-
UNIVERSITY OF CINCINNATI
Date:
I, ,
hereby submit this original work as part of the requirements for
the degree of:
in
It is entitled:
Student Signature:
This work and its defense approved by:
Committee Chair:
8/18/2010 996
16-Aug-2010
Andrew James Myer
Doctor of Philosophy
Criminal Justice
Hurricane Katrina, Citizen Displacement, and Social Control: A
Test of the
Threat and Benign Neglect Hypotheses and an Investigation of
the
Crime-Arrest Relationship
Mitchell Chamlin, PhDMitchell Chamlin, PhD
Andrew James Myer
-
Hurricane Katrina, Citizen Displacement, and Social Control: A
Test of the Threat and BenignNeglect Hypotheses and an
Investigation of the Crime-Arrest Relationship
A dissertation submitted to the
Graduate School
of the University of Cincinnati
in partial fulfillment of the
requirements for the degree of
Doctor of Philosophy
In the School of Criminal Justiceof the College of Education,
Criminal Justice, and Human Services
by
Andrew J. Myer
B.A., Saint John’s University, 2004M.S., University of
Cincinnati, 2005
Dissertation Committee: Mitchell B. Chamlin, Ph.D.
(Chair)Francis T. Cullen, Ph.D.James Frank, Ph.D.William Alex
Pridemore, Ph.D.
-
ii
ABSTRACT
A guiding theory often used by criminologist to examine the
application of social control
is the conflict perspective. This perspective asserts that
social control is used as a tool to protect
elite segments of society. The conflict theory posits an
increase in social control when elite
groups are threatened, the threat hypothesis. While the benign
neglect hypothesis predicts a
decrease in crime control when the population under threat is a
non-elite population. One of the
main concerns of this paper is to address limitations in the oft
studied threat hypothesis and the
understudied benign neglect hypothesis.
Two limitations of previous studies of conflict hypotheses are
(1) lack of a direct
measure of threat and (2) aggregation bias. This research will
employ a more direct measure of a
threatening event: the displacement of citizens to Houston,
Texas, by Hurricane Katrina, moving
beyond the reliance between the correlation of structural
antecedents and social control. This
research will also investigate aggregation bias by evaluating
conflict hypotheses at a level below
the city—the traditional unit of analysis of past research. To
this end, this research will use time
series analysis to directly account for social control pre- and
post-displacement to determine if
any changes in the administration of social control occurred as
a result of the displacement.
Finally, this paper also attempts to examine the role that
social context plays in the
reciprocal relationship between crimes and arrests. Past
research on the crime arrest relationship
has been mixed. Cross sectional designs often corroborate
predictions of economic theory, while
the more rigorous time series designs fail to provide similar
results. Recent research suggests
that the social context of an area may exert an influence on
this relationship. To this end, this
investigation will employ bivariate time series techniques to
examine the role that social context
plays in the reciprocal relationship between crimes and
arrests.
-
iii
ACKNOWLEDGEMENTS
There are many people who deserve acknowledgement for their
contributions in my
completion of this dissertation and my doctoral degree. Many
professors in the department of
criminal justice influenced me greatly. Dr. Ed Latessa was
instrumental in helping me feed my
family, as well as feed my curiosity about practical
applications in correctional settings. I
learned a great deal from Dr. Smith on teaching and applying
practical applications. I also wish
to thank Dr. Jim Frank for not only serving on my committee, but
also for the great
conversations we had (some while golfing). I am in debt to Dr.
Frank Cullen for showing me
how to frame probing questions in an interesting and academic
fashion, as well as serving on my
committee. I am also grateful for the in-depth and
thought-provoking comments to my
dissertation provided by Dr. Bill Pridemore. I also wish to
thank others for all of their
contributions to my graduation: Dr. John Wooldredge, Dr. John
Eck, Dr. Bonnie Fisher, Dr. Matt
Makarios, Dr. Ben Steiner, John Schwartz, Janice Miller, and
Jean Gary.
Dr. Mitch Chamlin (maybe I will finally be able to call you
Mitch now?) could not have
been a bigger influence on me. I owe a great deal to my success
because of what you taught me,
the conversations we had, and the work we have done together. I
truly cannot express my
gratitude for you taking the time to guide me and not just show
me the way. Thanks for being a
great mentor and a wonderful friend.
My mom and dad also deserve credit for my accomplishment. I
would not be asking all
of these questions if they had not fueled my curiosity and
raised me to not fear asking questions.
Thanks for all you have done for me.
-
iv
To my wife, Leslie: I could not and would not be where I am if
it wasn’t for you. Thanks
for your support, your kindness, and warmth. Thanks for your
generosity and sacrifice. I love
you.
To my son, Benet: I love you. What follows is proof that you can
accomplish anything,
as long as you listen to your mother.
-
v
TABLE OF CONTENTS
Abstract…………..……………………………………………………………………….....
iAcknowledgements…………………………………………………………………………. iiiTable of
Contents………..………………………………………………………………….. ivList of
Tables…………..…………………………………………………………………… vi
CHAPTER ONE: THE ISSUES OF DIRECT MEASUREMENT, AGGREGATION BIAS,
AND SOCIAL CONTEXT UNDER THE CONFLICT PERSPECTIVE
Introduction……………...………………………………………………………………….. 1The Present
Research…………..…………………………………………………………… 7
CHAPTER TWO: A CRITICAL REVIEW OF THE CONFLICT PERSPECTIVE
Introduction………..……………………………………………………………………….. 9The Conflict
Perspective………..………………………………………………………….. 9
The Threat and Benign Neglect Hypotheses…….………………………………………
12Previous Research on the Threat and Benign Neglect
Hypotheses……………………….. 14
Summary..………………………………………………………………………………... 20Methodological
Issues……...………………………………………………………………. 20
Indirect
Measurement….…...............................................................................................
20Aggregation Bias………..……..………………………………………………………… 22
The Current Study…………..……………………………………………………………… 24
CHAPTER 3: SOCIAL CONTEXT AND THE CRIME-ARREST RELATIONSHIP:
EVALUATING THE PREDICTIONS OF THE CONFLICT PERSPECTIVE
Introduction……………………………………………………………………………….. 26The Causal
Mechanisms Underlying the Crime-Arrest Relationship…………………….. 27A
Review of the Crime-Arrest Literature……………….………………………………….
31Research on the Tipping Effect……………….…………………………………………… 37The
Current Study………………..………………………………………………………… 38
CHAPTER 4: METHODS
Introduction…………..…………………………………………………………………….. 40Research
Design………………..…………………………………………………………... 40
Hurricane Katrina as Threat……………..……………………………………………….
42Data…………..……………………………………………………………………………... 44
Crimes………..…………………………………………………………………………...
45Arrests…………..………………………………………………………………………... 47Police
Divisions………..………………………………………………………………… 51
Analytical Method…………..……………………………………………………………… 58Univariate
Estimation………..…………………………………………………………... 59Interrupted Time
Series……………………………………..…………………………… 62
-
vi
Bivariate ARIMA Analysis.……………………..……………………………………….
65Conclusion………………………..…………..…………………………………………….. 66
CHAPTER 5: RESULTS OF TIME SERIES ANALYSES OF CLEARANCES AND
CRIMES
Introduction…………..…………………………………………………………………….. 67Interrupted
Time Series Results………………….……………………………………...
67Summary…………….…………………………………………………………………... 88
Bivariate Analyses Results………………..………………………………………………...
89Conclusion……………..…………………………………………………………………… 111
CHAPTER 6: DISCUSSION AND SUMMARY OF FINDINGS
Introduction…………………………………………………………………………………. 116Interrupted Time
Series Discussion………………………………………………………… 116Bivariate Analyses
Discussion……………………………………………………………… 125Recommendations and
Limitations………………………………………………………….
