucky I am to have the family I do, and I truly mean
it.

viii

TABLE OF CONTENTS

LIST OF TABLES………………………………………………………………………………..xi
LIST OF FIGURES ...................................................................................................................... xii
CHAPTER 1: INTRODUCTION ................................................................................................... 1
CHAPTER 2: FOCUSED DETERRENCE……………………………………………………….4
Origins of Focused Deterrence………………………………………………………………....4
Applications of Focused Deterrence……………………………………………………………5
Gangs and Groups…………………………………………………………………………....6
Guns…………………………………………………………………………………………..8
Open-Air Drug Markets……………………………………………………………………..10
Key Components Directed at the Targeted Offenders………………………………………...11
Targeted Population………………………………………………………………………....12
Multi-agency Working Group……………………………………………………………....12
Deterrence Message………………………………………………………………………....13
Call-in and Community and Family Involvement…………………………………………..14
Social Services………………………………………………………………………………15
Follow-up and Enforcement Consequences………………………………………………...16
CHAPTER 3: DETERRENCE THEORY AND PROCEDURAL JUSTICE…………………...18
Deterrence Theory………………………………………………………………………….…18
General Deterrence……………………………………………………………………….…20
Specific Deterrence…………………………………………………………………………21
Deterrence Theory and Focused Deterrence……………………………………………….…24
Targeted Population and Multi-agency Working Group …………………………………...24
Deterrence Message, Follow-up and Enforcement Consequences………………………....25
Procedural Justice Theory…………………………………………………………………….26
CHAPTER 4: METHODOLOGY……………………………………………………………….29
Research Question…………………………………………………………………………….30
Hypotheses……………………………………………………………………………………30
The Data……………………………………………………………………………………....30
Sample………………………………………………………………………………………...32
Comparison Group – Propensity Score Matching…………………………………………..33
Inclusion Variables……………………………………………………………………….35
Propensity Score Estimation……………………………………………………………...36
Matching………………………………………………………………………………….39
Assessing Balance………………………………………………………………………..40
ix

Research Variables……………………………………………………………………………43
Dependent Variables………………………………………………………………………..43
Independent Variables………………………………………………………………………43
Analysis Approach – Survival Analysis……………………………………………………....46
CHAPTER 5: RESULTS………………………………………………………...……………....51
Sample Characteristics………………………………………………………………………...51
Re-arrest Characteristics……………………………………………………………………....52
Survival Analysis - Time-to-Failure…………………………………………………………..56
Survival Curves……………………………………………………………………………..57
Cumulative Hazard and Cumulative Incidence Functions……………………………….....61
Cox Regression…………………………………………………………………………..…64
CHAPTER 6: DISCUSSION AND CONCLUSION……………………………………………67
Summary and Discussion……………………………………………………………………..68
Prohibited Offenses…………………………………………………………………………68
Drug Offenses………………………………………………………………………………70
Any Offense………………………………………………………………………………...70
Limitations…………………………………………………………………………………….71
Future Research and Theoretical Implications………………………………………………..73
Social services…………………………………………………………………………….73
Family involvement…………………………………………………………………...….73
Follow-up…………………………………………………………………………………74
Enforcing consequences…………………………………………………………………..75
Call-in and Deterrence message…………………………………………………………..76
Communication Theory and Call-in Meetings…………………………………………...78
Policy implications…………………………………………………………………………….81
APPENDIX……………………………………………………………………………..………..86
REFERENCES…………………………………………………………………………………..88

x

LIST OF TABLES

Table 1.

Balancing Tests for Propensity Scores: Means and Standardized Mean
Difference…..……….…………………………………………………………....… 42

Table 2.

Background Characteristics by Group ……………………….……………………. 52

Table 3.

Re-arrest Characteristics by Group ……………………………………………….. 53

Table 4.

Logistic Regression Models Estimating the Effect if DMI on Prohibited
Re-arrest, Drug Re-arrest, and Any Re-arrest …………………………………….. 55

Table 5.

Median Lifetime, Survival Probability, and Log-rank and Breslow Tests
of Re-arrest for Three Year Observation Period ……………………………....….. 57

Table 6. Cox Regression Models Estimating the Effect of DMI on Prohibited
Re-arrest, Drug Re-arrest, and Any Re-arrest Hazard Rates ……..………………. 66

xi

LIST OF FIGURES

Figure 1.

Histogram of Propensity Scores: DMI vs Comparison …………………………. 38

Figure 2.

Kaplan-Meier Survival Curves Stratified by Group, Prohibited Re-arrest ……... 58

Figure 3.

Kaplan-Meier Survival Curves Stratified by Group, Drug Re-arrest …………… 59

Figure 4.

Kaplan-Meier Survival Curves Stratified by Group, Any Re-arrest ……………. 60

Figure 5.

Cumulative Incidence Functions Stratified by Group, Prohibited Re-arrest …… 61

Figure 6.

Cumulative Incidence Functions Stratified by Group, Drug Re-arrest …………. 62

Figure 7.

Cumulative Hazard Functions Stratified by Group, Any Re-arrest …………….. 63

xii

CHAPTER 1: INTRODUCTION

Researchers and law enforcement officials have started to develop new innovative
intervention strategies to help combat a number of different crime problems. In accordance with
these strategies, police have begun to shift their enforcement efforts to specific crime problems,
situations, and places, focusing police strategies on hot spots and crime mapping (Weisburd,
2008). Particular criminal justice interventions have employed a problem-oriented policing
approach, focusing on identifying specific crime problems that are affecting a community, and in
collaboration with that community, developing strategies that target the underlying conditions
causing those crime problems (Braga et al., 1999; Klofas, Hipple, & McGarrell, 2010; Goldstein,
1979). These interventions are implemented with the expectation that they will decrease the
crime rate by deterring offenders and potential offenders through intense focus on a specific
crime problem in a concentrated area. One such strategy that has become popular in recent years
is focused deterrence.
The focused deterrence strategy involves identifying a specific crime problem in a
targeted area, enhancing the threat of criminal sanctions for the individuals who are driving the
problem, and deliberately and consistently communicating that threat to them. By increasing the
threat of criminal sanctions for those most likely to be involved in committing crimes in the
targeted area, it is expected that crime in that area will be reduced and the targeted offenders will
be deterred from committing the prohibited behaviors. This strategy has quickly grown in
popularity and has been applied to a number of crime problems including gang and group
violence (Braga et al., 2001; Braga, McDevitt, & Pierce, 2006; Kennedy & Braga, 1998), openair drug markets (Corsaro, Brunson, & McGarrell, 2010; Frabutt, Shelton, DiLuca, Harvey, &
1

Hefner, 2009; Hipple, Corsaro, & McGarrell, 2010), gun violence (Corsaro & McGarrell, 2009,
2010; McGarrell et al., 2006; Papachristos, Meares, & Fagen, 2007), and most recently domestic
violence (Sechrist, Weil, Shelton, & Payne, 2012). Scholars argue that these types of police
interventions provide an effective approach for gaining both specific and general deterrence
against crime (Braga & Weisburd, 2012). Focused deterrence strategies have been found to be
effective at reducing crime rates and improving the conditions of affected communities (Braga &
Weisburd, 2012), suggesting a positive general deterrent effect at the community level. This
approach also has potential to be effective at the individual level, deterring offenders specifically
targeted by the interventions; however research conducted on focused deterrence strategies
typically have not examined what happens to these individuals.
In 2003, in response to the high levels of overt open-air drug activity in their city, the
High Point Police Department in North Carolina developed the Drug Market Intervention (DMI).
The DMI uses a focused deterrence strategy in an attempt to eliminate open-air drug markets and
prevent and control crime and violence related to those markets (e.g., prostitution, homicide, gun
assault, sexual assault, and other serious violence). Studies have found the DMI in High Point to
be effective at reducing crime and violence associated with their drug markets (Frabutt et al.,
2009; Hipple, Corsaro, & McGarrell, 2010). High Point noted an average decline of roughly
seven percent for violent crimes, nine percent for property crimes, and six percent for drug and
nuisance offenses following the intervention (comparing 24 months pre-and post-intervention)
(Hipple, Corsaro, & McGarrell, 2010). While studies on the DMI have found it to have a
positive crime-control effect in the communities where it is implemented, participant outcomes
at the individual level have not been examined.

2

The purpose of this research is to analyze the deterrent effect of the High Point Drug
Market Intervention on the offenders targeted by the intervention, also known as B-listers. The
current study extends previous research by providing a more complete understanding of the
DMI’s impact on those targeted to participate in the program, rather than the larger communitylevel effects of the intervention. Although focused deterrence strategies like the DMI are meant
to be community-based interventions, developed and implemented to reduce violence and crime
problems in affected communities1 (Braga et al., 2008; Corsaro & McGarrell, 2010; Rivers,
Norris, & McGarrell, 2012), it is important to know and understand how the intervention affects
those specifically targeted. If no deterrent effect on the targeted offenders is found the strategy
could possibly be modified to produce the same crime-control effect in the community without
exerting all the time and intense focus on the offenders. However, if the DMI is found to be
effective at the community and individual level it enhances our understanding of the
effectiveness of this intervention, and the focused deterrence strategy.

1

In a study conducted by Rivers, Norris, and McGarrell (2012) examining a mental model of the High Point DMI
formed through interviews with police officials, service providers, and community members involved in the DMI,
they found that those interviewed realized that the success of the intervention was ultimately “contingent on
addressing community disorganization and reestablishing levels of informal social control and collective efficacy
necessary to prevent the drug market from reemerging in the targeted neighborhood” (pg. 71).

3

CHAPTER 2: FOCUSED DETERRENCE

Origins of Focused Deterrence
The focused deterrence strategy was originally developed in the mid-1990s as a tactic
aimed at reducing gang and group-involved violence in Boston (Braga et al., 2001; Braga et al.,
2008; McGarrell et al. 2006; Corsaro & McGarrell, 2010). Operation Ceasefire’s Boston Gun
Project began in 1995 as an attempt to use problem-solving policing in response to the city’s
youth homicide problem. A multi-agency working group was formed to research, design, and
implement a strategy that would be used to help reduce violent crime and youth homicides in the
city. First, their analysis indicated that gang members with prior involvement in the system
committed most of the homicides; about 1,300 gang members (who accounted for less than one
percent of their age group in the city) were responsible for 60 percent of all the youth homicide
in the city (Braga et al., 2001; Kennedy, 1998; McGarrell et al. 2006). Based on this finding, the
Projects’ central goal became deterring violence by these chronic violent offenders (gang
members).
The intervention that was designed consisted of two parts. Part one, the Ceasefire,
involved a law enforcement attack directed at the illicit market that supplied guns to youths. Part
two, the “pulling levers” strategy involved “deterring violent behavior by chronic gang offenders
by reaching out directly to gangs, setting clear standards for their behavior, and backing up that
message by ‘pulling every lever’ legally available when those standards were violated”
(Kennedy, 1998, pg. 3). This strategy was built on the realization that “chronic offending made
these youths and the gangs they formed, extremely vulnerable” (Kennedy, 1998, pg.5). There
were a number of ways (levers to pull) that law enforcement could sanction these offenders and
4

impose costs on the gangs. These levers included anything that could disrupt the gang’s street
drug activity, such as the police focusing on: low level street crimes (e.g., trespassing or public
drinking); serving outstanding warrants; cultivating confidential informants for their
investigations of gang activities; delivering strict probation and parole enforcement; seizing drug
proceeds and other assets; ensuring stiffer plea bargains and prosecutorial attention; and
focusing Federal attention on gang-related drug activity (Kennedy, 1998).
Offender notification or “call-in” meetings were held with high risk gang members,
communicating directly to them that if they continued to be involved in violence, all legally
available sanctions would be levied against them and their gang. These meetings were coupled
with crackdowns on gangs who continued to be involved in gun violence. The adult members of
these gangs were prosecuted federally, which allowed officials to show the gang members that
they were serious about reducing gun violence in the city (Kennedy, 1997; McGarrell et al.
2006). Law enforcement continually communicated to the gangs that violence would not be
tolerated and if it continued so would the crackdowns. If one gang member faltered the whole
gang would suffer the consequences. This strategy encouraged these groups to police themselves
to avoid becoming a priority to law enforcement. The evaluation of the Ceasefire intervention
suggests that it was associated with a significant reduction in youth homicide, victimization and
gun assault incidents in Boston (Braga et al., 2001). The success of Boston’s Operation Ceasefire
spurred the implementation of a number of similar focused deterrence initiatives.
Applications of Focused Deterrence
The central goal of the focused deterrence strategy is that it can be tailored to the specific
crime problem in a community. As a problem-oriented policing approach, the focus is on using
an iterative strategy to identify and address the specific crime problems affecting a community.
5

Emphasis is placed on identifying and analyzing the specific crime problems, responding to these
problems, and assessing and adjusting the response (Braga et al., 1999; Braga et al. 2008;
Goldstein, 1979). After the success of Boston’s Operation Ceasefire, focused deterrence
strategies have been applied to a number of different crime and public safety problems, such as
gun violence, firearms homicide, open-air drug markets, and most recently domestic violence.
Although there have been a number of different focused deterrence initiatives, the
strategies they use follow the same general framework. A multi-stage approach is utilized, which
typically consists of : (1) focusing on a specific crime problem; (2) creating a multi-agency
working group of relevant law enforcement agencies; (3) systematically identifying key
offenders, groups, and behavior patterns driving the crime problem; (4) developing a clear
deterrent message to offenders and groups of offenders that employ a wide range of sanctions
(pulling levers) to persuade them to cease their behavior; (5) focusing social services and
community resources on targeted offenders and groups to complement law-enforcement efforts;
and; (6) directly, clearly and repeatedly communicating to offenders why they are receiving this
special attention (Kennedy, 1997, Braga et al., 2008). This approach has been effective at
reducing gang violence (Braga et al., 2001; Braga, McDevitt, & Pierce, 2006; Engel et al., 2011;
Kennedy & Braga, 1998), gun violence (Corsaro & McGarrell, 2009, 2010; McGarrell et al.,
2006; Papachristos, Meares, & Fagen, 2007), and reducing crimes associated with drug markets
(Corsaro, Brunson, & McGarrell, 2010; Frabutt et al., 2009; Hipple, Corsaro, & McGarrell,
2010).
Gangs and Groups
As mentioned above, Boston developed the focused deterrence pulling levers strategy in
response to their youth homicide problem. They found that a majority of the youth homicide in
6

the city were committed by a small number of gang members. The strategy involved directly
informing gangs that violence would not be tolerated and every available lever would be pulled
if violence occurred. An analysis of the intervention strategy in Boston found a 63 percent
decline in monthly average homicides among individuals 24 years old and younger, a 32 percent
decrease in shots fired calls, and a 35 percent decrease in monthly gun assaults. When compared
to 39 other large US cities for the same time period (1991-1997), Boston experienced the largest
statistically significant decline in youth homicide (only three other cities showed statistically
significant declines) (Braga et al., 2001).
Lowell, Massachusetts and Minneapolis used a focused deterrence strategy to combat
gang problems. Lowell concentrated on general gang violence and it was reported that the city
experienced a 24 percent reduction in gun assaults and a 50 percent reduction in homicide
following the intervention (Braga, McDevitt, & Pierce, 2006). In an impact evaluation that used
seven other Massachusetts cities as a comparison, Braga et al. (2008) found that the pulling
levers strategy in Lowell was associated with a statistically significant reduction in gun-related
homicides and aggravated assaults, which was not experienced in the other cities. The
Minneapolis’ intervention targeted gangs but focused on reducing homicide and reported a
reduction in the pre- and post-intervention yearly totals of homicide victims (Kennedy & Braga,
1998).
The Cincinnati Initiative to Reduce Violence (CIRV) in Cincinnati, OH used a focus
deterrence strategy to reduce serious violence by at-risk group members. Specifically, the
objective was to reduce homicide and gun-related violence perpetrated by these groups. Engel et
al. (2011) used a pooled time-series regression analysis to estimate the impact of the CIRV on
group member involved homicides and non-group member involved homicides 24 and 42
7

months post-intervention. They found that group member involved homicides and violent
firearm incidents significantly declined after implementation of the intervention.

These

homicides declined by 38 percent 24 months post-intervention and by 41 percent 42 months
post-intervention. No such decline was found for the corresponding non-group involved
homicides.
Guns
In Indianapolis, the focused deterrence strategy was used to target illegal gun carrying
and gun use among known groups of chronic offenders. They focused on chronic offenders with
connections to gangs, guns, and drug markets. Using an interrupted time-series design,
McGarrell et al. (2006) reported a 34 percent statistically significant decline in monthly
homicides following implementation of the intervention (comparing 27 months pre-and postintervention). Corsaro and McGarrell (2009) tested whether the downward trend in homicides
found previously was similar across gang and non-gang homicides, if so it could suggest that
something other than the intervention was responsible for the decline. However, they found that
the decline in gang homicides was significantly greater than the decline in non-gang homicides
suggesting an intervention effect. Similarly, Corsaro and McGarrell, (2010) observed changes in
homicide patterns for the highest risk populations (individuals age 15 to 24) in Indianapolis after
a pulling levers intervention was implemented. High-risk populations experienced a statistically
significant reduction in homicides. The authors also found that in neighborhoods where specific
community level strategies were employed there was a statistically significant decline in highrisk homicide rates.
Chicago’s Project Safe Neighborhood (PSN) targeted convicted felons who carried or
used guns and coupled the deterrence message with increased federal prosecution and increased
8

sentence lengths. Using a quasi-experimental design, Papachristos, Meares, and Fagen (2007)
found that Chicago experienced a 37 percent decline in quarterly homicide rates in the treatment
area. They also found that decreases in gang related homicides were directly linked to the
percentage of offenders who attended the call-ins. Additionally, another evaluation of the
program found shooting declines ranging from 16 percent to 34 percent across the treatment sites
in Chicago (Skogan et al., 2009).
Chermak (2008) and Papachristos et al (2013) examined the effects of their respective
focused deterrence strategies on program participant recidivism. First, Chermak (2008)
evaluated probationers and parolees who participated in the Indianapolis Lever-Pulling
experiment. The study consisted of two treatment groups: the law enforcement group in which
participants attended a call-in meeting at a court house and the deterrence message was primarily
delivered by law enforcement officials, and the community group in which the participants
attended a call-in at a community center and the deterrence message was primarily delivered by
community members. He found no difference in the likelihood of recidivism or time to arrest
post-intervention between the treatment groups and the control group. About 33 percent of the
law enforcement group and the control group, and 28 percent of the community group were
arrested post-intervention. And, on average, the law enforcement group survived 140 days
before getting arrested, the community group survived 147 days, and the control group survived
149 days; there was no discernible difference.
Second, Papachristos et al (2013) examined the effect of PSNs offender notification
forums on the recidivism of program participants relative to a control group. They found that
offenders who participated in the offender notification forums experienced a significantly longer
time on the street before re-arrest than those in the control group. They also found that forum
9

participants had a 30 percent lower hazard of committing a new offense and had a lower
incidence of weapons crimes, murder, and violent crimes than those in the control group.
Open-Air Drug Markets
The application of the focused deterrence strategy to open-air drug markets was first
implemented in High Point, NC, in the form of the Drug Market Intervention (DMI). Since this
initial implementation the strategy has been implemented in a number of cities including
Rockford, Illinois, Nashville, Tennessee and Peoria, Illinois. The DMI adopted the pulling
levers model by focusing on specific drug markets, the small group of individuals working in
those markets (i.e. distributors, street-level sellers, and look-outs) that are driving the problem,
and directly communicating deterrence messages to them (Kennedy, 1998; Hipple, Corsaro, &
McGarrell, 2010).
Studies on the DMI have found it to be effective in reducing violent, property, and drug
crimes, and other behaviors associated with drug markets (Corsaro, Brunson, & McGarrell,
2010; Frabutt et al., 2009; Hipple, Corsaro, & McGarrell, 2010; Kennedy & Wong, 2009). In a
study examining the High Point DMI 24 months pre-and post-intervention, Hipple, Corsaro, and
McGarrell (2010) found an average decline of roughly seven percent for violent crimes, nine
percent for property crimes, and six percent for drug and nuisance offenses in the targeted areas
following the intervention. Examining specific neighborhoods in High Point where the
intervention was implemented, Kennedy and Wong (2009) found that 5 years post-intervention
the West End (the first neighborhood where the DMI was implemented) experienced a 57
percent reduction in violent crime and a 50 percent drop in gunshot calls for service. Four years
post-intervention, drug crime fell by 26 percent. In the South Side, 2 years post-intervention,

10

violent crime dropped at an average rate of 16 percent per year; however, drug crimes increased
by about 15 percent.
In Nashville, there were statistically significant reductions in property, narcotic, and drug
equipment offenses and a decrease in calls for service in the target area and adjoining areas
(Corsaro, Brunson, & McGarrell, 2010). Similarly, Rockford experienced a statistically
significant reduction in property crime, drug crimes, and nuisance offenses associated with the
pulling levers deterrence strategy in the targeted areas (Corsaro, Brunson, & McGarrell, 2009).
In Peoria, however, Corsaro and Brunson (2013) found that the pulling levers intervention had
no significant effect on the crime offense rate in the targeted neighborhood; although they
suggest that this may be due to the lack of community awareness and involvement in the
program.
Key Components Directed at the Targeted Offenders
As shown above, studies measuring the effectiveness of focused deterrence approaches
have primarily concentrated on community-level outcomes (e.g. reductions in gun violence,
firearm homicides, criminal activity related to drug markets, and youth violence), and have
generally shown positive results. A meta-analysis on the effects of focused deterrence strategies
also found that they had an overall statistically significant crime reduction effect (Braga &
Weisburd, 2012), suggesting that overall the focused deterrence strategy is having a positive
deterrent effect on the targeted crime in the targeted location. Although research in this area has
principally focused on community-level outcomes, many components of the focused deterrence
strategy are aimed at specifically deterring the individuals targeted by these strategies. The
following section discusses these components and how they are directed at influencing the
behavior of the targeted individuals.
11

Targeted Population
Focused deterrence strategies take advantage of the problem-oriented and place-based
policing literature which has shown that a relatively small number of offenders are responsible
for a disproportionally large amount of crime, and that these crimes are clustered in relatively
small geographic locations within a city (Braga et al.1999; Braga et al., 2010; Weisburd et al.,
2009). Given this, the problem-oriented policing literature suggests that police interventions
would be most effective when they are highly focused on the people, places and the context
driving the problem (Braga et al., 1999; Kennedy, 1998; Sherman & Weisburd, 1995). This
allows police attention to be focused on a specific crime problem (e.g. gun violence, gang
violence, drug markets) and the individuals that are driving that problem. For example,
Indianapolis and Rochester targeted gun violence (Indianapolis specifically focused on firearm
homicides) focusing on high risk probationers and parolees (Klofas et al., 2007; McGarrell et al.,
2006); while Chicago focused on gun violence but specifically targeted convicted felons
(Papachristos, Meares, & Fagen, 2007). High Point, Nashville, Rockford, and Peoria targeted
open-air drug markets focusing on low level street dealers (Corsaro &Brunson, 2013; Corsaro,
Brunson, & McGarrell, 2010; Frabutt, et al., 2009; Hipple, Corsaro, & McGarrell, 2010).
Multi-agency Working Group
The focused deterrence strategies also employs a problem-oriented policing approach by
encouraging law enforcement agencies to work in collaboration with the community to develop
proactive responses to crime problems. The introduction of a multi-agency working group,
which includes local and federal law enforcement and prosecutors, allow a wide range of
sanctions to be imposed on the targeted offenders. According to Kennedy (1997), because these
offenders commit crimes so often they leave themselves vulnerable to a number of legal
12

sanctions. Law enforcement agencies look to exploit the criminal records of these individuals by
coordinating to create consequences, by pulling every “lever” legally available, for individuals or
groups involved in crime (Braga et al., 2001; McGarrell et al., 2006; Tillyer et al., 2010). These
levers may include serving outstanding warrants or strictly enforcing their probation, basically
disrupting any criminal activity they may be involved in. This multi-agency group helps ensure
that a variety of sanctions can be used against these chronic offenders.
Deterrence Message
A defining feature of the focused deterrence strategy is the direct communication of the
strategy and threat of sanctions to the target population – those at risk for violent offending and
victimization. The small group of targeted offenders are directly confronted with a deterrence
message and told that their offending will not be tolerated. The strategy involves explicit
communication (typically in a group setting) to a specific group of chronic and habitual
offenders that have been identified by police and prosecutors (Hipple, Corsaro, & McGarrell,
2010). This component of the strategy involves key criminal justice officials (i.e., police,
prosecutors, probation and parole officers) warning the targeted offenders that continued
criminal behavior will result in the harshest possible legal sanctions (employed by the multiagency working group) (Kennedy, 1998). At the same time, they are offered help in the form of
social services for those who want to stop the behaviors that put them at risk for offending
(McGarrell et al., 2006; Tillyer et al., 2010). Essentially, law enforcement makes it known to the
offenders that they are ready to “take out” anyone that engages in violence and help is available
to those who want it.

13

Call-in and Community and Family Involvement
Interventions using the focused deterrence strategy have used a number of methods to
deliver the deterrence message to the targeted offenders: through probation or parole officers,
meetings with inmates in prison, and contact through gang outreach workers, members of the
clergy, or radio bulletins. For example, in Minneapolis they delivered the message to gang
involved victims in hospitals. However, the most common method of delivery is through call-in
or offender notification meetings. For these meetings, groups of targeted offenders are invited or
required to show up at a designated place (e.g., courtroom, church, community center, police
station, etc.) at a certain time. Meetings generally follow a common format: representatives from
law enforcement speak about the purpose of the program (e.g., violent behavior will not be
tolerated and will invoke an immediate and intense response), then representatives from the
community speak about the impact of violence or crime on the community at large, and finally
social service representatives detail program options for those who want to change their
behavior. Family and friends of the targeted offenders are also invited to the meetings to provide
support for the individuals.
The addition of community impact statements and the presence of the offender’s family
and friends are meant to serve as a form of deterrence. The central element of the call-in process
is the community, family, and loved ones of the offenders telling them that they need to stop
their behavior (Kennedy, 1998; Kennedy & Wong, 2009). In their meta-analysis of focused
deterrence strategies Braga and Weisburd (2012) suggested that family members and community
members in the interventions are powerful informal social controls for the offenders. In High
Point, they operated on the premise that “…dealers would stop selling drugs when people around
them made it clear they should…” leading them to identify “influentials” to invite to the call-in
14

meetings; “…the hope was to enlist those close to the offenders – parents, grandparents,
guardians, older members of the communities, ministers, ex-offenders – to create and reinforce
positive norms and expectations” (Kennedy & Wong, 2009, pg. 23).
Social Services
As noted above, in addition to the message that violence will no longer be tolerated by
law enforcement or the community, help is also offered to the targeted offenders in the form of
social services (e.g. job training, housing, mentoring, GED classes, housing assistance, parenting
assistance and substance abuse treatment) (Kennedy, 1997; Kennedy & Wong, 2009). As a
complement to the law enforcement efforts, social services and community resources are used as
a way to reintegrate the targeted offenders into the community (Kennedy, 1997; Rivers, Norris,
& McGarrell, 2012). The targeted offenders are given an opportunity to use social services to
help improve their situation, however if they continue to engage in violence, along with the legal
sanctions they will incur, the social service opportunities may also be taken away. According to
Williams and Hawkins (1986) this potential loss of legitimate opportunities can be a deterrent to
committing crime. Individuals may refrain from offending, not because of the legal sanctions
(e.g. jail), but because they do not want to risk losing a job or risk potentially not getting a job.
The use of these social services can be viewed as a legitimate opportunity (to cultivate vocational
skills or get an education) for these offenders, opportunities that they may have never had access
to before (Rivers, Norris, & McGarrell, 2012).
There is considerable variation in the nature and extent of services available across
intervention sites. Some programs offered direct services to offenders while others like WinstonSalem, High Point, Providence and Cincinnati had resource coordinators and case management
15

systems to assist offenders. Although the use of social services by the targeted offenders in these
programs have not typically been well documented, a study by Engel et al (2011) of the
Cincinnati Initiative to Reduce Violence (CIRV) provided a detailed description of their social
service model and examined its impact on the decline found in group member involved homicide
and gun violence incidents. Their analysis indicated that the social services did not have an
impact on the reduction in violence. However, they suggest this outcome may be due, in part, to
the social service program not being achieved as expected; less half of the clients that enrolled in
the social service program were those specifically targeted by law enforcement for CIRV (22
percent) and of the clients that were enrolled less than nine percent actively engaged in the
programs.
Follow-up and Enforcement Consequences
In the deterrence messages communicated to the offenders, a number of sanctions
(levers) are used to try to deter individuals from offending, such as parole and probation checks,
warrant enforcement, saturated patrol, increased prosecutorial attention (e.g., being prosecuted
federally), disruption of drug markets, and housing and property code enforcement (Kennedy,
1997). The success of a focused deterrence strategy can depend on whether or not the threats that
are made regarding law enforcement follow-ups are kept. Studies detail intended consequence
but they rarely have details of the actual consequences offenders faced if they were not deterred
by the message and continued to offend. Studies that have provided details on actual
consequences have described successful and unsuccessful efforts to enforce the deterrent
message. In High Point, offenders who attended call-ins were flagged in the police record
management system and any subsequent drug dealing resulted in immediate arrest. In Boston,
Minneapolis, and Chicago they performed crackdowns on gangs that continued to commit
16

violent crimes. These crackdowns gave credibility to the deterrence message (Kennedy, 1997).
Winston-Salem, however, had difficulty enforcing the promised consequences especially in the
case of minors. Judges were reluctant to impose harsh penalties for anything but the most serious
cases, so prosecution efforts were not successful.
Although studies have found focused deterrence strategies to be successful across crime
types and sites, the outcomes are primarily at the community level. As shown above, a number of
components in these strategies are aimed at deterring offenders specifically selected for increased
attention by program officials, yet little is known about the outcomes of these offenders. Given
that these components are geared toward influencing the behavior of the targeted offenders, an
examination of offender outcomes is needed to determine the effectiveness of focused deterrence
at the individual level.

17

CHAPTER 3: DETERRENCE THEORY AND PROCEDURAL JUSTICE

To explore the relationship between involvement in a focused deterrence strategy and
reoffending by the targeted offenders, deterrence theory and procedural justice will be explored.
Deterrence Theory
Deterrence is the practice of employing threats of punishment to influence an individual’s
behavior. The effectiveness of deterrence rests on the assumption that people have free will, act
in a rational manner and choose actions that result in the greatest pleasure and the least pain
(Brown & Esbensen, 1988; Paternoster, 2010). It is assumed that people will be deterred from
crime if the costs associated with punishment outweigh the pleasure associated with crime.
Rational individuals are capable of making decisions in a logical, calculating fashion, taking into
account the costs and benefits of alternative courses of action. This is based on Jeremy
Bentham’s idea of hedonic calculus. He believed that individuals acted as human calculators,
weighing the probability of present and future pleasures against present and future pain. They
then determine a course of action based on these calculations (Paternoster, 2010). Having free
will enables people to act as they choose and because they are hedonistic, their final decision is
based on the desire to maximize pleasure and minimize pain (Brown & Esbensen, 1988).
Most of the literature on deterrence focuses on the effect of changing the certainty,
swiftness, and severity of punishment associated with certain acts on the prevalence of crime
(Nagin, 1998; Paternoster, 1987). First, certainty refers to the probability of experiencing
punishment, such as the chance of being arrested, prosecuted, convicted, and sentenced. As the
perceived certainty of punishment increases, the probability of criminal behavior decreases.
Second, swiftness refers to the length of time that elapses between the crime and the punishment.
18

As the response to punishment becomes swifter, the probability of criminal behavior declines.
Finally, severity is the amount of punishment that is necessary to make it more painful for the
offender than the pleasure gained from the crime (Brown & Esbensen, 1988). Certainty and
swiftness are seen as more important than severity. Research has found certainty and swiftness
to have a modest deterrent effect on crime, while most studies suggest that changes in the
severity of punishment have little to no effect on crime (Matsueda et al, 2006; Nagin, 1998;
Nagin & Paternoster, 1993; Paternoster, 1987).
Scholars argue that crime is no different than other behavior and criminals are no
different than non-criminals, what differs are people’s assessment of the costs and benefits of an
action (Paternoster, 2010). Since everyone is motivated by self-interest, everyone has the
capacity to commit crime when the benefits exceed the costs. Becker argued that crime occurred
due to rational self-interest and could be understood like any other kind of economic activity,
“a person commits an offense if the expected utility to them exceeds the utility he could
get by using his time and other resources at other activities…Some persons become
“criminals,” therefore, not because their basic motivation differs from that of other
persons, but because their benefits and costs differ” (as quoted by Paternoster, 2010; pg.
776)2.
He argued that the decision to offend is made up of the costs and benefits of crime and noncrime, including the cost of formal legal punishments. If this is the case, the best way to prevent
crime is through punishments that are swift, certain, and appropriately severe. Deterrence
theorists focus primarily on the impact of official punishments of crime. Essentially, as an

2

Gottfredson and Hirschi (1990) also make this point. Gottfredson and Hirschi suggested, “[al]though there will be
little variability among people in their ability to see the pleasures of crime, there will be considerable variability in
their ability to calculate potential pains” (1990, pg. 95).