128Conclusions…………………………………………………………………………………. 130
References…………………………………………………………………………………... 132
-
vii
List of Tables
Table 4.1: Crimes and Clearances by Year for the City of
Houston, 1997-2008…..…..... 49Table 4.2: Crimes and Clearances by
Police Division for the City of Houston, 1997-2008. 53Table 5.1:
Final Univariate and Intervention Models for the City…………………………
68Table 5.2: Social Context of Westside Police Divisions in Houston
Using Non-proportional Census Distribution……………………………………………………………
71Table 5.3: Final Univariate Models for Westside Police
Divisions……………………….. 72Table 5.4: Zero-Order Intervention Models
for Westside Police Divisions………………... 73Table 5.5: Final
Univariate Models for Crime in Westside Police Divisions………………
76Table 5.6: Zero-Order Intervention Models on Crime for Westside
Police Divisions…….. 77Table 5.7: Final Univariate Clearance Models
for Eastside Police Divisions……………… 79Table 5.8: Social Context of
Eastside Police Divisions in Houston Using Non-proportional Census
Distribution…………………………………………………………………………. 81Table 5.9: Zero-Order
Intervention Clearance Models for Eastside Police Divisions……..
82Table 5.10: Final Univariate Crime Models for Eastside Police
Divisions………………… 85Table 5.11: Zero-Order Intervention Crime Models
for Eastside Police Divisions……….. 87Table 5.12: Final Univariate
Models for Full, Pre-, and Post-displacement Series……….. 91Table
5.13: Estimated CCFs for the Full Series for All
Aggregates……………………….. 96Table 5.14: Contextual Transfer Function
Models, Full Series……………………………. 101Table 5.15: Estimated CCFs for
the Pre-displacement Series for All Aggregates…………. 103Table 5.16:
Contextual Transfer Function Models, Pre-displacement Series……………….
106Table 5.17: Estimated CCFs for the Post-displacement Series for
All Aggregates……….. 107Table 5.18: Contextual Transfer Function
Models, Post-displacement Series……………... 111Table 6.1: Social
Context of Westside Police Divisions in Houston Using
Non-proportional Census Distribution…………………………………………………………… 122
-
1
CHAPTER ONE
THE ISSUES OF DIRECT MEASUREMENT, AGGREGATION BIAS, AND SOCIAL
CONTEXT UNDER THE CONFLICT PERSPECTIVE
INTRODUCTION
The focus of the current research is to determine the indirect
effects of Hurricane Katrina
on crime, arrests, and the relationship between crime and
arrests. More specifically, this research
assesses the impact of the displacement of citizens to Houston,
Texas on crime and arrests in
Houston, a city not directly exposed to the natural disaster.
This study occurs under the guidance
of the conflict perspective and addresses the issues of: (1)
direct measurement of the conflict
perspective’s notion of threat, (2) the identification and
selection of an appropriate unit of
analysis, and (3) consideration of the social contexts when
examining crimes, arrests, and the
relationship between the two. The remainder of this chapter
outlines the importance of
addressing these issues when examining the conflict
perspective.
The conflict perspective states that crime control is a tool
used to protect the interests of
elite populations (Black 1976; Blalock 1967; Turk 1969;
Chambliss and Seidman 1971; Spitzer
1975; Quinney 1970, 1977; Jackson 1989; Liska 1992). Research on
crime control under this
perspective generally focuses on two hypotheses. The first is
known as the threat hypothesis.
This hypothesis states that symbolic or physical threats to
elite populations will lead to an
increase in crime control (Liska 1992). This increase in crime
control is to protect the position of
elites in society. The second hypothesis is known as the benign
neglect hypothesis. The focus of
this hypothesis is on symbolic or physical threats to non-elite
populations. When threat is
experienced by a non-elite population, the benign neglect
hypothesis predicts a decrease in crime
control (Liska 1992). This occurs because the non-elite
population lacks the ability to legitimize
-
2
their complaints to police, or because the acts are committed
intra-group and are viewed as a
personal matter deemed unnecessary for formal intervention
(Liska and Chamlin 1984). In
general, past research finds support for the threat hypothesis
(Brown and Warner 1992;
Carmichael 2005; Corzine, Creech, and Corzine 1983; Eitle,
D’Alessio, and Stolzenberg 2002;
Greenberg, Kessler, and Loftin 1985; Holmes 2000; Jackson and
Carroll 1981; Jacobs 1979;
Jacobs and Britt 1979, Kane 2003; King and Wheelock 2007; Liska,
Lawrence, and Benson
1981; Liska, Chamlin, and Reed 1985; Nalla, Lynch, and Leiber
1997; Ruddell and Urbina 2004;
Smith and Holmes 2003; Stults and Baumer 2007), but research on
the benign neglect
hypothesis, having far fewer tests, receives mixed support
(Chamlin and Liska 1992; Liska and
Chamlin 1984; Eitle, Stolzenberg, and D’Allesio 2005; Myer and
Chamlin 2007; Parker et al.
2005; Stolzenberg et al. 2004). Two topics that are common to
the testing of both the threat
hypothesis and the benign neglect hypothesis are (1) the
measurement of the intervening variable
(i.e., the concept of threat), and (2) the unit of analysis used
to study the threat and benign
neglect hypothesis (i.e., the city level of analysis).
Previous research on the conflict perspective typically
conceptualizes threat as the size of
minority groups. The size of a non-elite population (e.g.,
percent black) is thought to be
indicative of perceived threat, because an increase of this
subordinate population poses a threat
to the economic and political position of elite populations. As
such, previous research has
assessed the relationship between the size of minority groups
and the size of police institutions
(Chamlin, 1989; Greenburg, Kessler, and Loftin 1985; Jackson
1986, 1989; Jackson and Carroll
1981; Jacobs 1979; Liska, Lawrence, and Benson 1981; Nalla,
Lynch, and Leiber 1997; Stults
and Baumer 2007) or the activity of police institutions (Brown
and Warner 1992; Chamlin and
Liska 1992; Eitle, D’Alessio, and Stolzenberg 2002; Liska and
Chamlin 1984; Liska, Chamlin,
-
3
and Reed 1985; Parker, Stults, and Rice 2005; Stolzenburg,
D’Alessio, and Eitle 2004). The
emergence of a significant relationship between the size of a
minority group and formal social
control is often used to infer perceived threat by elites. The
consequence of this approach is that
the intervening variable of threat is not directly measured
(Chamlin 1989; Chamlin 2009, Eitle et
al. 2002; King and Wheelock 2007; Jacobs 1979; Parker, Stults,
and Rice 2005). By not directly
measuring the theoretical construct of threat, the statistical
association between minority group
size and some outcome (e.g., crime control) is susceptible to
multiple interpretations. For
example, the positive statistical relationship between minority
size and arrests can also be
thought of as a challenge to the social solidarity of the social
system, a structural functionalist
perspective (for example, see Inverarity 1976). It is the
intention of the present research is to
directly account for threat, thus eliminating multiple
theoretical interpretations.
In addition to the measurement of threat, this research also
addresses the issue of
aggregation bias. The examination of data at an improperly
specified unit of analysis may lead
to incorrect or null results. To avoid this, research must
confront a theory with data collected at
the appropriate unit of analysis in order to falsify it (Bailey
1984; Greenberg, Kessler, and Logan
1981; Land, McCall, and Cohen 1990; Lieberson 1985). For
example, Bailey (1984) states that
cities are more appropriate than SMSAs when studying the
poverty-crime relationship.
Greenberg and colleagues (1981) demonstrate that state-level
evaluations of general deterrence
theory have little probative value because the causal mechanism
operates at a smaller level of
aggregation.
Aggregation bias is relevant to the present research because
previous research has
evaluated threat and benign neglect across municipalities, when
in fact the mechanisms
underlying the threat and benign neglect hypotheses may occur at
a level below the city (Myer
-
4
and Chamlin 2007, Stolzenberg, D’Allesio, and Eitle 2004).
Recall that the benign neglect
hypothesis predicts a decrease in crime control because of an
inability of non-elite populations to
legitimize their complaints or because of an increase in
intragroup crime. This is significant to
the conflict perspective because previous research has also
evaluated the relationship between
the distribution of minority groups (via a measure of
segregation at the city level) and crime
control (Brown and Warner 1992; Carmichael 2005; Chamlin and
Liska 1992; Eitle et al. 2005;
Liska and Chamlin 1984; Liska, Lawrence, and Benson 1981; Liska
et al. 1985; Parker et al.
2005; Parker, MacDonald, Jennings, and Alpert 2005; Stolzenberg
et al. 2004).
The segregation of less powerful populations increases the
likelihood of intraracial crime
on populations that already have limited sociopolitical power to
legitimize their complaints to
police (Blauner 1972; Spitzer 1975). Past research typically
measures segregation at the city
level through the use of the dissimilarity index, which compares
the racial composition of city
blocks to the racial composition of the city as a whole. This is
done for all blocks in order to
measure the percentage of one racial group that would have to
move to a different neighborhood
to be distributed similarly as another racial group (See Massey
and Denton, 1988). The resulting
index score provides a measure of segregation for the entire
city. If results produce a negative
relationship between segregation and social control, support for
benign neglect is revealed. The
issue paramount to this approach is that the city level, the
typical level of aggregation when
examining the benign neglect (and threat) hypothesis, may be
masking observable effects that
occur at a level beneath the city. As an example, imagine a
sample of cities, each containing
only one or two neighborhoods that are segregated. If such were
the case, there is likely to be
little variation in the level of segregation across the sample
of cities. Necessarily, measures of
segregation would have little, if any, impact on crime control
outcomes. The effects of
-
5
segregation that operate across neighborhoods within cities are
then being masked by the use of
city-level data (Blauner 1972; Spitzer 1975). If such were the
case, previous research has
neglected to use the appropriate unit of analysis. An objective
of this study is to examine
whether the predictions of the threat and benign neglect
hypotheses are evidenced at sub-city
level by taking into account contextual characteristics of
neighborhoods within the city.
Accounting for the social context is important because previous
research has demonstrated that
social control can be a function of the social context within
which police officers operate
(Klinger 1997; Klinger 2004; Smith 1986; Smith, Visher, and
Davidson 1984; Terrill and Reisig
2003).
The final intent of this research is to examine the relationship
between crime and arrests.
While theory posits a relationship between crimes and arrests
(see Becker 1968; Gibbs 1975;
Tittle 1980), different hypotheses predict different
interactions between the two variables. The
overload hypothesis predicts that increases in crime will lead
to a decrease in arrests because the
resources of the police are taxed (Fisher and Nagin 1978;
Geerken and Gove 1977), whereas the
public choice hypothesis predicts that an increase in crimes
will lead to an increase in arrests
because there is more pressure put on police to address the
up-rise in crime (Borcherding and
Deacon 1972; Deacon 1978; Tridimas 2001). Alternatively, the
deterrence hypothesis predicts
that an increase in arrests will lead to a decrease in crime
because the costs of committing crimes
exceed the benefits (Becker 1968; Geerken and Gove 1977; Gibbs
1975).
Initial tests of the above hypotheses were undertaken using
cross sectional designs.
While many of these tests produced significant findings, there
was growing concern about the
limitations of cross sectional designs in the evaluation of the
reciprocal relationship between
crimes and arrests. Accordingly, researchers began evaluating
the predictions of these
-
6
hypotheses through time series analyses in an attempt to examine
the temporal order and lag
structure of the relationship between crimes and arrests—both
limitations of cross-sectional
designs. Generally, time series analyses did not replicate the
findings of cross sectional designs
(Chamlin 1988; Greenberg et al. 1979; Greenberg and Kessler
1982; Logan 1975). This has led
some to argue that findings from the cross-sectional analyses
are corrupted by model-
misspecification error (Greenberg et al. 1979). Recently, it has
been argued that the inability of
time-series analyses to replicate findings from cross-sectional
analyses on the crime-arrest
relationship may in fact be due to the inability to account for
the social contexts of cities
(Chamlin and Myer 2009). This research will build off this new
line of research to determine if
social context does in fact play a role in the reciprocal
relationship between crimes and arrests.
The threat hypothesis and the benign neglect hypothesis vary in
their predictions of crime
control based on the status of the group experiencing threat,
but the effects of their predictions
may not be mutually exclusive. As such, the predictions of these
hypotheses may have utility in
the investigation of the reciprocal relationship between crimes
and arrests. If there is an increase
in arrests to protect the interests of social elites (the threat
hypothesis), it would be expected to
simultaneously reduce law enforcement in areas dominated by
non-elites. This is then likely to
generate a reduction in the amount of offenses cleared by
arrest. Accordingly, the benign neglect
hypothesis predicts that after the experience of symbolic or
physical threat, the level of crime
will begin to negatively affect the level of arrests in
predominately black (non-elite) areas of the
city. For the purposes of the current research, this means that
changes in the level of arrest will
affect changes in the level of crimes post-displacement, but not
before. These changes will also
be a function of the social contexts of Houston, Texas.
-
7
THE PRESENT RESEARCH
The intent of this study is to determine the indirect effects of
Hurricane Katrina on crime,
arrests, and the relationship between crimes and arrests. The
current research utilizes a quasi-
experimental, time series design to account for the occurrence
of Hurricane Katrina and the
resulting displacement of citizens to Houston. Previous research
has evaluated the effects of
disasters such as hurricanes (Brezina and Kaufman 2008; Davila,
Marquart, and Mullings 2005;
Faust and Kauzlarich 2008; Kirk 2009; Kulkarni, Bell,
Beausoleil, Lein, Angel, and Mason
2008; Lanza-Kaduce, Dunham, Akers, and Cromwell 1998; LeBeau
2002; Rojek and Smith
2007; Thomas 2008), riots (Bergesen and Herman 1998; Chamlin
2009; Chamlin and Cochran
2000; Chamlin and Myer 2009; King and Waddington 2004; Myer and
Chamlin 2007),
earthquakes (Green 2005), maritime disasters (Rothe, Muzzatti,
and Mullins 2006), and terrorism
(Bornstein 2005; Frank, Dewart, Schmeidler, and Demirjian 2006;
Pandiani, Knisley, Banks,
Simon, and Blackburn 2005; Pridemore, Chamlin, and Trahan 2008)
on various outcomes. Most
of the research on Hurricane Katrina has evaluated the direct
effects on New Orleans; however,
the indirect effects of Katrina also merit inspection. Assessing
the effects of Hurricane Katrina
on Houston, Texas is relevant because of the amount of citizens
that were affected, directly
(displaced New Orleans residents) and indirectly (Houston
residents)—Houston eventually
housed approximately 250,000 evacuees (Berger 2006). The
displacement of citizens to
Houston, Texas constitutes a threatening event for both the
displaced citizens of Louisiana, as
well as the residents of Houston, due to the abrupt nature of
the event.
This research will assess the predictions of the threat
hypothesis and the benign neglect
hypothesis by providing a more direct measure of threat, the
displacement of citizens. The
predictions of these two hypotheses will also be assessed at a
level below the city. Houston
-
8
Police divisions and their social contexts will be examined.
Examination of social context is
important because it can influence the administration of social
control (Klinger 1997; Klinger
2004; Smith 1986; Smith, Visher, and Davidson 1984; Terrill and
Reisig 2003). Finally, this
paper will also explore the predictions of these two hypotheses
in the context of the crime-arrest
literature to determine if the reciprocal relationship between
crimes and arrests is affected in
different social contexts of the city, post-displacement.
To this end, a number of objectives must be accomplished. A
review of the research on
the conflict perspective’s threat and benign neglect hypotheses
will be presented in Chapter 2.
This discussion will highlight how previous research measures
threat and segregation and the
level at which these analyses have been conducted. The
crime-arrest relationship literature will
be discussed in Chapter 3. Because one of the aims of the
present research is to examine if the
relationship between crimes and arrests was influenced by the
displacement of citizens to
Houston, it is important to review how previous research
attempts to examine the association
between crimes and arrests. Chapter 4 provides the research
strategy and methodology used to
examine the threat hypothesis, the benign neglect hypothesis,
and the relationship between
crimes and arrests. The results of the time series analyses on
crimes, arrests, and the relationship
between the two will be presented in Chapter 5. Finally, Chapter
6 provides a discussion on the
results, as well as the implications of the findings.
-
9
CHAPTER TWO
A CRITICAL REVIEW OF THE CONFLICT PERSPECTIVE
INTRODUCTION
The conflict perspective has a rich tradition in criminological
research. In order to better
understand the foundations of the current research, this chapter
offers a discussion on the origins
of the criminological conflict perspective’s views on crime and
its predictions on the
administration of social control. This includes a review of the
threat hypothesis and the benign
neglect hypothesis. A second priority of this chapter is to
present past research on the threat and
benign neglect hypothesis. For each study reviewed, the
measurement of threat and unit of
analysis will be offered. Third, because it is the intent of the
current research to provide a more
direct measure of threat and identify the proper unit of
analysis to inspect the two hypotheses, a
brief review of the methodological concerns of indirect
measurement and aggregation bias will
be provided. Finally, an outline of how the present research
will address the issues of threat and
aggregation bias will be offered.