19

individual’s perception of the risk of being sanctioned for a crime (e.g., through fines, probation,
imprisonment) increases, the likelihood that they will commit the crime should decrease (Gibbs,
1968; Piguero & Pogarsky, 2002). The belief is that these sanctions will restrain those who have
been punished from committing more crimes (specific deterrence) and inhibit those who have
not yet offended from committing crimes (general deterrence).
General Deterrence
Deterrence theory makes a distinction between two types of deterrence: general and
specific. General deterrence refers to the idea that punishment deters offending among people in
the general population, those not specifically punished for a crime. Early deterrence research was
based upon aggregate-level data in which certainty and severity of punishment at the state level
was correlated with official crime rates. These studies mostly used secondary data and certainty
was generally measured as the aggregate risk of arrest or imprisonment for a particular crime,
while severity of punishment was measured in terms of average length of sentence (Paternoster,
1987). Studies focusing on aggregate analyses examined the effect of targeted and specific
policy interventions such as an increase in prison sentences (severity) or an increase in the
number of police (certainty), and estimated deterrent effects by looking at natural variations in
the crime rates and sanction levels across time and space (Nagin, 1998; Paternoster, 2010). The
criminal justice system is based, in part, on the expectation that having strictly enforced laws and
appropriate penalties that are applied with reasonable swiftness will prevent crime by deterring
offenders from reoffending and noncriminal individuals from committing crime (Paternoster,
2010).

20

Specific Deterrence
Specific deterrence refers to the idea that punishment reduces the risk of reoffending for
those specific people who are punished. In the 1980s, scholars began to understand deterrence
theory as a theory about perceptions of sanction threats and the relationship between those
perceptions and offender behavior. Perceptual deterrence assumes that motivation to commit
crime is affected by the perceived consequences of criminal behavior (Paternoster & Bachman,
2013). This restatement of deterrence as perceptual moved deterrence research from the macrolevel to the individual or micro-level of study (Nagin, 1998; Paternoster, 1987). In the 1980s
deterrence also began to expand beyond looking solely at the deterrent effect of formal sanctions
to include basic expected utility propositions, family and peer influences, and moral judgments
(Paternoster, 1987).
Deterrence theory typically focused on the State, and threats of fines and incarceration, as
agents of social control; however, significant others in the individual’s social environment play a
similar role. They pose threats of potential punishment which increase the subjective cost of
nonconformity (Grasmick & Bursik, 1990; Williams & Hawkins, 1986). Rather than focus
solely on the deterrent effect of formal legal sanctions, informal sanctions were introduced into
research. Williams and Hawkins (1986) argued that inhibition to crime could be brought about
directly and indirectly, through social censure, loss of legitimate opportunities (e.g. losing job),
and self-imposed costs (e.g. guilt) by the threat of legal sanctions. Studies have consistently
found that informal sanctions were more responsible for inhibiting criminal conduct than fear of
formal sanctions (Nagin, 1998; Paternoster, 1987; Pratt et al, 2006). Additionally, perceptual
deterrence studies have consistently found that informal sanctions were a more effective
deterrent than formal sanctions. Rather than fear of arrest, evidence points to the importance of
21

social censure from important others as a more significant factor in inhibiting crime than fear of
legal consequences (Paternoster & Bachman, 2013).
Deterrence research also began to examine how perceptions of sanction threats were
formed. Sanction threat perceptions are constantly modified and updated over time, and change
in response to criminal behavior committed by the actor, and the consequences of that behavior
(Stafford & Warr, 1993; Piquero & Pogarsky, 2002; Pogarsky et al., 2004). Facts and
perceptions – like the perceived certainty of punishment – that were once salient can fade from
an individual’s memory, and the new information they continuously acquire can prompt them to
revise their perceptions (Pagrasky et al., 2004). Studies examining how offending experiences
and their consequences affect sanction threat perceptions have found that individuals update their
sanction risk estimates in response to new information (Paternoster et al., 1995; Piquero &
Pogarsky, 2002; Pogarsky et al., 2004). For example, Paternoster et al. (1995) conducted a twowave panel study on the effects of involvement in petty theft and bad check writing on
perceptions of the certainty of being sanctioned. They found that as involvement in both
behaviors increased, the perceptions of sanction certainty decreased. They also found that
reduction in perceived certainty was significantly related to increased involvement on both
offenses and that being sanctioned between the two waves was related to an increase in
perceived certainty. Similarly, Pogarsky et al. (2004) found that being arrested increases the
perceived certainty of getting arrested in the future among high school students, particularly for
those with low prior estimates of certainty.
Recently scholars have argued that police interventions provide an effective approach for
gaining both specific and general deterrence against crime (Braga & Weisburd, 2012). Police
22

interventions such as problem-oriented policing and place-based policing target small areas
where crime is concentrated, individuals in those areas that are committing the crimes, and/or the
situations that perpetuate those crimes, with the expectation that they will decrease the crime rate
through deterring offenders and would be offenders.
In place-based policing, places are important for understanding and controlling crime and
emphasis is placed on reducing opportunities for crime at places, not simply reacting to crime
after it occurs; it shifts focus from the people involved in crime to the contexts of criminal
behavior (Weisburd, 2008; Weisburd et al., 2009). According to Weisburd (2008) place-based
policing is more efficient than focusing on only targeted individuals, provides more stable targets
for police than individual offending patterns, and has a strong evidence base. In this context,
places are considered small micro units of analysis like buildings, addresses, blocks, and street
segments or clusters of addresses; and crimes concentrated at these places are referred to as hot
spots (Braga et al., 2010; Weisburd, 2008; Weisburd et al., 2009). Studies have shown that crime
is generally clustered in micro places, which generate a disproportionate amount of criminal
events in a city, and the high concentration of crime at these micro places has been found to be
stable over long periods of time. For example, Braga et al. (2010) found that Boston gun
violence was concentrated at a small number of street segments rather than spread across the city
and that the micro places with volatile concentrations of gun assaults over time represented less
than three percent of street segments and intersections but generated more than half of all gun
violence incidents.
Similarly, Weisburd et al. (2009) assessed the extent to which juvenile crime is concentrated
at hot spots and assessed the stability and variability of crime at street segments. They found that
juvenile crime was strongly concentrated at hot spots and that a high rate of juvenile crime
23

remained relatively stable at street segments across a 14 year time period. A study conducted by
Groff et al. (2010) went further and showed that even street segments right next to each other
tend to have very different levels and patterns of crime over time. This shows the importance of
focusing on micro places because blocks and street segments in the same neighborhood could
have very different levels of crime, which could not be captured by just examining crime at the
neighborhood level. The National Research Council’s review of police practices and policies
studies found that when resources were focused on crime hot spots they showed strong evidence
of police effectiveness (Weisburd, 2008).
Deterrence Theory and Focused Deterrence
Focused deterrence utilizes core deterrence ideas, attempting to increase the certainty,
swiftness, and severity of punishment associated with committing crime, by increasing the risks
offenders face in new ways, such as, communicating incentives and sanctions directly to targeted
offenders (Braga & Weisburd, 2012). Focused deterrence strategies also look to capitalize on
the recent scholarship which argues that police interventions provide an effective approach for
gaining both specific and general deterrence against crime (Braga & Weisburd, 2012). This
approach incorporates focusing on micro places while also targeting the specific individuals in
those places that are driving the problem. The following section discusses how certain
components of the focused deterrence strategy are directed at core deterrence ideas.
Targeted Population and Multi-agency Working Group
The specialized attention on the targeted offenders should increase the certainty of getting
arrested and sanctioned if they engage in the targeted criminal activity. The introduction of a
multi-agency work group and the wide range of sanctions they are able to employ to persuade
offenders to cease their behavior also serves as a way to increase the certainty and severity of the
24

sanctions for committing the targeted behavior. Law enforcement agencies are able to exploit the
criminal records of these individuals by serving outstanding warrants or by strictly enforcing
their probation; disrupting any criminal activity they may be involved in. Federal involvement in
these multi-agency work groups also increases the possibility of severe sanctions for certain
behaviors (Kennedy, 1997). The harsh penalties imposed on similarly charged persons are also
used as proof regarding the seriousness of the message and the certainty of a criminal justice
response (Corsaro, Brunson, & McGarrell, 2010). With the increased perceived certainty offered
by targeting a specific population and the increased perceived severity gained by using a multiagency working group, theory suggests that the targeted offenders would be deterred from
reoffending.
Deterrence Message, Follow-up and Enforcement Consequences
Deterrence scholars argue that sanction threat perceptions are constantly modified and
updated over time. Individuals change their criminal behavior in response to the consequences
that are incurred (or not) by their behavior (Nagin, 1998; Piquero & Pogarsky, 2002; Pogarsky et
al., 2004), therefore, it is important that the threats laid out in the deterrence messages are carried
out by program officials. It is also important that the deterrence message is continually
communicated to the offenders. Theory suggests that facts and perceptions, like the thought of
being watched and/or sanctioned, that were once salient in an individual’s mind can fade from
memory over time, and that newly acquired information can prompt individuals to revise their
perceptions (Pagrasky et al., 2004). According to McGarrell and colleagues (2006) the direct
communication of the deterrence message to the targeted offenders in the offender notification
meetings and the knowledge of the follow-up consequences are the types of new information that
may cause offenders to reassess the risk of continuing to commit crime. Therefore, the
25

effectiveness of the deterrence message for the targeted population depends on how well the
cause and effect consequences are communicated to them. If threats are enforced and the
deterrence message is continually communicated to the targeted offenders they should be
deterred from committing the prohibited offense.
Overall, the basic premise of the focused deterrence approach involves enhancing the
threat of criminal sanctions for the highest risk offenders and directly communicating that threat
to them, in an attempt to deter the targeted offenders. Deterrence theory posits that the deterrent
effect for these offenders will be determined by their perceptions of risk and uncertainty. By
increasing the threat of sanctions for those most likely to be involved in committing crimes in the
targeted area and increasing the perceived likelihood, severity, and swiftness of the sanctions, it
is expected that the offenders’ perceptions of risk will be altered and they will be deterred from
offending (Kennedy, 1998; McGarrell et al., 2006).
Procedural Justice Theory
Deterrence theory assumes that individuals obey the law because they fear the
consequences of failing to do so. Other theories recognize that certain circumstances such as
criminal justice processes also influence individual behavior. Procedural justice focuses on the
way individuals make judgments about the fairness and outcome of a process through their
interactions with others during that process. In criminal justice, the focus is the subjective sense
of being treated fairly and with respect by law enforcement officials (Tyler, 2006). Scholars
argue that compliance with the law stems from an individual’s belief that the law is just and/or
the authority enforcing the law has the right to do so (Tyler, 1990). The key issue is not
necessarily whether sanctions are applied; rather, the individual’s view of the sanctions fairness,

26

and the legitimacy of the procedures through which law enforcement exercise their authority
(Tyler, 2003).
Studies of procedural justice in criminal justice began in 1990 with Tyler’s, Why people
obey the law. Since then research on procedural justice has grown, with studies exploring its
effect on public cooperation and compliance with the police. This research has consistently
found strong associations between procedural justice, perceptions of police legitimacy, selfreported cooperation with police, and satisfaction with the process (Tyler, 2006). In 2005,
Tyler’s study on ethnic group differences in trust in the police found that trust in the police was
strongly influenced by perceptions about the fairness of the procedures police follow when
exerting their authority. He also found that this trust, or lack of trust, was related to an
individual’s willingness to cooperate with the police (also see Hinds & Murphy, 2007; Murphy,
Hinds, & Fleming, 2008; Tyler, Schulhofer, & Huq, 2010). Similarly, positive and negative
interactions with the police also influence an individual’s perception of law enforcement.
Skogan (2006) examined people’s outcome experiences (positive and negative) with the police
on their general confidence in the police. He found that negative outcomes, such as arrest, that
were also perceived as unfair, led to negative perceptions of police legitimacy and less of a
willingness to cooperate with police. However, negative outcomes that were perceived as fair
had no impact on perceptions of police legitimacy and cooperation. This suggests that
procedural justice and perceptions of fairness are especially important when interacting with
individuals during negative experiences (e.g., arrest, sentencing).
If offenders targeted for participation in the focused deterrence strategies judge the
process and law enforcement authorities to be fair and just, they will be more likely to comply
with program rules than if they found the process unfair. According to Tyler (2003), if there is a
27

sense of fairness in the procedures, individuals may feel a sense of obligation to comply with the
law. Although program participants are told they will incur harsher sanctions if they continue to
offend, they are also given a second chance and offered services; elements that could lead to a
greater sense of procedural justice and subsequently a decreased likelihood of reoffending.

28

CHAPTER 4: METHODOLOGY

As shown in Chapter 2, focused deterrence strategies contain a number of components
geared toward influencing the behavior of offenders targeted for involvement in interventions
that use the approach. Despite these targeted attempts to influence behavior, only Chermak’s
(2008) study on the Indianapolis Lever-Pulling experiment and Papachristos et al. (2013) study
on PSN Chicago’s offender notification forums have examined the outcomes for these
individuals. The current study adds to the existing body of literature by providing another
examination of the individual-level effects of an intervention that uses a focused deterrence
approach. Specifically, this study examines the rate of re-arrest and time to re-arrest for lowlevel drug dealers (i.e., B-listers) involved in the Drug Market Intervention (DMI) three years
post-intervention. Generally, criminal justice policies rely on formal sanctions such as arrest and
imprisonment to deter offenders. However, punishments for low-level drug dealers are rarely
severe (e.g., probation or fines) or swift, leading to chronic offending for these individuals who
know there will be no real consequences for their behavior. The DMI tries to deter these chronic
offenders using a model built on the focused deterrence pulling levers concept – directly
communicating a deterrence message to the small group of lower-level drug dealers involved in
the drug market driving the crime problem in the area (Hipple, Corsaro, & McGarrell, 2010).
Although this process uses formal sanctions such as threat of arrest and receiving the maximum
possible sentence allowed for committing a prohibited offense, it also relies on informal
sanctions by involving community members, family, and friends in the call-in meetings where
the deterrence message is delivered, as well as offering the offenders social services.

29

Research Question
The main objective of this study is to determine the deterrent effect of involvement in the
DMI by examining the impact of the DMI on the outcomes of the targeted offenders; therefore
the study will focus on the following research question:
Do offenders who attend a DMI call-in meeting have a lower incidence and longer time
to re-arrest than offenders who experience traditional criminal justice processing?
Hypotheses
Based on the review of components of focused deterrence geared toward influencing
offender behaviors in Chapter 2 and the theoretical explanations presented in Chapter 3, the
following outcomes are expected:
1) Offenders who attend a DMI call-in meeting will have a lower incidence of re-offending
than those of an equivalent comparison group.
2) Offenders who attend a DMI call-in meeting will have a longer time to re-arrest (time-tofailure) than an equivalent comparison group.
The Data
Data for this study were obtained from the High Point Police Department in North
Carolina, which implemented the Drug Market Intervention. In 2003, in response to high levels
of open-air drug activity and poor police/community relations, the High Point Police
Department, in conjunction with a multi-agency working group, developed and implemented a
strategy to eliminate overt open-air drug markets throughout the city. This strategy consisted of
first identifying a narrowly defined target area through a data-driven process. This was the area
with the highest level of drug-related crime and violence. Next, they identified the street-level
drug dealers in that area. Law enforcement documented criminal cases against these drug
30

dealers through buys by undercover officers or confidential informants that were audio or video
taped. The cases built against the low-level non-violent dealers, (known as B-listers) were held
(i.e., banked or not filed with the prosecutor’s office) and these individuals were invited to attend
a call-in. Dealers that were violent (had violent records or gun charges) or those with a probation
or parole violation were immediately arrested and prosecuted. These were known as the Alisters and were used to set an example for the B-listers of what could happen if they continued
to deal drugs. The call-ins were held at High Point police headquarters. During the call-in, law
enforcement, community members, and service providers communicated to the B-listers that
violence and drug dealing would no longer be tolerated. If they continued to deal drugs or
committed certain prohibited behaviors3 they would be arrested immediately and prosecuted to
the fullest extent (which includes the original banked charges), however, if they stopped dealing,
nothing would happen to them.
Through a review process, law enforcement also identified what they called
“influentials” and invited them to the call-in. Influentials were individuals close to the offender
(e.g., parents, grandparents, peers) who program officials felt could be a positive influence and
hoped could help them stop offending. Services in the form of job training, housing,
transportation, substance abuse counseling, and educational programs were also offered to
individuals who attended the call-in meetings. High Point Communities Against Violence4

3

There were certain offenses that were prohibited for these offenders. If they committed the prohibited offenses
then they would incur the sanctions laid out by DMI officials. See Appendix A for a list of prohibited offenses.
4

High Point Communities Against Violence is a community group of service providers, health care workers,
ministers, and educators designed to provide DMI participants with basic services and resources. These resources
included assistance with employment, transportation, substance treatment, housing, and enrolling in educational
courses.