THE CONFLICT PERSPECTIVE
The conflict perspective stems from the early works of the
sociologists Karl Marx and
Friederich Engels. Their writings focused on the conflict
between the working class and the
property owners. In short, they argue that it is in the best
interest of property owners to maintain
their elite status through the exploitation of the working class
and by determining the ruling ideas
of society. While much of the writings of Marx nor Engles did
not place an emphasis on crime
and the criminal justice system, modern day scholars have
produced a theoretical work on crime
and the criminal justice system based on Marx and Engle’s
scholarship. This line of reasoning is
the focus of this research.
-
10
Building on the thinking of Marx and Engles, the conflict
perspective takes a unique
perspective on crime in the sense that it does not view crime as
an inherent condition. For
example, other theories explain crime in terms of an
individual’s attachment to institutions of
society (Hirschi 1969), the level of personal self-control
(Gottfredson and Hirschi 1990), the
ability to learn crime (Burgess and Akers 1966), or the amount
of strain an individual has
(Agnew 1985, 1992). Instead, the conflict perspective asserts
that crime results from the conflict
between powerful, elite groups of society and less powerful,
non-elite groups of society. It is
this conflict between the groups that produces crime.
One of the first criminologists to build off of Marx and Engles
was Willem Bonger
(1916). Bonger follows the original underpinnings of the
conflict perspective and argues that a
consequence of capitalism is crime (see also Currie 1997). More
specifically, Bonger argues that
capitalism stresses focus on individual level success, what
Bonger labels “egoism.” Produced by
capitalism, egoism stresses individual profits through the
exploitation of others at the benefit of
individual success and the exclusion of the larger social good.
This leads to an unequal
distribution of resources and moves the cultural towards
individualism. Crime emerges from the
increase in self-interest, as the powerless commit crimes out of
a sense of need or injustice.
Turk (1969) agrees that the ongoing struggle between elite,
powerful groups and non-
elite, powerless groups produces crime. But crime is not
necessarily the result of differences in
economic standing so much as it is the social construction of
law. According to Turk (1969),
perceptions of group behavior are relative—the behavior of any
one group is going to be
perceived as negative by the other group. Turk argues that elite
groups of society label the
behavior of non-elite groups of society as threatening and
devote resources to curtail these
“threatening” behaviors. The labeling of non-elite’s behaviors
as threatening allows elite groups
-
11
to protect their development of the cultural order and maintain
their status and control as an elite
group (Chambliss 1975; Chambliss and Seidman 1971). This idea
was further embraced by
Richard Quinney (1969; 1970), who argues that social segments of
society conceptualize and
develop their own normative systems that define acceptable
behavior. Each segment of society
then views their own system’s definitions as normal and the
other’s as abnormal. These clashing
definitions lead to crime, with the end result being that the
powerless have their definitions
labeled as criminal.
Imbedded in the above discussion is the notion that laws are
created to protect the
interests of the powerful. In order to protect interests, laws
are created that function as a
mechanism to control the powerless. Thus, a main feature of the
conflict perspective is
explaining the administration of social control. It is also this
aspect of the conflict perspective
emphasizes the notion of threat. One of the leading discussions
of threat comes from the work of
Hubert Blalock.
Blalock (1967) argues that increases in social control will
result when the political and
economic status of elite groups are threatened by non-elite
groups. Up to this point, there had
been no clear articulation of what constitutes elite and
non-elite groups. Addressing this, Blalock
couches his theory in terms of minority-group relations.
Succinctly stated, Blalock argues that a
nonlinear relationship between discrimination and the size of
minority groups is a result of threat
felt by non-minority groups (pp. 144-154). Accordingly, when
increases in minority groups
threaten the socioeconomic position, competition, and power of
non-minority groups, increases
in discrimination (or motivation to discriminate) will result
(see also Black 1976).
The idea that laws are created to benefit some and not all has
been amalgamated with the
theoretical underpinnings of Blalock’s theory on minority-group
relations. This marriage of
-
12
ideas has led to the conflict perspective’s rationale on the
administration of social control. One
of the first attempts to apply Blalock's theory of majority
group discrimination to the relationship
between minority group size and macro social control is Jackson
and Carroll’s (1981)
examination of variations in policing resources among 90
non-southern cities. Drawing on
Blalock’s power threat hypothesis, they contend that as minority
group size increases, so too
does the level of policing resources. This relationship holds
until a point at which racial
minorities achieve a numerical superiority (around 50%) and
thereby begin to resist the use of
policing resources to suppress them. The current body of
research uses two main hypotheses to
examine social control as a function of racial tensions. These
two hypotheses are known as the
threat hypothesis and the benign neglect hypothesis.
The Threat and Benign Neglect Hypotheses
Both the threat and the benign neglect hypotheses make
predictions on the use of social
control based on group exposure to threat. Key to both the
threat and the benign neglect
hypotheses is the concept of threat. The difference between the
two hypotheses is the group
(elite or non-elite) who experiences the threat.
The threat hypothesis predicts an increase in social control
when elite groups experience
threat to their social status. In order to protect their social
standing and position in society, elite
groups utilize the formal mechanisms of crime control to protect
their interests from groups that
are viewed as threatening their status quo. The ancillary benign
neglect hypothesis predicts a
decrease in crime control. This is because the groups
experiencing the threat are powerless, non-
elite groups. There are two mechanisms underlying the benign
neglect hypothesis (Liska and
Chamlin, 1984). First, non-elite groups may not be able to
legitimize their complaints because of
-
13
their lack of socio-political power. Second, the threat may lead
to increases in intra-group crime
which may be viewed by the police as a personal matter
undeserving of formal intervention.
While there remains debate on the constitution of elite groups
(Bierne 1979; Liska 1992;
Spitzer 1975), the operationalization of key variables involved
in the testing of the threat and
benign neglect hypotheses has been established in the
literature. Namely, tests of the threat
hypothesis look at the relationship between the size of minority
groups and crime control, while
tests of the benign neglect hypothesis evaluate the relationship
between the distribution of
minority groups and crime control. Past tests of the threat and
benign neglect hypotheses have
evaluated the relationship between perceived threat and the size
of police institutions (Chamlin,
1989a; Greenburg, Kessler, and Loftin 1985; Jackson 1986, 1989;
Jackson and Carroll 1981;
Jacobs 1979; Liska, Lawrence, and Benson 1981; Nalla, Lynch, and
Leiber 1997; Stults and
Baumer 2007), police use of force (Chamlin 1989b; Holmes 2000;
Liska and Yu 1992; Parker,
MacDonald, Jennings, and Alpert 2005; Smith and Holmes 2003;),
incarceration (Carmichael,
2005; Myers 1990; Tittle and Curran 1988), executions and
lynchings (Phillips 1986; Tolnay,
Beck, and Massey 1989, 1992) or the use of arrests by police
(Brown and Warner 1992;
Chamlin 2009, Chamlin and Liska 1992; D’Alessio, Stolzenberg,
and Eitle 2002; Eitle,
D’Alessio, and Stolzenberg 2002; Eitle, Stolzenberg, and
D’Alessio 2005; Liska and Chamlin
1984; Liska, Chamlin, and Reed 1985; Parker, Stults, and Rice
2005; Stolzenburg, D’Alessio,
and Eitle 2004). The focus of the present research is on the
relationship between the use of arrest
as social control. As such, the following review of the
literature on the threat and benign neglect
hypotheses will focus on the related research on the
relationship between threat and police
aggressiveness. Special attention will be paid to the
measurement of threat, segregation, and the
unit of analysis at which a study is conducted.
-
14
PREVIOUS RESEARCH ON THE THREAT AND BENIGN NEGLECT
HYPOTHESES
Williams and Drake (1980) produced the first study to evaluate
the relationship between
threat and police use of arrest. Investigating a sample of SMSAs
with a population above
500,000 in 1970, Williams and Drake inspect the role of economic
inequality, percent black,
percent unemployed, population size, and victimization rate on
arrest rates for each of the Type I
index crimes. Results showed no relationship between the
structural variables—including their
measure of percent black—and any of the seven index crimes.
One of the drawbacks of Williams and Drake’s (1980) research was
the inability to
control for the crime rate. To address this, Liska and Chamlin
evaluated a sample of 76 cities
with a population over 100,000. The authors used regression
analyses and controlled for the
crime rate, and also the city population, police force size,
percent poor, income inequality,
segregation, and percent nonwhite. Liska and Chamlin also took
the unique step of disaggregated
property and personal arrest rates by race. This allowed for the
creation of race specific models
of arrests. Analyses for the aggregate models (i.e., models that
did not disaggregate arrests by
race) demonstrate a positive, significant relationship between
percent nonwhite and personal and
property arrests. Arrests were then disaggregated by race,
revealing no significant effect of
percent nonwhite and white personal or property arrests. This is
contrary to the predictions of
the threat hypothesis. However, percent non-white was
significantly and negatively related to
black arrest rates for both personal and property crimes, being
the least important predictor in the
white arrest model and the most important predictor in nonwhite
arrest model. Inspection of the
role of segregation revealed a significant, negative
relationship to arrest rates for both white and
nonwhite arrest rate models. The results of Liska and Chamlin’s
(1984) analyses provide
support for the benign neglect hypothesis, but not the threat
hypothesis.
-
15
Liska, Chamlin, and Reed (1985) pit the predictions of the
conflict perspective against
the predictions of an economic approach in explaining the
certainty of punishment. To do this,
Liska and colleagues evaluate the clearance rates (ratio of
arrests to crimes) for 77 cities with a
population over 100,000. Regression analysis is used to control
for the effects of workload,
resources, population, percent poor, income inequality,
segregation, and percent nonwhite on
each of the seven index crimes (i.e., homicide, rape, assault,
robbery, burglary, larceny, and auto
theft). Results of the analyses indicate that after controlling
for economic and conflict
exogenous variables, percent nonwhite—the indicator of
threat—was positively and significantly
related to the certainty of arrest for all index crimes but
assault, lending support to the threat
hypothesis. Moreover, segregation was significantly and
negatively related to arrests for all
crimes except homicide, supporting the benign neglect
hypothesis.
To account for the possibility of period effects of their
earlier research (Liska and
Chamlin 1984), Chamlin and Liska (1992) examine similar data 10
years later. As such,
regression analyses were conducted on race specific arrest rates
for property and personal crimes
for 97 cities. Analyses of the effects of percent nonwhite on
white personal and property arrests
rates again revealed no support for the threat hypothesis. In
fact, for property arrests, whites
were now significantly less likely to be arrested as the
percentage of nonwhites in a city
increased. Percent nonwhite was not significantly related to the
arrest rate for white personal
crimes. The relationship between segregation and arrest rates
for both the white and nonwhite
arrest models was non-significant, unlike in previous findings.
In short, this analysis now
revealed no support for benign neglect.
Brown and Warner (1992) argue that the previous tests of the
threat hypothesis may be
subject to a historical effect—the turbulent social unrest of
the 1960s. More specifically, Brown
-
16
and Warner argue that the reliance of the measure of percent
minority may be affected by the
social movements of the period under investigation. To determine
if the findings are relevant to
a different historical time period, Brown and Warner examine
data from the 50 largest cities in
the United States in 1900. Rather than employ percent minority
as their measure of threat, the
authors institute a measure percent foreign born to capture
threat felt by native born Americans
in predicting arrests for drunkeness. Results demonstrated that
percent foreign born was a
significant, positive predictor of arrests for drunkenness
despite controls for alcohol consumption
and police force size. In later models, this significant effect
was not observed, as it appears to be
mediated by the political and social structure of a city. With
regards to segregation, there was no
support for the benign neglect hypothesis as this variable
demonstrated no statistically significant
relationship to arrests.
Eitle, D’Alessio, and Stolzenberg (2002) take issue with the
operationalization of threat
as the relative size of minority population. To address this,
the authors utilize three distinct
measures of threat: (1) the ratio of black-to-white voting; (2)
the ratio of black-to-white
unemployment; and (3) the black-on-white crime rate. Eitle and
colleagues use NIBRS data
from 46 counties in South Carolina to estimate pooled
cross-section time series equations.
Analysis on the effects of black-on-white crime on black arrest
rates revealed a positive,
significant relationship. Specifically, as interracial crime
increases, the probability of a black
individual being arrested also increases, which is consistent
with the threat hypothesis.
However, there was no statistical relationship between the other
two measures of threat, the ratio
of black-to-white voting or the ratio of black-to-white
unemployment, and arrest rates. Eitle and
colleagues did not include a measure of segregation in any of
their analyses.
-
17
In a subsequent study, D’Alessio, Stolzenberg, and Eitle (2002)
evaluate the effects of
both the ratio of black-to-white voting and the ratio of
black-to-white unemployment, as well as
the effect of percent nonwhite, on four different dependent
variables. These four endogenous
variables are: (1) white-on-black crime rate; (2) black-on-white
crime rate; (3) white-on-white
crime rate; and (4) black-on-black crime rate. Again, data were
from the NIBRS on 46 counties
in South Carolina. Analyses from the pooled cross-section time
series equations revealed no
significant effects for the percent nonwhite or the ratio of
black-to-white voting across any of the
four dependent variables. The ratio of black-to-white
unemployment was a significant, negative
relationship with white-on-black crime. This variable has no
significant impact on any of the
other victim-offender dyads. Segregation was not included as a
variable in any of the models.
Kane (2003) explores a different unit of analysis than past
studies. Specifically, Kane
argues that threat hypothesis research is better suited for a
smaller unit of analysis. Kane
suggests that police districts are the most appropriate units
because this aggregate level is where
police develop work groups and norms of behavior. To this end,
Kane hypothesizes that
increases in nonwhite population, disaggregated into percent
black and percent Latino, will lead
to increases in police deployment. A panel design is used on
three years of data from the New
York City Police Department. Results from the analyses showed no
relationship between police
deployment and percent black. However, once the Latino
population of a police district reached
23 percent, there was a significant, positive relationship with
police deployment. Thus, Kane’s
police district analysis provided mixed support for the threat
hypothesis.
While Kane (2002) did not include a specific measure of
segregation in his models, his
data were able to capture the percent black and percent Latino
of each district under
investigation. In this sense, he did succeed in capturing the
relative segregation of each police
-
18
district. Accordingly, Kane found no support for the benign
neglect hypothesis. Increases in
black population (i.e., increases in district segregation) were
unrelated to arrests, and increases in
Latino population in each district (i.e., increases in district
segregation) were in the opposite
direction predicted by the benign neglect hypothesis.
Stolzenburg, D’Alessio, and Eitle (2004) combine microlevel
NIBRS data with macro
level census data to develop a multilevel test of the threat
hypothesis. In short, Stolzenberg and
colleagues wanted to control for individual factors (e.g., race)
as well as structural factors (e.g.,
segregation) to determine the important influences on the
probability of arrest for a violent
crime. The results of the multilevel analysis failed to support
the threat hypothesis. Instead, the
authors found that the percent population black was
significantly and negatively related to the
probability of arrest. That is, as a city’s black population
increased, the likelihood of arrests
decreased, which is not predicted by the threat hypothesis.
While this finding is in disagreement
with the threat hypothesis, it appears to be consistent with the
benign neglect hypothesis.
However, further analysis revealed that percent black was not
significantly related to the
likelihood that black offenders will be arrested or the
probability that intraracial crimes will
result in an arrest. Analyses of segregation revealed that there
was a significant and negative
relationship between a city’s level of segregation and (1) the
overall probability of arrest (black
or white offender), (2) the likelihood of black offenders being
arrested, and (3) the probability
that intraracial crimes (black-on-white) resulted in an arrest.
Taken as a whole, these findings
provided no support for the threat hypothesis, and mixed support
for the benign neglect
hypothesis.
Parker, Stults, and Rice (2005) offer different
operationalizations of threat by
incorporating three different measures of threat in their
analyses. Specifically, Parker and
-
19
colleagues measure threat through the traditional percent black
population, but also include the
measures of percent black immigration and a racial inequality
index. The racial inequality index
was developed using principal components factor analysis with
the following factors: percent
persons living below poverty, income inequality, percent males
jobless aged 16 and older,
percent children not living with both parents, white-black
unemployment ratio, white-black
bachelor degree ratio, and white-black high school diploma
ratio. Data were from 245 cities
with a population over 100,000 in 2000. The authors estimated
separate models for arrests
disaggregated by race. Results demonstrated no relationship for
the racial inequality index for
either white or black arrests. Moreover, the measures of percent
black and percent black
immigration were not related to white arrests, but were
significantly and negatively related to
black arrests. Specifically, increases in the size of the black
population and increases in the
number of black immigrants were associated with a decrease in
the number of black arrests.