31

coordinates the social service resources for the DMI. After the call-in, street officers and
detectives patrolled the targeted area to make sure the targeted offenders (B-listers) were not
dealing. In the targeted areas “regular patrol officers developed a permanent strategy that
included maintaining systematic contact with notified offenders, their families, and the
community” to make sure that the message stuck (Kennedy & Wong, 2009, pg.19). High Point
implemented the DMI strategy in five target areas over a six-year period: West End in May of
2004, Daniel Brooks in April of 2005, Southside in June of 2006, East Central in August of
2007, and Washington Drive in February of 2010.
Criminal history data were collected from arrest records provided by the High Point
Police Department. These data were collected for all subjects three years pre-and postintervention. The data include demographics such as race, sex, and age at time of initiating
offense, along with number of arrests, type, and dates of arrests post-intervention. Criminal
history data were used to develop the control group for this study (see Sample below). They
were also used to measure the outcome variables: incidence of re-offending and time-to-failure
(i.e., time to re-arrest after the date of the call-in meeting). Corrections data from the North
Carolina Department of Public Safety was used to determine actual time at risk for re-arrest by
accounting for time incarcerated during the study period.
Sample
The sample for this study consists of two groups, a treatment group and a comparison
group. The treatment group included all individuals who attended a call-in meeting during each
round of the DMI (i.e., B-listers). DMI officials selected dealers for call-in meetings based on
certain criteria. First, dealers were identified through a review process. Then dealers had to be
caught selling drugs to an undercover officer or informant. The multi-agency working group
32

then reviewed the offenders’ records and offenders with violent records, gun charges, or
probation and parole violation were excluded and became A-listers. The remaining offenders
became B-listers and were invited to attend a call-in meeting (n=81). The comparison group
consists of a matched group of offenders arrested for selling drugs or possession with intent to
sell drugs within one month before or after a call-in (n=162). The procedure for developing the
comparison group is described below.
Comparison Group – Propensity Score Matching
The comparison group was constructed using a propensity score-matching method. One
of the benefits of estimating causal effects using randomized experiments is “that the treated and
comparison groups are guaranteed to be only randomly different from one another on all
background covariates, both observed and unobserved” (Stuart, 2010, pg 1). When comparing a
treatment group with a nonexperimental comparison group to estimate the causal effect of a
treatment, self-selection problems or systematic bias in the selection of units assigned to the
treatment group could influence the effects (Cook & Campbell, 1979). Therefore, when
estimating causal effects using observational data it is best to emulate a randomized experiment
as closely as possible by obtaining treatment and control groups with similar covariate
distributions. Choosing well-matched treatment and control groups will reduce the bias that may
occur due to the covariates (Stuart, 2010). This is done by using a procedure known as
matching, a technique that attempts to identify, for each individual in the treatment group, at
least one individual in the comparison group that “looks like” the treated individual based on
some set of observational characteristics (Apel & Sweeten, 2010).

33

The matching method used in the current study is propensity score matching. This
method offers a way to select a subsample of treated and untreated individuals that are similar
(Apel & Sweeten, 2010). A propensity score is the probability of being assigned to the treatment
group given a set of observed pretreatment covariates. Once propensity scores are found for
individuals in the treatment group and individuals in the comparison group, subjects from those
groups are matched based on the value of their propensity scores. Propensity scores are
essentially balancing scores, “at each value of the propensity score, the distribution of the
covariates defining the propensity score is the same in the treated and comparison groups. Thus,
grouping individuals with similar propensity scores replicates a mini-randomized experiment, at
least with respect to the observed covariates” (Stuart, 2010, pg. 6). Propensity scores also allow
construction of matched sets with similar distributions of covariates without requiring close or
exact matches on all individual variables (Rosenbaum & Rubin, 1983). Propensity score
matching is a widely accepted method for adjusting for existing group differences to create an
equivalent comparison group, therefore this statistical technique is suitable for determining the
best comparison matches for the DMI participants in the treatment group (Freedman & Berk,
2008).
Given that the treatment group for this study consisted of individuals caught selling drugs
to undercover agents, the comparison group was drawn from pools of offenders that were
arrested for a drug offense in High Point within a month before or after each call-in5 (n= 1370).
That offense will be the referred to as the “initiating offense” for the comparison group. The

5

The range of dates for each pool were: West End April 18, 2004 - June 18, 2004; Daniel Brooks March 5, 2005 –
May 5, 2005; Southside May 8, 2006 – July 8, 2006; East Central July 14, 2007 – September 14, 2007; and
Washington Drive January 9, 2010 – March 9, 2010.

34

pool of offenders was refined by excluding individuals arrested for mere possession of drugs or
drug paraphernalia for their initiating offense since this would not rise to the level of seriousness
needed to come to the attention of DMI officials. Individuals charged with violent offenses or
possession of a firearm at the same time as the initiating offense or that had prior weapons
charges were also excluded from the pool because they would not have met the criteria for
inclusion as a B-lister6. After excluding these individuals, 211 offenders were available for
inclusion in the matching procedure.
There are several key steps when conducting propensity score matching: 1) determining
variables to include for matching, 2) estimating the propensity score, 3) implementing a
matching method, and 4) assessing the quality of the resulting matched samples by
demonstrating balance.
Inclusion Variables. To begin, a propensity score model is developed which includes variables
that potentially confound the relationship between treatment status and the outcome. A
confounding variable is a variable that is associated with both treatment selection and the
outcome, when unaccounted for, these variables can affect the ability to make causal inferences
about the effect of the treatment on the outcome (i.e., the effect of DMI on re-arrest) (Apel &
Sweeten, 2010; Lanza, Moore, & Butera, 2013). It is important to include all variables known
to be related to treatment assignment and related to the outcome in the propensity score model
(Rubin & Thomas, 1996; Stuart, 2010). It is also acceptable to include variables that are not
related to treatment because they will have little influence on the propensity score model.

6

Could not account for probation and parole violations at the time of initiating offense.

35

Therefore according to Stuart (2010), when possible, researchers should be liberal in including
variables that may be associated with treatment assignment or the outcome. The goal is to model
non-random elements of being selected for treatment using observed information on potential
confounders (Apel & Sweeten, 2010). However, it may not always be possible to include a large
number of variables, in those cases; Stuart (2010) says priority should be given to variables
thought to be related to the outcome. The variables included in this model were: age at time of
initiating offense, sex, race, age at first offense, type of first offense (violent, property, drug,
other), previous number of arrests, and the most common type of previous arrest (violent,
property, drug, other).7
Propensity Score Estimation. Once the covariates were identified the propensity scores were
estimated. The propensity score is defined as the probability of receiving treatment conditional
on a set of observed pretreatment covariates:
𝑃 (𝑥) = Pr(𝑇 = 1 | 𝚾 = 𝐱)
where for each individual whose pretreatment covariates, X, have observed values X=x,
Pr is probability and T is a binary indicator that takes values 1 for treatment and 0 for
comparison. Any model relating a binary variable (i.e., the treatment assignment) to a set of
predictors can be used to estimate propensity scores, which is the predicted probability of
receiving treatment (Stuart, 2010). In the current study, the propensity scores (predicted
probability of being selected for DMI) were obtained using logistic regression. After propensity
scores are obtained it is important to assess common support between the treatment and

7

Models with different variations of variables were also tested; this model yielded the best results.

36

comparison individuals to determine if matching is possible. Common support occurs when the
distribution of propensity scores for the treatment and comparison individuals overlap (Apel &
Sweeten, 2010; Lanza, Moore, & Butera, 2013). Figure 1 shows the histograms of the
propensity score distributions for the treatment and comparison group. The histograms indicate
considerable overlap of propensity scores, meaning for nearly all DMI participants (treatment
group) there is an individual in the comparison group with a similar propensity score to match
on.

37

38

Matching. After estimation of the propensity scores, matching is performed to construct
equivalent groups. When there are a large number of comparison individuals compared to the
number of treatment individuals it is possible to get multiple good matches for each individual in
the treatment group, therefore a 2:1 nearest neighbor matching (ratio matching) method is used.8
Each treated individual is matched with two comparison individuals who have similar estimated
propensity scores (Stuart, 2010). Individuals in the treatment group are randomly sorted and
then the first individual from the treatment group is matched with the two individuals from the
comparison group who have the closest propensity score estimates. Once the matches are made
they are set aside. This process continues until each individual in the treatment group is matched
to two comparison individuals. Any comparison individuals not matched are discarded.9 This
method involves a trade-off between variance and bias. Bias increases because on average,
poorer matches may occur since the second closet match for each treated individual is further
away (i.e., the propensity scores are not as close) than their first closest match. However, using
multiple matches reduces variance because there is a larger matched sample size (Stuart, 2010).
A caliper of .2 standard deviations of the logit of the propensity score was imposed to improve
matches made on the estimated propensity scores. A caliper is the maximum distance a treated
and comparison individual can be apart on their propensity score in order to be matched. Having
a small caliper will result in better matches on the propensity score because they are more

8

According to Caliendo and Kopeinig (2005) when using oversampling one must decide how many matched
partners should be chosen for each treated individual. Since the number of comparison offenders available to be
included in the matching procedure was 211 a 2:1 method was used. This allowed every treatment individual to
have the same number of comparison matches.
9

Matching was done using the program “psmatching3” in SPSS via an SPSS R-Plugin.

39

similar, therefore likely resulting in better balance on the covariates. The procedure produced
162 matched individuals for the comparison group.
Assessing Balance. After matching, tests were performed to assess the quality of the resulting
matched groups. The goal of matching is to get the best possible balance between the treatment
and comparison groups on the pretreatment covariates (i.e., confounders) (Apel & Sweeten,
2010; Lanza, Moore, & Butera, 2013; Stuart, 2010). A demonstration of balance makes a
stronger case for assuming that treatment is “as good as randomly assigned,” between the groups,
at least with respect to the variables included in the propensity score (Apel & Sweeten, 2010).
Balance was assessed by comparing different statistics for the treatment and comparison group
pre- and post- matching. First, a t-test (continuous variables) and Chi square test (categorical
variables) were used to determine whether there were statistically significant differences between
the treatment and comparison group on the pretreatment covariates. Before matching,
differences are expected, but after matching, imbalance should be reduced. Second, one of the
most common techniques for assessing balance, the standardized mean differences (standardized
bias), is examined pre- and post- matching. The standardized mean differences are similar to an
effect size (e.g., Cohen’s d), therefore the treatment and comparison groups are considered
balanced on the covariates if their standardized mean differences are less than the absolute value
of 0.25 after matching (Apel & Sweeten, 2010; Lanza, Moore, & Butera, 2013). Table 1
presents the differences and standardized mean differences between the treatment (DMI
participants) and comparison group on each covariate included in the propensity score model as
well as pretreatment covariates not included in the propensity score model pre-and post
matching. Even before matching, the treatment and comparison groups were not statistically
different on the observed covariates and the standardized mean differences showed that most
40

covariates were balanced, with the exception of race, number of arrests, number of drug arrests,
age at first arrest, and age at first drug arrest. After matching the groups were balanced on all
covariates.

41

Table 1. Balancing Tests for Propensity Scores: Means and Standardized Mean Difference
Before Matching

Variable
Age
Sex
Male
Female
Race
Black
White
First arrest type
Violent
Drug
Property
Other
No arrests
# Arrests
# violent arrests
# drug arrests
# property arrests
Most common arrest
Violent
Drug
Property
Other
No arrest
Age 1st arrest
Age 1st drug arrest

DMI
Control
(n=81)
(n=211)
Mean (SD) Mean (SD)
/n(%)
/n(%)
27.6(9.3)

27.0(8.8)

72(88.9)
9(11.1)

188(89.1)
23(10.9)

After Matching

P
value
.629
.959

Control
(n=162)
Mean (SD)
/n(%)
27.0 (8.8)

188(89.1)
23(10.9)

11(13.6)
49(60.5)
14(17.3)
4(4.9)
3(3.7)
20.7(8.9)
17.9(11.4)

Before

After

.659
.883

.06
-.01

.06
-.02

.209

-.53

-.23

.724

-.03

-.10

.317
.699
.193
.858
.240

.27
.15
.31

.14
.06
.18

.17

.13

.355
.151

.34
.37

.12
.21

152(93.8)
10(6.2)
.100

3(3.7)
19(23.5)
13(16.0)
43(53.1)
3(3.7)
11.0(8.4)
0.6(0.8)
2.6(2.6)
1.1(1.7)

P
value

145(89.5)
17(10.5)
.069

79(97.5)
2(2.5)

St. Mean
Difference

9(4.3)
32(15.2)
32(15.2)
109(51.7)
29(13.7)
8.8(8.6)
0.4(0.9)
1.8(2.5)
1.2(2.0)

.046
.282
.017
.890
.005

16(7.6)
91(43.1)
55(26.1)
20(9.1)
29(14.1)
17.6(10.3)
13.7(13.0)

.018
.011

9(5.9)
27(16.7)
25(15.4)
93(57.4)
8(4.9)
9.9(8.6)
0.5(0.9)
2.2(2.7)
1.2(1.8)
15(9.3)
81(50.0)
43(26.5)
15(9.3)
8(4.9)
19.6(8.7)
15.5(12.8)

Note: T-test (continuous variables) and Chi square (categorical variables) were used to test significance.
Means and standard deviations are displayed for continuous variables and count and frequency for
categorical variables.

42

Research Variables

Dependent Variables
The primary outcomes of interest are incidence of re-arrest and time-to-failure (re-arrest)
for any prohibited offense three years after the intervention. Included in the prohibited offenses
are UCR Part I crimes, weapons offenses, assault on law enforcement officers, and possession
with the intent to sell drugs or higher (see Appendix A for a full list of prohibited offenses).
Incidence of re-arrest and time-to-failure will also be examined for any offense and any drug
offense. Incidence of re-arrest is operationalized as a dichotomous variable used to indicate
whether or not offenders were re-arrested (1= re-arrested, 0 = not re-arrested). In instances
where arrest records showed multiple charges for an arrest, the most serious charge was used.
Time-to-failure is time in days from the call-in date for the treatment group and from the date of
the initiating arrest for the comparison group, to the date of re-arrest (e.g., date of call-in
subtracted from date of re-arrest).
Independent Variables
The independent variables used for this study include, participation in DMI,
demographics, and time varying crime rates.
Participation in DMI is the primary predictor variable. This is a dichotomous variable
used to indicate group membership. The treatment group (DMI) is coded as 1, and the
comparison group (traditional criminal justice process) is coded as 0.
Race is a dichotomous variable with 0 indicating Black, and 1 indicating white.10

10

Only Black and White offenders were present in the sample.

43

Sex is a dichotomous variable with 0 indicating female, and 1 indicating male.
Age is conceptualized as the age in years at time of the call-in date for the treatment
group and age at time of initiating offense for the comparison group.
Number of Previous Arrests is a continuous variable indicating number of arrests prior to
the call-in date of the treatment group and initiating offense for the comparison group.
Most Common offense is a categorical variable conceptualized as the most common type
of arrest prior to the call-in date of the treatment group and initiating offense for the
comparison group. Violent is coded as 1, property is coded as 2, drug is coded as 3, and
other is coded as 4. Violent crimes included UCR part 1 crimes (e.g., murder, rape,
robbery, aggravated assault) and violent prohibited offenses outlined by High Point PD
(see Appendix A). Property crimes included UCR part 1 crimes (i.e., burglary, larceny,
motor vehicle theft). Drug crimes included all crimes coded as involving drugs by High
Point PD. Other included all crimes not in the previous categories.
Violent crime rate (2004-2013) is the rate of violent crimes per 1,000 residents. This
will be treated as a time-varying covariate and will be calculated monthly for each DMI
neighborhood.
Property crime rate (2004-2013) is the rate of property crimes per 1,000 residents. This
will be treated as a time-varying covariate and will be calculated monthly for each DMI
neighborhood.

44

Drug crime rate (2004-2013) is the rate of drug crimes per 1,000 residents. This will be
treated as a time-varying covariate and will be calculated monthly for each DMI
neighborhood.