These findings are inconsistent with the threat hypothesis, but
consistent with the benign neglect
hypothesis. Nevertheless, analysis of segregation produced no
statistical relationship with
arrests, leading to mixed findings regarding the benign neglect
hypothesis.
Eitle, Stolzenberg, and D’Alessio (2005) employed hierarchal
linear modeling to examine
the impact of individual, city, and police organization
variables on probabilities of arrest for
aggravated assaults and simple assault separately in 105 small
cities. There was no relationship
between a city’s percent black population and simple or
aggravated assaults. Thus, Eitle and
colleagues did not find support for the threat hypothesis.
However, with regards to the benign
neglect hypothesis, there was a significant negative
relationship between a city’s level of
segregation and probability of arrest for aggravated assault and
simple assault. In fact, the
-
20
authors noted that the difference between arrest probabilities
for whites and blacks was relatively
large. This finding supports the predictions of benign
neglect.
Chamlin (2009) also attempts to provide a more direct measure of
threat. Rather than
rely on a compositional measure of minority population, Chamlin
uses the occurrence of a race
riot as a measure threat. Using time series data on robbery
arrests in the city, Chamlin finds an
increase in robbery arrests directly after the threatening event
of the race riot, as predicted by the
threat hypothesis. To control for the possibility of a history
effect, similar data were examined
for the cities of Dayton and Columbus, Ohio, where no riots
occurred, and no changes were
observed in the level of robbery arrests. There were no controls
for segregation in Chamlin’s
analysis.
Summary
The above discussions have demonstrated that results of tests of
the threat hypothesis and
benign neglect hypothesis are mixed. Two themes emerged in the
above discussion: (1) tests of
the threat hypothesis either lacked direct measures of threat or
the exploration of more direct
measures was a focal concern; and (2) the vast majority of tests
were conducted at the city level
of analysis. Because findings for both hypotheses were clearly
mixed, this dissertation will
address these two themes. As such, the next section provides a
discussion surrounding the
concerns of the two methodological issues: indirect measurement
and aggregation bias.
METHODOLOGICAL ISSUES
Indirect Measurement: The Threat Hypothesis and the Size of
Minority Groups
The size of minority group is the most frequently used
operationalization of threat. The
use of this measure stems from Blalock’s (1967) assertion that
increases in black minority groups
are viewed as threatening to the status of the white majority.
Tests of the threat hypothesis
-
21
evaluate the correlation between percent black and the size of a
police force or the use of arrest
by police. A significant positive association between the size
of minority group and crime
control is taken as an overall indication that there is a
response to perceived threat to elites.
Therefore, the indirect measurement of threat has become a
staple in previous research on the
conflict perspective’s hypotheses on the administration of crime
control.
The inability to directly measure threat presents an immediate
obstacle for theory testing,
thus leaving the interpretation of the correlation between the
size of a minority group and crime
control susceptible to multiple interpretations. For example, a
structural functionalist argument
would say that increases to the size of minority groups
represent a remonstrance to the
universalistic values of the public good (Durkheim 1933;
Inverarity 1976; Lauderdale 1976).
This argument was made by Inverarity (1976:278) who argues that
the observed positive
relationship between percent black and lynchings across
Louisiana Parishes is a function of the
detrimental effects of racial heterogeneity on the stable
operation of societal order. A similar
example can be provided using social disorganization theory
(Shaw and McKay 1972).
One of the key exogenous variables in social disorganization
theory is population
heterogeneity. According to social disorganization theory,
racial heterogeneity impedes
communication and centralizing community interaction. This in
turn weakens informal social
control and increases the need for formal social control (Bursik
1988; Sampson and Groves
1989; Shaw and McKay 1972). But this relationship between
heterogeneity and crime is
inherently nonlinear. As population heterogeneity lessens
(increasing minority group
homogeneity), reliance on formal mechanisms of social control
(such as arrest) should also
decrease. As noted above, Blalock (1967) argued for a nonlinear
relationship between
discrimination and percent minority. In his power-threat
hypothesis, increases in percent
-
22
minority threaten non-minority populations. However, once the
minority groups numbers are
large enough, they are able to defy the discrimination by
non-minority groups. The presence of
this curvilinear relationship has been supported by research on
the conflict perspective. Jackson
and Carroll (1981) identified this concave relationship between
percent black and police
expenditures, while Greenberg, Kessler, and Loftin (1985) found
the same form in the
relationship between percent non-white and police force
size.
The same relationships predicted by conflict theory can also be
explained by competing
theories’ hypotheses, which fails to falsify the predictions of
any theory. In order to avoid the
problem of multiple interpretations and failing to falsify
conflict theory’s threat hypothesis, it
stands to reason that a more direct measure of threat is needed.
This dissertation offers a more
direct measure of threat—the displacement of citizens by
Hurricane Katrina—which allows for
the falsification of competing theoretical explanations.
The Proper Unit of Analysis: The Distribution of Minority Groups
and the Benign Neglect
Hypothesis
The notion of group distribution within society has a place in
the conflict perspective.
For example, Quinney (1974; 1977) argues that the powerful
utilize class fragmentation to get
the powerless class to fight within themselves. Blauner (1972)
suggests that segregation can
influence the administration of social control by affecting the
relative probabilities of the
offender-victim dyad. Specifically, the segregation of blacks
(the less powerful) lessens the
likelihood of interracial victimizations and simultaneously
increases the probability of intraracial
victimizations. Thus, based on the benign neglect hypothesis,
one would predict that the
segregation of racial minorities would reduce reliance on more
overt and formal mechanisms of
social control (Blauner 1972; Spitzer 1975).
-
23
Evaluation of the benign neglect hypothesis looks at a city’s
level of segregation and its
relationship to crime control. If a significant, negative
relationship is observed, this is taken as
evidence of benign neglect. Measurement of segregation is
typically done through the use of the
dissimilarity index, which compares the racial composition of
city blocks to the racial
composition of the city as a whole. This is done for all blocks
in order to measure the
percentage of one racial group that would have to move to a
different neighborhood to be
distributed similarly as another racial group, and produces an
index score that measures
segregation for the entire city (See Massey and Denton, 1988).
Thus the benign neglect
hypothesis is studied at the city level of analysis. This
research contends that this level of
analysis may be inappropriate for the study of the benign
neglect hypothesis given the
mechanisms underlying benign neglect.
Picture a sample of cities, each containing only one or two
neighborhoods that are
segregated. If this is the circumstance, there would likely be
little variation in the level of
segregation across a sample of cities. Hence, measures of
segregation probably would have
little, if any, impact on crime control outcomes. Therefore, the
segregation on crime for the one
neighborhood of a city would go unnoticed (Blauner 1972; Spitzer
1975). In addition to the
mechanisms of benign neglect warranting further investigations
of the proper unit of analysis,
there is another reason to suspect a more appropriate unit of
analysis—police use of social
control is a function of the social context within which they
operate (Klinger 1997; Klinger
2004; Smith 1986; Smith, Visher, and Davidson 1984; Terrill and
Reisig 2003). Because the
conflict perspective is interested in explaining social control,
it stands to reason to evaluate social
control at a level where effects may differ.
-
24
Research has demonstrated that social control is a function of
social context. Smith,
Visher, and Davidson (1984) included a measurement of
neighborhood poverty in a multivariate
analysis exploring the antecedents of arrest across sixty
residential neighborhoods. Their
analysis shows that the probability of arrests by police
increased as the poverty level of the
neighborhood increased. Smith (1986) analysis on the effects of
11 neighborhood characteristics
(e.g., poverty, residential mobility, racial heterogeneity)
identified that: police are more likely to
stop suspicious people in racially heterogeneous neighborhoods;
in neighborhoods of lower
socioeconomic status, the odds of arrest were higher; and there
was an inverse relationship
between neighborhood crime rates and the likelihood of police
officers taking a crime report.
Smith (1986: 337) concludes that “officers act differently in
different neighborhood contexts.”
Terrill and Reisig (2003) found that police use of force was a
function of the
neighborhood context. More specifically, Terril and Reisig’s
analysis included a factor of
neighborhood poverty consisting of percent in poverty, percent
unemployed, percent of female-
headed families, and percent black and found a significant,
positive relationship between
neighborhood poverty and police use of force, as well as
neighborhood crime level and police
use of force. The totality of these findings has spawned the
advance of a theory that
acknowledges community characteristics when explaining police
action (Klinger 1997).
Moreover, this research supports the arguments made in this
dissertation that past conflict
perspective studies using the city as a unit of analysis may
have masked observable effects.
THE CURRENT STUDY
Clearly the inability to provide a direct measure of threat and
addressing aggregation bias
are important considerations when testing the threat and benign
neglect hypotheses. The current
research utilizes a quasi-experimental, time series design to
account for the occurrence of
-
25
Hurricane Katrina and the resulting displacement of citizens to
Houston and determine whether
or not this event impacted arrest levels of various police
districts. I believe the use of citizen
displacement by Hurricane Katrina constitutes a more direct
measure of threat, and inspection of
police districts rather than just the city addresses the issue
of aggregation bias.
Citizens were ordered to evacuate New Orleans on short notice,
and eventually Houston
housed approximately 250,000 evacuees (Berger 2006). I believe
that the abrupt nature and
magnitude of citizen displacement constitutes a threatening
event for both the displaced citizens
as well as the citizens that make up communities that now must
house a larger population. By
accounting for the displacement of citizens by Hurricane
Katrina, this research provides a more
direct measure of threat than the traditional size of minority
population.
Furthermore, data on clearance levels will be inspected by
police division, not just at the
city level of analysis. The use of police divisions allows for
the characteristics of the each
division to be accounted for. As police behavior can be
influenced by the divisions within which
they operate (Klinger 1997; Klinger 2004; Smith 1986; Smith,
Visher, and Davidson 1984;
Terrill and Reisig 2003), it is important to account for changes
in clearances within a city and
across divisions. The time series design of this study allows
for this to occur. Moreover,
because displaced citizens were housed within specific police
divisions, this allows for specific
predictions on which areas of Houston should experience threat
or benign neglect.
By directly accounting for the occurrence of a threatening event
to all populations and
evaluating the effects of this event within a city, this paper
is positioned to provide a more direct
test of the threat hypothesis and benign neglect hypothesis.
-
26
CHAPTER THREE
SOCIAL CONTEXT AND THE CRIME-ARREST RELATIONSHIP: EVALUATING THE
PREDICTIONS OF THE CONFLICT PERSPECTIVE
INTRODUCTION
The previous chapter was devoted to the predictions of the
conflict perspective’s threat
and benign neglect hypotheses, and presented the argument that
issues of measurement and
aggregation bias may play a role in the proper identification of
the level at which threat and
benign neglect occur. Central to the argument was the role of
social context. Succinctly stated,
Chapter Two argued that it may be important to take into account
characteristics of the social
context when evaluating conflict theory hypotheses. This chapter
extends the argument that
social context is an important component of the conflict
perspective to research on the crime-
arrest relationship.
It has long been posited that crimes and arrests are interlinked
(Becker 1968; Gibbs 1975;
Tittle 1980). Three findings from the research on the
crime-arrest relationship are paramount to
the present discussion: (1) most studies rely on the predictions
of hypotheses derived from
economic theories and fail to take into account the predictions
of the conflict perspective; (2)
results from previous research appear to be a function of the
methodology used to examine the
reciprocal relationship between crimes and arrests; and (3) the
role of social context appears to
be an important part of the crime-arrest relationship.
There are many hypotheses surrounding this relationship. Because
the current study is
concerned with the macro level crime-arrest relationship,
attention will be paid to the relevant
macro-social literature. This chapter is presented accordingly:
first, a review of the theoretical
predictions regarding the crime and arrest relationship is
discussed. This is provided so in order
-
27
to demonstrate how the predictions of the conflict perspective
can be incorporated into the
research on the crime-arrest relationship. Second, a review of
the research on the reciprocal
relationship between crimes and arrests is presented. This
section will organize the relevant
literature around the methodology used to examine the
crime-arrest relationship (i.e., cross
sectional or longitudinal). This vital section demonstrates how
past research has relied on
economic theories of crime, although, as demonstrated in the
first section of this chapter, the
conflict perspective may have utility in explaining the
crime-arrest relationship. This section
also demonstrates how the methodology of past research on the
crime-arrest relationship has
driven findings. Finally, an argument is offered on how threat
may influence the social context
of an area and may be useful in examining the relationship
between crime and arrests.
THE CAUSAL MECHANISMS UNDERLYING THE CRIME-ARREST
RELATIONSHIP
The attempt to identify the causal linkage between crimes and
arrest has a long history
(see Geerken and Gove 1977). Much of this history surrounds an
attempt to unearth the true
causal ordering of the relationship, as well as the direction of
this relationship. For example, one
can imagine how crimes would affect arrests (e.g., as crimes go
up, police respond with more
arrests). One could also picture how arrests affect crimes
(e.g., as arrests increase, criminals
become more aware and crime decreases). In the case of the
direction of the relationship, it is
viable to picture a situation in which crimes exert a positive
influence on arrests, but it is also
likely that crimes could exert a negative impact on arrests, for
example. The discussions below
will illustrate hypotheses dedicated to explaining the
relationship between arrests and crimes.
Economic theory has been at the forefront of reasoning behind
research into this
relationship (Becker 1968; Ehrlich 1973; Schmidt and Witte
1984). Economic theory
-
28
approaches the relationship between crime and arrests using a
precise model based on a general
economic perspective. Namely, economic theory assumes that
individuals have relatively stable
interests. Behavior exhibited by individuals is a result of this
stability and the weighing of costs
versus benefits to alternative behaviors. Individuals choose to
exhibit behavior that maximizes
the ratio of benefits to costs. Therefore, individuals will not
exhibit particular behaviors when
the costs of that behavior outweigh the benefits of displaying
that behavior (see, Becker 1968).
The role of crime control (e.g., arrests), from an economic
standpoint, is to raise the costs
of criminal behavior so that individuals will choose to exhibit
non-criminal behavior because the
ratio of benefits to costs is reduced. Failure to increase the
costs of crime will result in more
criminal behavior. Thus, the crime arrest relationship can be
viewed as part of an overall general
model (see, Liska 1987). Further, crime control can be affected
by workload and resources (see
Phillips and Votey 1981; Liska et al. 1985). The following
section details predictions of an
economic approach to the crime arrest research.
The first hypothesis is termed the public choice hypothesis. It
predicts that increases in
crime lead to increases in arrests. This hypothesis makes the
assumptions that public citizens are
knowledgeable about the costs and consequences of municipal
functioning (Borcherding and
Deacon 1972: 892) and the polity respond to voter outcry
(Tridimas 2001: 300). Therefore, this
hypothesis states that increases in crime lead to citizens’
placing more pressure on the
government (i.e., police, politicians) to crack down on crime in
their area. As an example,
politicians, serving at the pleasure of voters, may cry out for
more crime control because of
pressure from the public. This may then result in more arrests
as crime increases (Borcherding
and Deacon 1972; Deacon 1978; Tridimas 2001).
-
29
Another hypothesis put forth by economic theory predicts that
increases in crime lead to
decreases in clearances. This hypothesis is known as the
overload hypothesis (Chamlin and
Brandl 1998; Fisher and Nagin 1978; Geerken and Gove 1977;
Schulenberg 2003). The
overload hypothesis predicts that increases in crime can
overwhelm the capacities of law
enforcement. When police capacities have reached their limit,
the ability to respond to other
crimes is greatly reduced. Therefore, the lessening of police
resources leads to fewer aggregate
arrests (Fisher and Nagin 1978; Geerken and Gove 1977).
Economic theory also hypothesizes how arrests can influence
crime, the deterrence
hypothesis (in general, see Becker 1968; Geerken and Gove 1977;
Gibbs 1975; Lynch 1999;
Nagin 1998; Paternoster 1987; Petersilia 1998; Pratt, Cullen,
Blevins, Daigle, and Madensen
2006). The deterrence hypothesis also operates under a
utilitarianism approach by arguing that
humans are rational actors who base their decisions on the
maximization of profit. Because
rational actors weigh the costs of exhibiting a behavior against
the benefits of exhibiting the
same behavior, an increase in arrest (i.e., the certainty of
punishment) should sway rational
actors away from performing criminal behaviors. The deterrence
hypothesis argues that
increases in arrests (cost) make crime a less profitable option
and, therefore, leads to a decrease
in crime (Becker 1968; Geerken and Gove 1977; Gibbs 1975).