45

Analysis Approach – Survival Analysis
To examine whether the DMI has a deterrent effect on the targeted offenders who
attended the call-ins, this study utilizes survival analysis employing a continuous-time event
model. Survival analysis allows for the examination of length of time to an event occurrence,
with event occurrence representing an individual’s transition from one state to another (Singer &
Willett, 2003). There are three major features needed to conduct a survival analysis: 1) a well
defined event; 2) a clear “beginning of time;” and 3) a meaningful metric for measuring time.
First, the primary event for the current study is re-arrest for any prohibited offense. The
prohibited offenses are those that DMI officials specifically warned program participants that, if
they engaged in them, would lead to the harsh sanctions. Arrest for drug offenses and any
offense will also be examined. Second, the beginning of time refers to the point in the study
when all individuals in the population are at risk of experiencing the event (i.e., re-arrest) (Singer
& Willett, 2003). For this study, the beginning of time for the treatment group is the date of the
call-in that the individual attended; this is when their risk of being re-arrested started.11 For the
comparison group, the beginning of time starts on the date of their initiating offense.12 Finally,
since the exact day of re-arrest can be captured with the official arrest records, time to the event
will be measured in days.
A major advantage of survival analysis is the ability to account for right-censored data.
Right censoring occurs when an individual’s time to event (re-arrest) is unknown. This may

11

To get actual time at risk, corrections records were used to account for times during the study period that
subjects were incarcerated and therefore not at risk of being re-arrested.
12

The drug offense that made them eligible for inclusion in the control group (see section on control group for
more details).

46

occur for two reasons: 1) some individuals may never experience the event and 2) others may
experience the event but not during the study period (Singer & Willett, 2003). In this case, the
time period of the study (three year or 1095 days) may end before an individual gets re-arrested
for the specified offense; this does not mean that they were never re-arrested for that offense, just
that they did not get re-arrested for it during the study period. In this study individuals were also
censored because they became incarcerated during the study period and were no longer at risk
for re-arrest. Survival analysis allows you to analyze both observed and censored cases.
Analysis Procedure
Time to event can be measured in continuous or discrete time. Continuous time is
recorded in precise units (e.g., days, minutes, seconds) where as discrete time is recorded in
thicker intervals (e.g., months, years, semesters) (Singer & Willett, 2003). According to Singer
and Willett (2003) time should be recorded in the smallest possible unit relevant to the study.
Since time for this study is measured in days, a continuous-time event model is used to analyze
the data.
The first step in a survival analysis is to describe the distribution of event occurrences,
whether and when events occur. To describe the distribution of continuous-time event
occurrences the Kaplan-Meier estimate of the survival function and the cumulative hazard
function should be examined (Singer & Willett, 2003). The Kaplan-Meier method for estimating
the survivor function involves constructing time intervals so that each interval only contains one
observed event time. That is, each interval will begin at one observed event time (re-arrest) and
end just before the next (Singer & Willett, 2003). The estimate of the survivor function is
obtained by first computing the conditional probability of event occurrence:

47

𝑝̂ (𝑡𝑗 ) =

𝑛 𝑒𝑛𝑣𝑒𝑛𝑡𝑗
𝑛 𝑎𝑡 𝑟𝑖𝑠𝑘𝑗

where 𝑝̂ (𝑡𝑗 ) is the conditional probability that an individual who is at risk at the
beginning of interval j will experience the event by the end of the interval, 𝑛 𝑒𝑣𝑒𝑛𝑡𝑗 is the
number of individuals who experience the event during interval j, and 𝑛 𝑎𝑡 𝑟𝑖𝑠𝑘𝑗 is the number
of individuals at risk during the beginning of interval j. Once the conditional probabilities are
found for each interval, the estimate of the Kaplan-Meier survivor function is constructed by
multiplying the successive conditional probabilities of surviving through each interval. The
survivor function is the probability that an individual will not experience the event during a
given interval (i.e., the probability of surviving (not being re-arrested) past a given interval).
Kaplan-Meier survival curves are plotted for the treatment and comparison group out of the
estimates of the survivor curve. To determine whether the survival curves for the groups are
different the log-rank and Breslow tests were used. The log-rank and Breslow tests are large
sample chi-square tests which uses observed versus expected cell counts for each of the intervals
to compare the survival curves (Kleinbaum & Klein, 2012). When comparing the curves, the
log-rank test emphasizes events that occur later in time while the Breslow test emphasizes events
that occur early in time.
Normally, the hazard function would be examined to determine the risk of an event
occurring at each interval among those still at risk (i.e., the conditional probability that individual
i will experience the event in interval j, given that they did not experience it in an earlier
interval). However, the hazard function cannot be well estimated in a continuous-time event
model, therefore the cumulative hazard function is examined. The cumulative hazard function,
denoted by H(𝑡𝑖𝑗 ) assesses, “at each point in time, the total amount of accumulated risk that
48

individual i has faced from the beginning of time until the present” (Singer & Willett, 2003, pg.
488). By examining the changing hazard levels over time, the unique risk of experiencing an
event at each particular point in time can be deduced. The cumulative hazard for any offense
will be explored graphically to learn about the underlying shape of the hazard function.
For prohibited offenses and drug offenses a competing risk model was used. Competing
risks arise when subjects are exposed to more than one cause of failure, and failure from one
cause prevents the occurrence of the others (Lin, So, Johnson, 2012). For example, if an
offender is arrested for a non prohibited offense it may prevent them from getting arrested for a
prohibited offense during the study period, or affect their time at risk, leading to an
underestimation of the hazard rate. Because only one event can occur first, these multiple events
are said to compete (Singer & Willet, 2003). To resolve this issue the event-specific hazard
function and cumulative incidence function are used. In the context of competing risks an
individual can leave the risk set by experiencing the targeted event, through censoring, or by
experiencing a competing event. The event-specific hazard function assesses the risk of
experiencing a specific event, given that an individual is at risk of doing so; specifically, it is the
instantaneous rate of failure due to cause k (competing risks) conditional on survival until time t.
From the event-specific hazard function, the cumulative incidence function is analyzed which
allows for the use of a single event time variable to analyze the cumulative probability of failure
from a specific cause k over time. For the prohibited offenses analysis the primary event of
interest will be arrest for prohibited offenses with the competing risk being any other offense.
To examine whether variation in the risk of re-arrest varies systematically with
participation in the DMI a Cox regression model is used. For continuous-time event data the
Cox regression model (also known as the proportional hazards model) models the cumulative
49

hazard as a linear function of the covariate (participation in DMI), then transforms it using the
log cumulative hazard function (Singer & Willett, 2003). The Cox regression model can be
written as:
log H(𝑡𝑖𝑗 ) = logH0 (𝑡𝑗 ) + 𝛽1 𝐷𝑀𝐼𝑖
where logH0 (𝑡𝑗 ) is the general baseline log cumulative hazard function at each time
interval (this represents the value of the outcome when DMI is set at 0), and 𝛽1 (the parameter
associated with DMI) measures the size of the vertical displacement (or the constant shift per
unit difference in the value). When the model is extended to include multiple predictors it can
be written as:
log H(𝑡𝑖𝑗 ) = logH0 (𝑡𝑗 ) + [𝛽1 𝑋1𝑖𝑗 + 𝛽2 𝑋2𝑖𝑗 + ⋯ + 𝛽𝑃 𝑋𝑃𝑖𝑗 ]

where logH0 (𝑡𝑗 ) is the general baseline log cumulative hazard function at each time
interval (this represents the value of the outcome when the predictors are set at 0), 𝑋1 through 𝑋𝑃
represents the predictors, and β measures the predictors of the log cumulative hazard.

50

CHAPTER 5: RESULTS

This chapter begins by presenting descriptive statistics for the sample and then examining
the effects of DMI on re-arrest for prohibited offenses, drug offenses, and any offense after three
years. Time-to-failure (i.e., re-arrest) is then analyzed for the DMI and comparison groups using
Kaplan-Meier survival curves and cumulative hazard and cumulative incidence distributions.
Finally, the effect of involvement in DMI on time-to-failure is examined using the Cox
regression model.
Sample Characteristics
This study compared a sample of 81 offenders who attended a DMI call-in meeting to a
sample of 162 offenders arrested for selling drugs or possession with the intent to sell drugs
around the time of a call-in meeting. Table 2 summarizes the general background characteristics
of the two groups, including demographic characteristics (age at the beginning of the study
period, sex, and race); and criminal history characteristics (number of arrests three years preintervention and most common type of offense three years pre-intervention). Due to propensity
score matching the groups did not differ significantly with respect to the background
characteristics.
Offenders in the DMI group were approximately the same age as those in the comparison
group (𝑋̅ = 27.58 and 27.04, respectively). Most offenders were male, 88.9 percent for the DMI
group and 89.5 percent for the comparison group, and most were Black, 97.5 percent for the
DMI group and 93.8 percent for the comparison group. The demographic characteristics indicate
that the sample consisted mostly of young Black males.

51

Criminal history characteristics showed that on average the DMI group incurred more
arrests than the comparison group (𝑋̅ = 5.40 and 4.74, respectively). Most offenders involved in
DMI (62.8% or N=49) and a little more than half of those in the comparison group (52.6% or
N=81) were most commonly arrested for drug offenses. This was followed by property crime
arrests (DMI = 17.9%; comparison = 27.9%), violent crime arrests (DMI= 14.1%; comparison =
9.7%), and other arrests (DMI= 5.1%; comparison= 9.7%).
Re-arrest Characteristics
Table 3 presents data on whether and how often offenders in the study were re-arrested
during the three-year observation period. While both groups’ average number of arrests dropped
during the observation period, the offenders in the DMI still averaged more arrests than the

52

comparison group. The mean number of arrest for the DMI group was 4.27, while the mean
number of arrest for the comparison group was 3.35, and this difference was statistically
significant (p<0.10).
Results of the bivariate analysis of re-arrest for prohibited offenses, drug offenses, and
any offense are also presented in Table 3. About 48 percent of offenders involved in DMI and
36 percent of offenders from the comparison group were re-arrested for a prohibited offense.
The results show that involvement in DMI is associated with re-arrest for a prohibited offense.
Contrary to the hypothesis, offenders involved in DMI were re-arrested more for a prohibited
offense than those in the comparison group. About half the offenders in both groups (DMI=
51.9%; comparison= 50.0%) were re-arrested for a drug offense and most offenders in both
groups (DMI= 79.0%; comparison= 85.2%) were re-arrested for any offense. No significant
differences were found between the groups with respect to the proportion of offenders re-arrested
for a drug offense or any offense.

53

To examine the effect of involvement in DMI on re-offending at the end of the three-year
study period, logistic regression was used to predict re-arrest for a prohibited offense, drug
offense, and any offense. Group membership (DMI vs. comparison) was used as the predictor
variable, while controlling for sex, race, age, number of previous arrests, and most common type
of offense. Table 4 presents the results of the logistic regression models for prohibited re-arrests,
drug re-arrests, and any re-arrest. The logistic regression coefficient 𝛽, can be interpreted as the
change in the log odds of the dependent variable (i.e., re-arrest) for a one-unit change in the
predictor. Exp(β) represents the “odds” of re-arrest given a unit change in the predictor variable
(controlling for other the other variables in the model). Values greater than one indicate
increased odds and values less than one indicate decreased odds. The model for prohibited rearrests indicates that the DMI group was more likely to be re-arrested for a prohibited offense.
Although this difference is not significant, the results did not support the hypothesis. Age and
prior number of offenses are statistically significant; the odds of re-arrest are higher for younger
offenders and offenders with more arrests.
The results of the logistic regression for drug re-arrests and any re-arrest indicate that the
DMI group was less likely to be re-arrested than the comparison group. While this difference
was not significant for drug offenses,13 it was significant for any offense (p<0.10). The odds of
getting re-arrested for any offense were lower by a factor of 4.94 for offenders involved in the
DMI. Age and number of previous offenses were significant in the drug re-arrest and any rearrest model, consistent with the prohibited re-arrest model.

13

The summary statistics for the drug re-arrest model did not indicate a significant “goodness of fit.” The
independent variables did not predict a significant amount of variance in re-arrest of drug offenses.

54

Table 4. Logistic Regression Models Estimating the Effect of DMI on Prohibited Re-arrest, Drug Re-arrest, and Any Re-arrest

DMI
control variables
Sex
Race
Age
Prior Arrests
Common Arrestᵃ
Violent
Property
Drug
Constant

β
0.267

Prohibited
S.E.
Sig.
0.309
0.388

Exp(β)
1.306

β
-0.052

S.E.
0.292

Sig.
0.859

Exp(β)
0.950

β
-0.681

S.E.
4.14

Sig.
0.092*

Exp(β)
0.506

-0.180
-1.709
-0.044
0.090

0.558
7.093
0.020
0.041

0.836
0.181
0.957
1.094

0.203
0.448
-0.029
0.096

0.500
0.668
0.018
0.039

1.225
1.225
0.971
1.071

-0.190
0.637
-0.046
0.276

0.694
1.144
0.022
0.086

0.701
0.625
0.591
0.909

4.818
1.227
1.786
2.838

0.278
-0.005
0.41
0.122

0.627
0.556
0.522
0.841

1.32
0.995
1.507
1.129

-1.663
-1.841
-1.696
4.044

1.214
1.142
1.086
1.511

0.785
0.578
0.042**
0.001***
0.446
0.171
0.107
0.118
0.007

0.827
1.891
0.956
1.318

1.572
0.204
0.580
1.043

0.685
0.685
0.099*
0.076*
0.646
0.658
0.993
0.432
0.885

0.747
0.118
0.029**
0.030**
0.083*
0.025**
0.744
0.327
0.251

Drug

Any

Model Chi-square
32.352**
10.482
29.654***
Degrees of Freedom
8
8
8
R²
0.175
0.059
0.205
ᵃOther arrest is the reference category
*p<0.10, **p<0.05, ***p<0.001
Note: The pseudo R² is not the proportion of the variance in re-arrest explained by the model. In logistic regression it tends to underestimate the
explanatory power, therefore it is a rough approximation of the effectiveness of the model.

55

0.190
0.159
0.183
57.073

Survival Analysis - Time-to-Failure

Step one in a continuous time survival analysis is describing the distribution of event
occurrences through examination of the survival curve, cumulative hazard and cumulative
incidence functions, and median lifetime (i.e., the distributions center). The median lifetime is
the time at which the value of the estimated survivor function is 0.5; specifically, the point at
which it is estimated that half the sample has experienced the event and half has not. This
identifies the average survival time for the sample.14
Table 5 presents the survival probability and median lifetime by group for prohibited rearrests, drug re-arrests, and any re-arrest. During the three-year period, thirty-nine offenders
involved in DMI were re-arrested for a prohibited offense (48.1%), forty-two were re-arrested
for a drug offense (51.9%), and 64 were re-arrested for any offense (79.0%). While fifty-nine
offenders in the comparison group were re-arrested for a prohibited offense (36.4%), eighty-one
were re-arrested for a drug offense (50.0%), and 138 were re-arrested for any offense (85.2%).
The probability of survival for prohibited offenses did not drop below 0.5 for either group; as a
result, the median lifetime could not be calculated. This indicates that for both groups the rate of
re-arrest for a prohibited offense was low. For drug re-arrests, the probability of survival just
drops below 0.5. The estimated median lifetime reveals that on average offenders involved in
DMI were re-arrested for a drug offense 924 days after the start of the study period, while an

14

Since events are censored, time is often skewed. Therefore, it is not appropriate to compute the mean, and the
median lifetime is used.

56

Table 5. Median Lifetime, Survival Probability, and Log-rank and Breslow Tests of
Re-arrest for the Three Year Observation Period
Number of
re-arrests
n(%)

Median
Lifetime
(days)

Survival
Log-rank
Probability
Sig.

Breslow
Sig.

0.144
Prohibited Re-arrestᵃ
0.113
DMI (n= 81)
39 (48.1)
0.50
Comparison (n=162)
59(36.4)
0.58
Drug Re-arrest
0.922
0.694
DMI (n= 81)
42(51.9)
924
0.45
Comparison (n=162)
81(50.0)
904
0.43
Any Re-arrest
0.458
0.68
DMI (n= 81)
64(79.0)
164
0.20
Comparison (n=162)
138(85.2)
186
0.11
ᵃSince survival probabilities barely reached 0.50 median lifetimes could not be calculated.

offender in the comparison group was re-arrested 904 days after the start of the study period
(about 24 days later than the DMI group). Finally, for any re-arrest, the probability of survival
drops well below .50 for each group, which indicates a high rate of recidivism for both groups,
with offenders in the comparison group having the lowest survival rate. Although offenders in
the comparison group had a lower probability of survival, the estimated median lifetimes reveal
that on average, an offender in the comparison group was re-arrested for any offense about 22
days later than offenders in the DMI group (186 days vs. 164 days).
Survival Curves
Survival and hazard functions are often best presented graphically because the key point
is determining the general shape of the patterns rather than the precise numeric values. Figure 2
displays Kaplan-Meier survival curves stratified by group for prohibited re-arrests. At the
beginning of time, when no one has been re-arrested everyone is surviving, so by definition the
value is one. Over time as events occur, the survivor function declines toward zero. The graph
shows that the comparison group had a consistently higher survival distribution that the DMI
57

group. This signifies that the comparison group committed prohibited offenses at a slower rate
than the DMI group. The log-rank test and Breslow test (Table 5) indicate that the differences
between the survival curves are not significant. Involvement in DMI did not change the survival
distribution for offenders during the three-year follow-up.

58

Figure 3 and Figure 4 display survival curves for drug re-arrests and any re-arrest,
respectively. The graph for drug re-arrests (Figure 3) shows that the two groups were re-arrested
at the same rate. For about the first year, offenders involved in the DMI had a lower probability
of survival; however, by the end of the three-year period there was no difference between the
two groups.

59

The graph for any re-arrest (Figure 4) shows that early in the study period the probability
of survival was lower for offenders in the DMI group, then just more than one year into the study
period their probability of survival became better than that of the comparison group and
consistently stayed higher. For both drug re-arrests and any re-arrest, there were no significant
differences between the survival curves for the groups (Table 5). Over the three-year period
offenders in the DMI group and offenders in the comparison group committed drug offenses and
any offense at about the same rate.

60

Cumulative Hazard and Cumulative Incidence Functions

When examining cumulative hazard function and cumulative incidence functions it is
important to assess how the level of the underlying hazard functions differ between the groups.
An examination of the level (e.g., the vertical distance between the functions) of the cumulative
hazard determines how different the underlying hazard functions are. Figure 5 and Figure 6
display the cumulative incidence functions for prohibited re-arrests and drug re-arrest stratified
by group, while Figure 7 display the cumulative hazard functions for any re-arrest.

61

For prohibited re-arrests (Figure 5), the cumulative incidence for those in the DMI group
was consistently higher and the distance between the functions widened as time went on. This
suggests that involvement in DMI was related to a higher risk of re-arrest for prohibited offenses.
The cumulative incidence functions for drug re-arrests (Figure 6) are similar and do not diverge;
suggesting that the risk of re-arrest for drug offenses was the same for both groups. Finally, for
any re-arrest (Figure 7), the cumulative hazard function for the DMI group starts higher, and a

62

little after one year the hazard function for the comparison group rises above the DMI group and
the distance between the functions widens. This suggests that the risk of re-arrest was higher for
the DMI group early on; however, those who survived the high risk time had a lower risk of rearrest than offenders in the comparison group.

63

Cox Regression

Cox regression analysis is used to examine the effect of involvement in DMI on time to
re-arrest. Cox regression measures the relative hazard risk for re-arrest not the absolute hazard
risk; therefore comparative statements about the hazard can be made, but not absolute statements
(Singer & Willett, 2003). Table 6 presents the results of the Cox regression for prohibited-rearrests, drug re-arrests, and any re-arrest. Model 1 displays the effects of DMI on time to rearrest controlling for time-invariant covariates. The hazard ratio is the predicted change in the
hazard of re-arrest for a one-unit change in the predictor (controlling for other variables in the
model). Values greater than one indicate increased hazard risk and values less than one indicate
decreased hazard risk. Model 1 shows that the risk of re-arrest for a prohibited offense was
greater for offenders in DMI than offenders in the comparison group; however, this difference
was not significant. Similar to the results of the logistic regression, these results did not support
the hypothesis. Offenders involved in DMI had a shorter time to re-arrest than offenders in the
comparison group. The results of Model 1 for drug re-arrest shows that the risk of re-arrest for
offenders in the DMI group did not differ significantly from those in the comparison group.
Although not significant, this outcome is moving in the hypothesized direction; the risk of rearrest for a drug offense was greater for the comparison group than for the DMI group.
A look at the cumulative hazard functions for any re-arrest (Figure 7) shows that the
functions cross during the observation period. This is a sign that the hazards are not proportional.
The effect of involvement in DMI may differ at different time periods; however, the difference is
not systematically related to time. A segmented time-dependent covariate was constructed by
taking the product of time and involvement in DMI (T_DMI). A test of this covariate indicates
64

that the effect of DMI on re-arrest was time dependent; therefore the term was added to the
model for any offense to control for this dependence. The results of Model 1 show that
involvement in DMI did not have an effect on re-arrest for any offense. There was no significant
difference between the groups on the risk of re-arrest for any offense.
To control for the fluctuating nature of crime rates which may affect risk differently
during different periods of time, violent, property, and drug crime rates were added to the model
as time-varying covariates (Model 2). The results indicate that the violent and property crime
rates affected the risk of re-arrest for prohibited offenses and drug offense, while only the
property crime rates had an effect on any offense. Although the addition of these crime rates
improved the model, they did not change the effect of DMI on re-arrest; it remained non
significant.

65

Table 6. Cox Regression Models Estimating the Effect of DMI on Prohibited Re-arrest, Drug Re-arrest, and Any Re-arrest Hazard Rates

DMI
control variables
Sex
Race
Age
Prior Arrests
Common Arrestᵃ
Violent
Property
Drug
T_DMI
Violent crime rate
Property crime rate
Drug crime rate

Prohibited
Model 1
Model 2
Hazard
Hazard
Ratio
Sig.
Ratio
Sig.
1.096
0.673
1.198
0.414
1.182
0.218
0.967
1.060
3.794
1.510
1.993
-

0.687
0.136
0.038**
0.010**
0.012**
0.010**
0.420
0.158
-

1.158
0.220
0.969
1.059
3.612
1.427
2.002

0.728
0.138
0.051*
0.011**
0.014**
0.013**
0.490
0.156

1.000
1.000
1.000

0.014**
0.002**
0.311

Drug
Model 1
Hazard
Ratio
Sig.
0.874
0.499
1.366
1.339
0.980
1.063
1.593
1.130
1.694
-

0.390
0.502
0.139
0.005**
0.266
0.306
0.767
0.173
-

Any

Model 2
Hazard
Ratio
Sig.
0.852
0.442
1.089
1.352
0.981
1.070
1.621
1.349
2.065
0.999
1.000
1.000

0.815
0.490
0.150
0.002**
0.148
0.293
0.475
0.066*
0.000***
0.000***
0.998

Model 1
Hazard
Ratio
Sig.
1.177
0.438
1.388
1.114
0.973
1.109
0.871
1.098
1.101
0.998
-

0.240
0.752
0.007**
0.000***
0.803
0.681
0.743
0.722
0.023***
-

Model 2
Hazard
Ratio
Sig.
1.154
0.499
1.270
1.033
0.970
1.111
0.739
1.026
1.137
0.998
1.000
1.000
1.000

0.402
0.924
0.004**
0.000***
0.449
0.376
0.969
0.637
0.032**
0.149
0.000***
0.543

Model Chi-square
38.368***
63.024***
18.136**
71.497***
63.902***
90.382***
Degrees of Freedom
8
11
8
11
9
12
-2 Log Likelihood
951.959
929.235
1185.083
1141.668
1795.41
1772.833
Note: The -2 Log Likelihood is the measure of model fit. The reduction of the -2 Log Likelihood between models signifies that the model was improved by adding the
crime rates.
ᵃOther arrest is the reference category
*p<0.10, **p<0.05, ***p<0.001

66

CHAPTER 6: DISCUSSION AND CONCLUSION

Research examining the effects of focused deterrence strategies have typically
concentrated on the community-level effects of the strategies, most notably, change in homicide
rates (Braga et al., 2001, 2008; Braga, McDevitt & Pierce, 2006; Corsaro & McGarrell, 2009,
2010; Engel et al., 2001; Kennedy & Braga, 1998; McGarrell et al., 2006; Papachristos, Meares,
& Fagen, 2007), gun assaults (Braga et al., 2001; Braga, McDevitt & Pierce, 2006; Skogan et al.,
2009), shots fired calls (Braga et al., 2001; Kennedy & Wong, 2009), and crimes associated with
drug markets (Corsaro & Brunson, 2013; Corsaro, Brunson, & McGarrell, 2010; Frabutt et al.,
2009; Hipple, Corsaro, & McGarrell, 2010; Kennedy & Wong, 2009). With the exception of
Chermak (2008) and Papachristos et al (2013), individual-level effects of the offenders targeted
for these interventions have not been examined. In High Point specifically, studies on the DMI
have found it to have a positive crime-control effect at the community level (Frabutt et al., 2009;
Hipple, Corsaro, & McGarrell, 2010; Kennedy & Wong, 2009), but little is known about the
outcomes of the individuals targeted for the intervention. Using survival analysis to compare the
recidivism patterns of offenders who attended a DMI call-in meeting relative to those of an
equivalent comparison group, this study examines the DMI’s specific deterrent effect on the
offenders targeted to participate in the intervention. This final chapter begins with a summary
and discussion of the findings, followed by a discussion of the limitations, theoretical
implications and suggestions for future research, and concludes with implications for policy.

67

Summary and Discussion

The hypotheses explored in this study were:
1) Offenders who attend a DMI call-in meeting will have a lower incidence of re-offending
than those of an equivalent comparison group.
2) Offenders who attend a DMI call-in meeting will have a longer time to re-arrest (time-tofailure) than an equivalent comparison group.
With certain components of the intervention aimed at influencing offender behavior (i.e., the
call-in where the deterrent message is delivered, family involvement, social services, law
enforcement follow-up, and enforcing consequences) theory would suggest that the extra
attention given to those in the DMI group would result in a lower level of re-offense and a longer
time to re-offense than those who received the traditional criminal justice treatment (comparison
group). The hypotheses were tested for prohibited offenses (those offenses that the DMI
officials told offenders would incur increased sanctions), any drug offense, and any offense.
Prohibited Offenses

The analyses of prohibited offenses revealed no support for the hypotheses. The results
of the bivariate analysis showed that the incidence of re-arrest was not lower for the DMI group;
in fact, offenders involved in DMI had a significantly higher number of re-arrests for a
prohibited offense than offenders in the comparison group. Findings from the logistic regression
revealed that participation in DMI was not a significant predictor of re-arrest for a prohibited
offense. Though not significant, the effect of participation in DMI on re-arrest was not moving in
the expected direction; the odds of re-arrest were more likely for offenders involved in DMI than
68

those in the comparison group. Similar results were obtained from the survival analysis. The
survival probability for both groups never dropped below 0.5 indicating that both groups had a
low rate of re-arrest for prohibited offenses. However, the survival curves and cumulative
incidence functions for prohibited offenses indicated that offenders in DMI were consistently at
greater risk of re-arrest and had a shorter time to re-arrest than the comparison group.
While these results are not consistent with the hypotheses, they may indicate success for
one of the components of the intervention: law enforcement follow-up. Offenders in the DMI
are supposed to be watched closely by DMI officials and evidence from the re-entry literature
has found that enhanced surveillance can increase officials’ ability to detect criminal behavior
(Petersilia & Turner, 1993). If DMI officials are keeping a closer watch on offenders in the
intervention, it follows that they would experience a greater risk of re-arrest than offenders in the
comparison group. Additionally, one of the main goals of the DMI is improving the
communities where it is implemented by reducing the crime rate. As a result, if offenders are
continuing to commit prohibited offenses after attending the call-in, getting them “off the streets”
is what matters. Because these offenders were thought to be the ones driving the drug problem
in the communities, the quick removal of the offenders who continued to commit prohibited
criminal behavior may have contributed to the success of the DMI at the community-level.
The results for prohibited offenses were also inconsistent with the study conducted by
Papachristos et al. (2013) for PSN Chicago, which used a focused deterrence strategy to reduce
the city’s homicide and gun violence rates. They found that offenders who participated in the
PSN offender notification forums, on average, stayed out of prison longer and had a lower
hazard of committing a new offense than offenders in the comparison group (e.g. those that did
69

not participate in an offender notification forum). The difference between these findings may be
because the PSN Chicago strategy did not call for enhanced surveillance of attendees after the
forum.
Drug Offenses

Results from the analyses of drug offenses showed no support for the hypotheses. The
bivariate analysis revealed no differences in the incidence of re-arrest; the proportion of
offenders re-arrested for a drug offense did not differ significantly between the groups. Results
from the logistic regression suggest that participation in DMI was not a significant predictor of
re-arrest for a drug offense. At the end of the three-year study period, involvement in DMI was
not found to have an effect on the incidence or odds of re-arrest for a drug offense. The survival
curves and cumulative incidence functions for drug re-arrests revealed no difference between the
groups. The Cox regression models also found that DMI had no effect on the hazard rate for
drug offenses. Time to re-arrest was not significantly influenced by involvement in DMI.
Any Offense

Results from the analyses of any offense showed little support for the hypotheses. The
bivariate analysis revealed no differences in the incidence of re-arrest for any offense; however,
the results of the logistic regression suggest that participation in DMI was a significant predictor
for any offense. The odds of being re-arrested for any offense were lower for offenders in the
DMI. The survival curves and cumulative hazard functions for any re-arrest revealed no
differences between the groups. Findings from the Cox regression model suggest that time to rearrest for any offense was not significantly influenced by involvement in DMI.
70

The findings for drug offense and any offense categories are consistent with Chermak’s
(2008) study of the Indianapolis Lever-Pulling experiment, which found that there was no
difference between the two treatment groups15 and control group on likelihood of recidivism and
time to arrest post-intervention meeting for all offenses and felony offenses. The DMI’s lack of
effect on re-arrest for drug offenses and any offense may stem from the components of the
intervention meant to influence offender behavior not working or being implemented as
intended. For example, Chermak (2008) found that probationers in the law enforcement and
community treatment groups were no more likely to take advantage of community programming
(education programs, treatment programs, work programs, and faith-based programs) than the
control group. This indicates that having services available does not guarantee that offenders
will take advantage of them. If offenders are not taking advantage of the services being offered,
then the services cannot have the intended influence on offender behavior.
Limitations
This study has several limitations. First, when conducting propensity score matching,
omitting important variables could increase the bias in the resulting estimates (Caliendo &
Kopeinig, 2005; Heckman, Ichimura, & Todd, 1997). Matching for this study was done using
variables constructed from the available data. Important potential confounders may have been

15

The study had two treatment groups. The law enforcement group attended a call-in at a courthouse with
several law enforcement officials in attendance. Probationers were given the message that violence would not be
tolerated and they would incur a number of sanctions if it continued. They were also told services were available
to them if they wanted to take advantage of them. The community group attended a call-in at a community
center. It was less formal with the only law enforcement presence being a probation officer to check probationers
in to the meeting. The deterrence message was delivered by community members who spoke about the problems
violence has caused in the community. Probationers were told services were available and they should contact
community leaders; however, no possible sanctions were discussed during the meeting.

71

omitted due to data limitations. Bootstrapping would have been one way to handle this issue;
however, this procedure could not be preformed due to limitations of the analysis program.
Second, the small sample size of the treatment (DMI) and comparison group limits the
ability to perform additional analyses that could detect conditional relationships between the
intervention and other variables. Analyzing separate models allows researchers to determine if
the combined effects of the control variables behave differently across models (Paternoster et al.,
1998). It would be beneficial to run separate models for the treatment (DMI) and comparison
group to examine the effects of the control variables on the populations separately. If, for
example, age and previous arrests lose significance in the treatment model but not in the
comparison model; the intervention could be said to have a conditional effect on the risk of rearrest. The small sample size of offenders in each call-in period also limited the ability to
examine effects of the intervention across intervention sites; effects of the intervention had to be
aggregated across call-in periods. Small sample sizes are closely related to measures of
statistical power and given the small number of offenders in each call-in (Lipsey, 1999).
Third, this study theorizes that certain components of the intervention such as social
services, family involvement, and enforcing consequences would influence offender behavior;
however, these components could not be tested directly. Without testing these components
directly their effects can only be speculated. This study was also unable to determine if these
components were implemented as intended (for example of services not being implemented as
intended see Engel et al. 2011). If components are not fully realized, then effects of the
components can not truly be tested, this is discussed further in the following section.

72

Future Research and Theoretical Implications

Although prior research suggests that the DMI intervention is having a positive effect on
neighborhood levels of crime and quality of life, the results from this study suggest that the callin meetings are not having the anticipated specific deterrent effect on the drug offenders that
participate in the call-in meetings. As noted above, components of the intervention that are
thought to influence offender behavior such as social services, family involvement, enforcement
consequences, the call-in process and deterrence message could not directly be tested. Future
research should focus on how these components directly affect offender behavior.
Social services. Studies have found that offenders in these interventions rarely take advantage of
the social service programs For example, Engel et al (2011) found that the social service program
for the Cincinnati Initiative to Reduce Violence (CIRV) was not being carried out as expected.
They suggest that the social service program showed no impact on the reduction of violence in
the city because it was not implemented as expected (also see Chermak, 2006; Tillyer et al.,
2010). If offenders are not taking advantage of the services, then the effect of the services can
not actually be measured. Future studies should examine if, and to what extent, services were
being utilized by offenders in the DMI to determine if the purpose behind having social services
was achieved as planned.
Family involvement. Family members and loved ones of the offenders (i.e., influentials) were
invited to the call-ins under the premise that individuals significant to the offenders would be
able to convince them to stop offending by creating and reinforcing positive norms and
expectations (Kennedy & Wong, 2009). Researchers suggest that significant people in an
individual’s social environment can be used as a form of informal social control (Grasmick &
73

Bursik, 1990; Williams & Hawkins, 1986). Additionally, evidence from perceptual deterrence
theory points to the importance of social censure from important others as a significant factor in
inhibiting crime, often more important than fear of legal sanctions (Paternoster & Bachman,
2013). However, it is not known whether these influentials continued to support the offenders
after the call-in, or if they were positive influences to begin with. For example, a DMI official
interviewed in Rivers, Norris, and McGarrell (2012) explained that some offenders sold drugs
because their mothers made them (pg. 72). The involvement of these types of family members
would defeat the purpose of family involvement. Future research on this component should focus
on the level and type of support that influentials provide to the offenders.
Follow-up. When looking at the combined results of the prohibited offense, drug offense, and
any offense analyses there are indications that the intervention may be having an effect on police
discretion. Whereas increased surveillance helps explain why DMI clients are arrested faster and
more often than offenders in the comparison group for prohibited offenses, it does not explain
why there were no differences between the groups for the drug and any offense categories. If
DMI clients are being surveilled more often, then it stands to reason that they should have been
arrested faster and more often for drug and any offenses as well. However, these results can be
explained if there were factors limiting officer discretion, especially for prohibited offenses. To
enforce the deterrence message officers were supposed to arrest DMI clients for committing
prohibited offenses. That is, as part of the DMI intervention officers may have felt they had no
choice. For non-prohibited offenses, however, officers may have been able to exercise more
discretion on how they handled the situations. This may reflect a situation whereby for DMI
clients the prohibited offenses reflected “non-normal” crimes that thereby limited officer
discretion and resulted in arrests (Sudnow, 1965). Therefore, even if officers did come in contact
74

with DMI clients committing minor offenses they may have let it go because it did not rise to the
level of a prohibited offense.16
Future research should examine whether the offenses offenders in each group were
arrested for can be subject to discretion. For example, prohibited offenses could be broken down
by level of seriousness to examine whether DMI clients are more likely to be arrested for less
serious prohibited offenses, thereby suggesting restricted officer discretion for prohibited
offenses. The number of police contacts could also be examined to determine if DMI clients had
more contact with officers as predicted by their increased surveillance. This may indicate that
officers have more opportunities to exert discretion but were restricted from doing so because of
DMI program rules.
Enforcing consequences. One of the consequences laid out by program officials were harsher
punishments for offenders who committed prohibited offenses. The increased certainty and
severity of sanctions put forth in the deterrence message was intended to act as a deterrent.
While program officials seemed to deliver on the increased certainty of arrest, they may not have
followed through on the increased severity. For example, Chermak (2008) found that
probationers’ in the treatment groups did not incur harsher penalties or have more levers pulled
than the control group. Therefore, if, as theory suggests, individuals change their criminal
behavior in response to the consequences that are incurred (or not incurred) by their behavior,
officials not delivering on the harsher sanctions may keep offenders from altering their behavior
in the intended way (Nagin, 1998; Piquero & Pogarsky, 2002; Pogarsky et al., 2004). Future

16

While conducting interviews of DMI experts to construct a mental model of the DMI for River, Norris, and
McGarrell (2012), some DMI officials mentioned that if they saw offenders in the program committing minor
offenses they would often just give them a warning. This was to let them know they were serious about trying to
help them and giving them a second chance.

75

research on this component should compare the length of sentences incurred by offenders in the
treatment and comparison group for prohibited offenses to determine if DMI clients actually
receiver harsher sanctions. Then, to fully examine this component, a recurrent event survival
analysis should be conducted to examine the effect of sentence length on subsequent re-arrests.
Program officials look to make examples out of offenders that do not comply; however, if they
do not deliver on the stated consequences then the intervention loses credibility.
Call-in and Deterrence message. Both deterrence theory (Nagin, 1998; Paternoster, 2010) and
procedural justice (Tyler, 2003, 2006) use offender perceptions to try to influence behavior;
deterrence theory through changes in the perception of certainty, severity, and swiftness of
sanctions and procedural justice through perceptions of the fairness of the process and legitimacy
of the authority. The DMI could be seen as utilizing the concepts of both theories
simultaneously to change the behavior of offenders in the program. During the call-in a
deterrence message is directly communicated to the offenders that looks to increase their
perceived certainty of getting arrested, the perceived swiftness with which it will happen, and the
perceived severity of their subsequent punishment. Although DMI officials are delivering a
deterrence message that could seem like a threat, they are supposed to communicate it in a
respectful manner, eliciting feelings of fairness from the offenders. Even though offenders in the
DMI are told they would be watched more closely and would incur harsh sanctions, they were
also being given a second chance (not being arrested for a case that was already made) and were
being offered support by program officials and members of the community. It was theorized that
offenders would change their behavior if they perceived an increase in the certainty, severity, and
swiftness of the sanctions laid out by the deterrence message delivered during the call-in and if
they found the policies and procedures of the intervention to be fair and just. This study,
76

however, was not able to test how offenders perceived the process, or the message that was
delivered to them.
If there are inconsistencies in the clarity of the deterrence message or in how the message
is delivered then it could change how the offenders perceive the message and their subsequent
behavior. For example, DMI officials could gain cooperation if offenders “buy-in” to the
intervention, but according to Tyler (2003) this usually cannot be done by using threats. If the
deterrence message is viewed as a threat, offenders may be less likely to buy-in to the
intervention and comply with program officials or take advantage of the services they offer. If
offenders feel like they are being threatened they may no longer feel like the process is fair or the
authority figures are being just. The context in which the message is delivered may also be
important. Offenders may view the process as threatening and unfair because the message is
delivered at police headquarters in front of the community. If offenders feel disrespected or
embarrassed by the process, they may be less likely to view the process as legitimate and
therefore, less likely to comply with intervention rules and policies.17

17

While Papachristos et al (2013) suggests that offender success in their study was due to the strengthened
perceptions of legitimacy and fairness gained from the PSN Chicago offender notification forums; they did not
actually test whether offenders perceived the process to be more legitimate. The PSN Chicago offender notification
forums were also conducted in a different setting and under a different context than DMI call-ins. For example,
Chicago forums were held in neutral locations like public buildings or community centers while DMI call-ins were
held at High Point police headquarters, a more formal and possibly intimidating location. This could have an effect
on how offenders view the legitimacy and fairness of the process. Future research is needed to better understand
how factors like meeting setting, number and nature of speakers, order of speakers (e.g., criminal justice official
versus community members and social service representatives) influence the perception of the meeting attendees.

77

Deterrence researchers have discussed the importance of measuring perceived deterrence;
especially how perceptions of sanction threats are formed and modified over time (Paternoster &
Bachman, 2013; Piquero & Pogarsky, 2002; Pogarsky et al., 2004). Perceptual deterrence theory
assumes that motivation to commit crime is affected by perceived consequences of criminal
behavior, positive or negative (Paternoster & Bachman, 2013). That is, offenders’ decisions will
be more influenced by what they think or perceive the sanctions to be, regardless of what they
actually are. Therefore, it is important to understand how offenders who attended call-in
meetings perceived the deterrence message and the consequences that were presented.
Communication Theory and Call-in Meetings. One way to examine how offenders process and
perceive the deterrence message is through the examination of fear appeals, specifically utilizing
the Extended Parallel Process Model (EPPM). The EPPM is a fear appeal theory which accounts
for responses to fear appeal messages that lead to both success (message acceptance) and failure
(message rejection) (Witte, 1992; Witte, Meyer, & Martell, 2001). The EPPM suggests that two
distinct processes occur when responding to fear appeals: a danger control process and a fear
control process (Witte, 1992; Witte, Meyer, & Martell, 2001). The danger control process is the
cognitive process that occurs when individuals are dealing with a danger or threat. It leads to
protection motivation, causing the individual to change their attitudes, intentions, or behaviors to
protect against the threat occurring (Witte, 1992; Witte, Meyer, & Martell, 2001). The fear
control process is the emotional process that occurs when individuals are faced with a significant
threat. This leads to defensive motivation, causing individuals to cope with the fear through
denial (e.g., a refusal to believe that they could experience the threat), defensive avoidance (e.g.,
distorting or blocking further information about the threat) or reactance (e.g., rejection or anger
at the message due to perceived manipulation) (Witte, Meyer, & Martell, 2001).
78

In the EPPM these two response processes are elicited through the interaction of three
constructs: fear, perceived threat, and perceived efficacy (Witte, 1992). Fear is a negatively
valenced emotion elicited through a threat which the individual perceives to be significant and
personally relevant (Witte, 1992). Perceived threat is the idea that a threat exists only because
the individual thinks it does and this perception can be real or imagined (Witte, 1992; Witte,
Meyer, & Martell, 2001). In the model, perceived threat is characterized by perceived
susceptibility and perceived severity. Perceived susceptibility refers to an individual’s beliefs
about their chances of experiencing the threat (Witte, 1992; Witte, Meyer, & Martell, 2001).
Perceived severity refers to an individual’s beliefs about how serious the threat is or how much
harm would occur if the threat were to take place (Witte, 1992; Witte, Meyer, & Martell, 2001).
Efficacy refers to the effectiveness of the recommended response (response efficacy) and the
individual’s ability to perform that recommended response (self-efficacy). Perceived response
efficacy is an individual’s “beliefs about the effectiveness of the recommended response in
deterring the threat” (Witte, Meyer, & Martell, 2001, p.20). Perceived self-efficacy refers to an
individual’s beliefs about his or her own ability to perform the recommended response to prevent
the threat (Witte, 1992; Witte, Meyer, & Martell, 2001).
The DMI’s deterrence message can be thought of as a fear appeal because the offenders
are warned that if they continue their criminal behavior it will result in the harshest possible legal
sanctions (Kennedy, 2009). They are presented with a message designed to scare them into
stopping their behavior and threatened with what will happen if they do not comply. The threat
perceived by the offenders includes the perceived susceptibility (e.g., whether the offenders
thinks they will go to jail if they continue their behavior) and the perceived severity (e.g., how
serious the offenders believe the harsh jail sentence to be). Efficacy involves whether the
79

offenders believe that stopping the behavior will keep them out of jail (i.e., perceived response
efficacy) and whether they believe they will be able to stop offending (i.e., perceived selfefficacy). The DMI tries to increase the offenders’ perception of self-efficacy by linking them
with programs and social services such as employment training, housing, mentoring, and
substance abuse treatment to provide them with options to help improve their situation.
When the offenders are presented with the deterrence message first the perceived threat
found in the message is appraised. If they perceive the threat to be low – there is no way they
will get arrested again and go to jail (e.g., perceived susceptibility) – and/or they do not think
going to jail is a big deal (e.g., perceived severity) then there is no response to the message; the
offenders stop processing the message. If they perceive the threat to be high – they believe they
will go to jail if they continue to sell drugs and/or they believe the stiff jail sentence to be serious
– and perceive efficacy to be high – the offenders believe that stopping their behavior will keep
them out of jail (e.g., perceived response efficacy) and they believe they will be able to stop their
behavior (e.g., perceived self-efficacy) – then danger control processes are initiated and the
offenders will try to control the danger of the threat by following the recommended response and
stopping their offending behavior. However, if the offenders perceive the threat to be high and
they do not believe that stopping their behavior will keep them out of jail – “the police will just
find something else to arrest me for because they are watching” (e.g., low perceived response
efficacy) – or they do not believe that they can stop offending – “there is no other way I can
make the kind of money I make from selling drugs” (e.g., low perceived self-efficacy) – then
fear control processes are elicited. As a result offenders might respond to the message through a
maladaptive response such as defensive avoidance (e.g., the thought of going to jail may be too
scary so they stop listening to the message), denial (e.g., “I won’t go to jail because I won’t get
80

caught again”), or reactance (e.g., “The police are just trying to manipulate me so I’m not going
to listen to them”) leading to message rejection and continued offending.
Procedural justice and perceptual deterrence theory stress the importance of
understanding how an individual’s perceptions influence behavior; whether it is the perception of
the legitimacy and fairness of the process or the perceptions of the certainly, severity, and
swiftness of the consequences. If future research can tease out how offenders are processing the
deterrence message and how it is delivered through the use of the EPPM, we may be able to
understand why the message is not having the expected effect on offender behavior. A better
understanding of how offenders who attend the call-in meetings perceive the process and the
deterrence message that is delivered to them during the meeting will also help inform better
implementation of DMI processes aimed at targeted offenders and delivery of the deterrence
message.
Policy implications

Overall, this study found that the DMI did not have a specific deterrent effect on the
targeted offenders. Although not the expected outcome, in the end, the DMI is designed to be a
community-based intervention, with the goal of eliminating drug markets and crime problems
associated with those markets and improving public safety in the neighborhoods. The DMI
appears to be effective at the community-level (Corsaro & Brunson, 2013; Corsaro, Brunson, &
McGarrell, 2010; Frabutt et al., 2009; Hipple, Corsaro, & McGarrell, 2010; Kennedy & Wong,
2009). And, although components of the intervention aimed at influencing offender behavior did
not have the intended effect on re-offending overall, that is, offenders in the DMI group were
arrested faster for prohibited offenses and outcomes of the offenders in the DMI group were no
81

better for drug offenses and any offense than those in the comparison group, findings from this
study suggest that certain components of the intervention aimed at individual-level effects (e.g.,
law enforcement follow-up and the call-in) may help contribute indirectly to the overall goal of
the intervention. Based on these findings and knowledge of the program, two implications for
policy are presented.
First, social services such as substance abuse treatment, housing assistance, and education
and employment assistance are thought to be important factors which will assist offenders in
reintegrating successfully in the community; however, participation in social services is
voluntary for offenders in the DMI. As stated earlier, other studies utilizing a focused deterrence
strategy have found that offenders targeted for the interventions rarely take advantage of the
services offered (Chermak, 2008; Engel et al., 2011; Tillyer et al., 2010). For example, in the
Cincinnati CIRV social service program, less than nine percent of the social service clients
actively participated in services (Engel et al., 2011). Program officials need to find a way to
increase participation in services while providing offenders with a caseworker or mentor to make
sure they stay on task and utilize the services they need. This could impact the effect of this
intervention component on offender outcomes. Rivers, Norris, and McGarrell (2012) suggested
that these services need to be resilient and redundant, and to accomplish this there needs to be a
wide range of overlapping services provided by multiple providers. Along with resilient and
redundant services, they suggest that the effectiveness of this component also lies in the
provider’s ability to follow through with the services. An example of this is illustrated by the
Boston Reentry Initiative (BRI). The BRI used a strategy similar to focused deterrence on
violent offenders being released from jail. Participants were identified while still incarcerated,
once identified they attended a BRI panel session where they were given the impression that
82

multiple law enforcement agencies would be watching them when they were released while also
being informed of community resources and institutional programs available to them. They were
assigned mentors and enrolled in a number of wrap-around services. An evaluation of the
program found that the BRI was associated with a 30 percent reduction in arrest rates for violent
and all offenses, comparing participants in the program to an equivalent control group (Braga et
al., 2009). Having treatment plans that offenders must follow can have a positive effect on their
reintegration in to the community and may result in the social services component having the
interned effect on the offenders.
Second, as noted above, findings from this study suggest that certain components of the
intervention aimed at individual-level effects may help contribute indirectly to the overall goal of
the intervention. As a result, even though the DMI does not seem to have a specific deterrent
effect, the intervention’s focus on the targeted offenders is still warranted. For example, results
from this study suggest that law enforcement follow-up may be working as intended with respect
to prohibited offenses. Contrary to what was hypothesized, offenders in the DMI were rearrested faster for committing prohibited offenses. This can be thought of as a success for the
overall goal of the intervention – improving the community – since offenders thought to be
driving the crime problem who are not complying with the program were arrested quickly.
Additionally, findings from this study suggest that the deterrence message delivered at
the call-in and other elements introduced at that time like the idea of family and community
support are not having the intended effect on the targeted offenders. If, as stated earlier,
delivering the deterrence message in front of the community has more of a negative connotation
for offenders, DMI officials should consider alternative ways to deliver the message while still
83

having a forum similar to a call-in. The true function of the call-in may not be to deter the
individual offenders, it may be to get the members of the community more involved and invested
in facilitating a change. In High Point, the community organization that offers services to
offenders, High Point Communities Against Violence, was borne out of the DMI. To be
effective the DMI requires strong community buy-in and in the end, the most visible aspect of
the DMI to the community is the actual call-in meeting (Rivers, Norris, & McGarrell, 2012).
Corsaro and Brunson (2013) suggest that the reason the intervention had no effect on crime rates
in Peoria was the lack of community awareness and involvement in the program. Ultimately,
community involvement is an important aspect of the DMI. DMI officials and community
members recognize that “…the intervention’s success is contingent on addressing community
disorganization and reestablishing levels of informal social control and collective efficacy
necessary to prevent the drug market from reemerging in the target neighborhood” (Rivers,
Norris, & McGarrell, 2012, pg. 71). The call-ins are a way to increase collective efficacy by
empowering communities to mobilize and exert informal social control (Sampson, Raudenbush,
& Earls, 1997). Corsaro et al. (2012) suggested that effects of the High Point DMI may be
wearing off at the community-level. A community forum similar to a call-in may be needed to
keep community members invested.
Although this study indicates that the DMI did not have the expected specific deterrent
effect on the targeted offenders, there is some evidence that continuing the focus on individual
offenders contributes to the overall program goals. Once components of the intervention such as
family involvement, social services, enforcement consequences, and perceptions of the call-in
and deterrence message are examined directly, researchers will have a better idea of how to

84

improve individual-level effects of the DMI while at the same time disentangling the
community-level and individual-level effects of the program.

85

APPENDIX

86

APPENDIX
High Point Police Department Prohibited Offenses for the Drug Market Intervention
Armed Robbery
Arson 1st & 2nd Degree
Assault Attempt to Inflict Serious Injury
Assault ISI
Assault ISI on LEO, School Official or State Employee
Assault by Pointing a Gun
Assault on a Handicapped Person
Assault with a Deadly Weapon on LEO, School official or State Employee
AWDW for Escape
AWDWISI
AWDWIKISI
AWDWWITKISI
Assault with intent to commit rape
Common Law Robbery
Deadly Weapon on State Property
Death by Duel
Discharging Firearm into Property
Felony Fleeing to Elude Arrest
Homicide
Intentional Death of Patient
Involuntary Manslaughter
Involuntary Manslaughter 2nd Offense
Kidnapping 1st & 2nd Degree
Kidnapping/Abduction
Malicious Castration
Malicious Maiming
Manslaughter
Manslaughter 2nd Degree
Murder 1st & 2nd Degree
Possession of Weapon of Mass Destruction
Possession of Firearm by Minor
Possession of Machine Gun
Possession of Weapon on Campus/School Property
Riot & Civil Disorders
Robbery
RWDW
Possession with Intent to Sell/Distribute
Trafficking any Amount of Drugs
87

REFERENCES

88

REFERENCES

Apel, R., & Sweeten, G. (2010). Propensity score matching in criminology and criminal justice.
A. Piquero and D. Weisburd (eds.), Handbook of Qualitative Criminology. Springer
Science + Business Media.
Braga, A., & Bond, B. (2008). Policing crime and disorder hotspots. Criminology, 46, 577-607.
Braga, A., Kennedy, D., Waring, E., & Piehl, A. (2001). Problem-oriented policing, deterrence,
and youth violence: An evaluation of Boston’s Operation Ceasefire. Journal of Research
in Crime and Delinquency, 38, 195-226.
Braga, A., McDevitt, J., & Pierce, G. (2006). Understanding and preventing gang violence:
Problem analysis and response development in Lowell, Massachusetts. Police Quarterly,
9, 20-46.
Braga, A., Papachristos, A., & Hureau, D. (2010). The concentration and stability of gun
violence at micro places in Boston, 1980-2008. Journal of Quantitative Criminology, 26,
33-53.
Braga, A., Piehl, A., & Hureau, D. (2009). Controlling violent offenders released to the
community: An evaluation of the Boston Reentry Initiative. Journal of Research in
Crime and Delinquency, 46, 411-436.
Braga, A., Pierce, G., McDevitt, J., Bond, B., & Cronin, S. (2008). The strategic prevention of
gun violence among gang-involved offenders. Justice Quarterly, 25, 132-162.
Braga, A. & Weisburd, D. (2012). The effects of focused deterrence strategies on crime: A
systematic review and meta-analysis of the empirical evidence. Journal of Research in
Crime and Delinquency, 49(3), 323-358.
Braga, A. A., Weisburd, D., Waring, E., Green Mazerolle, L., Spelman, W., & Gajewski, F.
(1999). Problem-oriented policing in violent crime places: A randomized controlled
experiment. Criminology, 37, 541-580.
Brown, S., & Esbensen, F. (1988). Thoughts on deterrence: Evolution of a theoretical
perspective. International Journal of Offender Therapy and Comparative Criminology,
32, 219-231.
Caliendo, M., & Kopeinig, S. (2005). Some practical guidance for the implementation of
propensity score matching. Discussion Paper No. 1588. Germany: The Institute for the
Study of Labor.
89

Chermak, S. (2006). Reducing violent crime and firearms violence: The Indianapolis Lever
Pulling Experiment. Washington DC: US Department of Justice.
Cohen, J., & Ludwig, J., (2003). Policing crime guns. In Ludwig, J., and Cook, P. (eds.)
Evaluating Gun Policy: Effects on Crime and Violence. Washington, D.C.: Brooking
Institution Press.
Cook, T. & Campbell, D. (1979). Quasi-Experimentation: Design and analysis issues for field
settings. Boston, MA: Houghton Mifflin Company.
Corsaro, N., & Brunson, R. (2013). Are suppression and deterrence mechanisms enough?
Examining the “pulling levers” drug market intervention strategy in Peoria, Illinois, USA.
International Journal of Drug Policy, 24, 115-121.
Corsaro, N., Brunson, R., & McGarrell, E. (2009). Problem-oriented policing and open air drug
markets: Examining the pulling levers deterrence strategy in Rockford, Illinois. Crime
and Delinquency.
Corsaro, N., Brunson, R., & McGarrell, E. (2010). Evaluating a police strategy intended to
disrupt an illicit street-level drug market. Evaluation Review, 34(6), 513-548.
Corsaro, N., & McGarrell, E. (2009). Testing a promising homicide reduction strategy:
Re-assessing the impact of the Indianapolis “pulling levers” intervention. Journal of
Experimental Criminology, 5, 63-82.
Corsaro, N., & McGarrell, E. (2010). Reducing homicide risk in Indianapolis between 1997 and
2000. Journal of Urban Health, 87, 851-864.
Engel, R., Tillyer, M., & Corsaro, N. (2011). Reducing gang violence using focused deterrence:
Evaluating the Cincinnati Initiative to Reduce Violence (CIRV). Justice Quarterly.
Online DOI:10.1080/07418825.2011.619559.
Frabutt, J. M., Shelton, T. L., DiLuca, K. L., Harvey, L. K., & Hefner, M. K. (2009). A
collaborative approach to eliminating street drug markets through focused deterrence.
Washington, DC: U.S. Department of Justice.
Freedman, D., & Berk, R.. (2008). weighting regressions by propensity
scores. Evaluation Review, 32, 392-409.
Gibbs, J. (1968). Crime, punishment, and deterrence. Social Science Quarterly, 48, 515-530.
Goldstein, H. (1979). Improving policing: A problem-oriented approach. Crime & Delinquency,
25, 236-258.

90

Grasmick, H.G., & Bursik, R. (1990). Conscience, significant others, and rational choice:
Extending the deterrence model. Law and Society Review, 24, 837-861.
Groff, E., Weisburd, D., & Yang, S. (2010). Is it important to examine crime trends at a local
micro level? Journal of Quantitative Criminology, 26, 7-32.
Heckman, J., Ichimura, H., & Todd, P. (1997). Matching as an econometric evaluation estimator:
Evidence from evaluating a job training programme. Review of Economic Studies, 64,
605-654.
Hipple, N. K., Corsaro, N., & McGarrell, E. F. (2010). The high point drug market initiative: A
process and impact assessment. Paper prepared for Bureau of Justice Assistance. East
Lansing: Michigan State University.
Hinds, L., & Murphy, K. (2007). Public satisfaction with police: Using procedural justice to
improve police legitimacy. Australian & New Zealand Journal of Criminology, 40, 2742.
Kennedy, D. (1997). Pulling levers: Chronic offenders, high-crime settings, and a theory of
prevention. Valparaiso University Law Review, 31, 449–484.
Kennedy, D. (1998). Pulling levers: Getting deterrence right. National Institute of
Justice Journal. Washington, DC: U.S. Department of Justice.
Kennedy, D., & Braga, A. (1998). Homicide in Minneapolis: Research for problem-solving.
Homicide Studies, 2, 263-290.
Kennedy, D., & Wong, S. (2009). The high point drug market intervention strategy.
Washington, DC: U.S. Department of Justice.
Kleinbaum, D., & Klein, M. (2012). Survival analysis: A self-learning text, Third Edition.
Springer Science + Business Media, DOI 10.1007/978-1-4419-6646-9_8.
Klofas, J., Hipple, N. K., & McGarrell, E. F. (2010). The new criminal justice. New York, NY:
Routledge.
Lanza, S., Moore, J., Butera, N. (2013). Drawing causal inferences using propensity scores: A
practical guide for Community Psychologists. American Journal of Community
Psychology, 52, 380-392.
Lin, G., So, Y., & Johnston, G. (2012). Analyzing survival data with competing risks using SAS
software. Statistics and Data Analysis, SAS Global Forum.

91

Lipsey, M. (1990). Statistical power for experimental research. Newbury Park, CA: Sage
Publications.
Matsueda, R., Kreager,D. & Huizinga, D. (2006). Deterring delinquents: A rational choice
model of theft and violence. American Sociological Review, 71, 95-122.
McGarrell, E., Chermak, S., Wilson, J., & Corsaro, N. (2006). Reducing homicide through a
“lever pulling” strategy. Justice Quarterly, 23, 214-231.
McGarrell, E., Chermak, S., Weiss, A., & Wilson, J. (2001). Reducing firearms violence through
directed police patrol. Criminology and Public Policy, 1, 119-148.
Murphy, K., Hinds, L., & Fleming, J. (2008). Encouraging public cooperation and support for
police. Policing & Society, 18(2), 136-155.
Nagin, D. (1998). Criminal deterrence research at the outset of the twenty-first century. Pp.
1-42 in M. Tonry (ed.), Crime and Justice: A Review of Research, Vol. 23. Chicago:
University of Chicago Press.
Nagin, D. & Paternoster R. (1993). Enduring individual differences and rational choice
theories of crime. Law and Society Review, 27(3), 467-96.
Papachristos, A. V., Meares, T. L., & Fagan, J. (2007). Attention felons: Evaluating project safe
neighborhoods in Chicago. Journal of Empirical Legal Studies, 4, 223-272.
Papachristos, A.W., Wallace, D., Meares, T., & Fagan, J. (2013). Desistance and legitimacy:
The impact of offender notification meetings on recidivism among high risk offenders.
Columbia Law School Public Law & Legal Theory Working Paper Group. Available at:
http://ssm.com/abstract=2240232.
Paternoster, R. (1987). The deterrent effect of the perceived certainty and severity of
punishment: A review of the evidence and issues. Justice Quarterly,4,173-217.
Paternoster, R. (2010). How much do we really know about criminal deterrence? The Journal of
Criminal Law & Criminology,100(3), 765-823.
Paternoster, R. & Bachman, R. (2013). Perceptual deterrence theory. In F. Cullen and P. Wilcox
(Eds.), The Oxford handbook of criminological theory. New York: Oxford University
Press.
Paternoster, R., Brame, R., Mazerolle, P., & Piquero, A. (1998). Using the correct statistical test
for the equality of regression coefficients. Criminology, 36(4), 859-866.
Paternoster, R., L. E. Saltzman, G. P. Waldo, & T. G. Chiricos (1985). Assessments of Risk and
Behavioral Experience: An Exploratory Study of Change. Criminology 23 (3): 417-33.
92

Petersilia, J., & Turner, S. (1993). Intensive probation and parole. In M. Tonry (Ed.), Crime and
Justice: A Review of research, Volume 17. Chicago: University of Chicago Press.
Piquero, A., & Pogarsky, G. (2002). Beyond Stafford and Warr’s reconceptualization of
deterrence: Personal and vicarious experiences, impulsivity, and offending behavior.
Journal of Research in Crime and Delinquency, 39, 153-86.
Pogarsky, G., Piguero, A., & Paternoster, R. (2004). Modeling change in perceptions about
sanction threats: The neglected linkage in deterrence theory. Journal of Quantitative
Criminology, 20(4), 343-369.
Pratt, T.C., F.T. Cullen, K.R. Blevins, L.E. Daigle, & T.D. Madensen. (2006). The empirical
status of deterrence theory: A meta-analysis. In Francis T. Cullen, John Paul Wright,
and Kristie R. Blevins (Eds.), Taking Stock: The Empirical Status of Criminological
Theory—Advances in Criminological Theory, Volume 15. New Brunswick, NJ:
Transaction.
Rivers, L., Norris, A., & McGarrell, E. (2012). Mental model of the drug market intervention.
Journal of Drug Issues, 42(1), 59-81.
Rosenbaum, P., & Rubin, D., (1983). The central role of the propensity score in observational
studies for causal effects. Biometrika, 70(1), 41–55.
Rubin, D., & Thomas, N. (1996). Matching using estimated propensity scores, relating theory to
practice. Biometrics, 52, 249-264.
Sechrist, S., Weil, J., Shelton, T., & Payne, C. (2012). Implementation documentation of the
offender focused deterrence violence initiative (OFDVI) in High Point, NC. University of
North Carolina at Greensboro: Greensboro, NC.
Sherman, L., & Rogan, D. (1995). Effects of gun seizure on gun violence: ‘hot spots’ patrol in
Kansas City. Justice Quarterly, 12(4), 673-693.
Sherman, L., & Weisburd, D. (1995). General deterrent effects of police patrol in crime “hot
spots”: A randomized controlled trial. Justice Quarterly, 12, 625-648.
Singer, S., & Willett, J. (2003). Applied longitudinal data analysis: Modeling change and event
occurrence. New York: Oxford University Press.
Skogan, W. (2006). Asymmetry in the impact of encounters with police. Policing & Society,
16(2), 99-126.
Skogan, W., Hartnett, S., Bump, N., & Dubois, J. (2009). Evaluation of ceasefireChicago. Washington, DC: US Department of Justice.
93

Stafford, M. and M. Warr. 1993. A reconceptualization of general and specific deterrence.
Journal of Research in Crime and Delinquency 30:123-135.
Stuart, E. (2010). Matching methods for causal inference: A review and look forward. Statistical
Science, 25, 1-21.
Sudnow, D. (1965). Normal crimes: Sociological features of the peanal code in a public defender
office. Social Problems, 12(3): 255-276.
Tillyer, M., Engel, R., & Lovins, B. (2010). Beyond Boston: Applying theory to understand and
address sustainability issues in focused deterrence initiatives for violence reduction.
Crime & Delinquency. Online at DOI: 10.1177/0011128710382343.
Tyler, T. (1990). Why people obey the law. New Haven: Yale University Press.
Tyler, T. (2003). Procedural justice, legitimacy, and the effective rule of law. Crime & Justice,
30, 283-357.
Tyler, T. (2006). Psychological perspectives on legitimacy and legitimation. Annual Review of
Psychology, 57, 375-400.
Tyler, T. (2005). Policing in black and white: Ethnic group differences in trust and confidence in
the police. Police Quarterly, 8(3), 322-342.
Tyler, T., Schulhofer, S., & Huq, A. (2010). Legitimacy and deterrence effects in
counterterrorism policing: A study of Muslim Americans. Law & Society Review, 44(2),
365-402.
Weisburd, D. (2008). Place-based policing. Washington: Police Foundation.
Weisburd, D., Morris, N., & Groff, E. (2009). Hot spots of juvenile crime: A longitudinal study
of arrest incidents at street segments in Seattle, Washington. Journal of Quantitative
Criminology, 25, 443-467.
Williams, K. R., & Hawkins, R. (1986). Perceptual research on general deterrence: A critical
review. Law and Society, 20, 545-572.
Witte, K. (1992). Putting the fear back into fear appeals: The Extended Parallel Process Model.
Communication Monographs 59, 329-349.

Witte, K., Meyer, G., & Martell, D. (2001). Effective health risk messages: A step-by-step guide.
Thousand Oaks, CA: Sage.

94