Another hypothesis, the incapacitation hypothesis, makes a
similar prediction of the
affect of punishment (e.g., arrests) on crimes (in general, see
DiIulio and Piehl 1991; Geerken
and Gove 1977; Gottfredson and Gottfredson 1994; Marvel and
Moody 1994; Petersilia 1992;
Piehl and DiIulio 1995; Sabol and Lynch 2000; Spelman 2000;
Visher 1987; Zimring and
Hawkins 1988). Increases in arrests lead to the removal of
offenders from the social order,
thereby reducing the amount of crime that could have been
committed had these individuals not
-
30
been apprehended. To the extent that an increase in arrests
removes individuals from society
and they cannot return to the general public while awaiting
criminal justice proceedings, the
aggregate level of crime should decrease (Geerken and Gove
1977).
The hypotheses outlined above provide predictions on the
reciprocal relationship between
arrests and crimes under the umbrella of economic theory. As
will be discussed in the
subsequent section, the vast majority (all but one) of the
macro-social studies on the crime-arrest
relationship evaluate the crime-arrest relationship using one or
more of these hypotheses. Far
fewer tests have used the predictions of the conflict
perspective to inform research on the
reciprocal relationship between arrests and crimes. The
following paragraphs will illustrate how
the predictions of the threat hypothesis and benign neglect
hypothesis can be utilized to better
understand the crime-arrest relationship.
As discussed in Chapter Two, the conflict perspective
conceptualizes crime control as a
tool used by elites to advance their own interests (Black 1976;
Blalock 1967; Quinney 1970,
1977; Turk 1969). This view-point argues that culturally
dissimilar groups are viewed as threats
to the social order by elites and law enforcement practices are
used as a mechanism to control
these threatening populations (Turk 1969). The threat and benign
neglect hypotheses are two
predictions used to explain the use of social control. The
threat hypothesis predicts an increase
in arrests when the status of the social elites is threatened.
The benign neglect hypothesis
predicts a decrease in arrests when the non-elite are
threatened. Therefore, the difference
between an increase or a decrease in social control is the group
that is experiencing threat.
The utility of the threat and benign neglect hypotheses role in
the crime-arrest
relationship is clear. If the displacement of citizens by
Hurricane Katrina to Houston, Texas,
increased fear of crime among elite populations (i.e., the group
experiencing threat), this would
-
31
be reflected in a positive association between crimes and
arrests (the threat hypothesis). If the
group experiencing threat were the non-elite populations, then
the benign neglect hypothesis
would be reflected by a negative association between crime and
arrests (see Chamlin and Myer
2009).
The relationship and direction between crime and arrests has
received significant
empirical examination. The following section reviews macro-level
research on this relationship.
Two issues are relevant—methodology used and social context. As
such, the section that follows
will begin with a discussion on cross-sectional research,
followed by discussions on longitudinal
research. Finally, research on the role of social context in the
examination of the crime-arrest
relationship is presented.
A REVIEW OF THE CRIME-ARREST LITERATURE
Examination of the crime-arrest relationship has typically
employed two different designs
in an attempt to investigate the reciprocal nature of crimes and
arrests: cross sectional designs
and longitudinal designs. Research on the reciprocal
relationship between crimes and arrests
began under examination of cross-sectional data. This section
begins with a review of research
on the crime-arrest association with cross-sectional data.
Sjoquist (1973) used data from 53 municipalities with
populations between 25,000 and
200,000 in 1960 to examine 1968 property crimes (i.e., burglary,
larceny, and robbery) and
arrests and convictions reported to the FBI. Multiple regression
was used to estimate the effects
of arrests on crimes. Results demonstrated that an increase in
the probability of arrest and
conviction was significantly related to a decrease in property
crime. This result offers support
for the deterrence hypothesis. Tittle and Rowe (1974) also
revealed support for the deterrence
hypothesis. Using data from cities and counties with populations
over 2,500, the authors found a
-
32
significant negative correlation between certainty of arrest and
crime rates for counties (-.65) and
cities (-.19).
Geerken and Gove (1977) used data on 65 SMSAs with populations
over 500,000. For
each of these units, the authors combined the reported number of
crimes (i.e., homicide and non-
negligent manslaughter, aggravated assault, rape, robbery,
burglary, larceny, and auto theft) for
the years 1970, 1971, and 1973. For each of these indices,
Geerken and Gove then created a
certainty of arrest measure by dividing the number of crimes
cleared by the number of crimes
reported. These two measures were then correlated to determine
if any significant relationship
existed. Geerken and Gove found a significant negative
association for all crimes (r = -.25), auto
theft (r = -.65), larceny (r = -.41), robbery (r = -.40),
burglary (r = -.22), and rape (r = -.26).
There were no significant findings for either assault or
homicide. Thus, the authors argue their
findings, at least in regards to economic crimes, support the
deterrence hypothesis.
Brown (1978) analyzed crime and arrest data in California cities
and counties. In regards
to the California counties (n=58), Brown found a negative
association between the arrest
clearance rate and the crime rate (r = -.39). When cities in
California were the unit of analysis (n
= 114), a negative association between arrest probability and
crime rate also emerged (r = -.17),
although smaller than the counties measure of association.
Using data from 35 large American cities in 1975, Wilson and
Boland (1978) used two-
stage least squares regression to examine the crime arrest
relationship. Their analysis revealed a
significant negative relationship between the robbery crime rate
and the robbery arrest ratio
while controlling for age, sex, race, unemployment, and
population. The same findings could
not be replicated for burglary or auto theft.
-
33
While the above studies demonstrate that the crimes and arrests
are in fact related to one
another, the above cross-sectional research fails to do is
explain the mechanisms underlying this
relationship. The above research cannot articulate whether
crimes influence arrest, or whether
arrest influence crimes. While one could employ simultaneous
equations to unravel the
reciprocal effects between crimes and arrests, it would involve
the ability to identify and employ
instrumental variables that allow one to uniquely identify the
two equations (Hanushek and
Jackson 1977). However, there has yet to be an identifiable
instrumental macro-level predictor
that affects one of these variables and not the other (for
example, see Fisher and Nagin 1978;
Greenberg and Kessler 1982; Nagin 1978). In short,
cross-sectional designs investigating this
relationship are unable to deal with simultaneous equation bias
(Greenberg and Kessler 1982).
As Logan (1975) laments, cross-sectional designs fail to divulge
the direction of the crime-arrest
relationship.
In an attempt to unearth the proper temporal order of the
crime-arrest relationship, many
studies began to use longitudinal designs. Specifically,
researchers employed the technique of
multi-wave panel designs to examine the reciprocal relationship
between arrests and crimes.
Logan (1975) used 1964 to 1968 state level data to examine the
effect of arrest clearances on
crimes. Arrest clearance was calculated as the number of crimes
cleared by arrest divided by the
number of crimes known to police. Seven indices were created for
each of the FBI Type I
crimes (i.e., homicide, rape, robbery, aggravated assault,
burglary, larceny, and auto theft).
Logan employed a two-wave panel design for successive years to
examine a lagged and
immediate relationship between yearly measures of crime and
arrest. Findings from Logan’s
analyses reveal no evidence of any relationship between crimes
and arrests or arrests and crimes.
-
34
Greenberg, Kessler, and Logan (1979) used a three-wave panel
design on yearly arrest
clearance and offense rates for 98 U.S. cities for the years
1964 to 1970. As above, the panel
design was used to examine possible lagged and instantaneous
effects of the reciprocal
relationship. Of the 24 estimated coefficients examining the
effect of crime on arrest, only two
were significant (assault and homicide). However, the assault
coefficient was small and unstable
and the homicide coefficient was not in the predictive direction
(i.e., positively related). For the
24 coefficients in their model predicting the effect of arrest
on crime, no coefficients emerged as
significant predictors. In an extension of this analysis,
Greenberg and Kessler (1982) again used
a three-wave panel design on the same data, this time
controlling for 12 structural predictors.
The results of these analyses provided few estimates reaching
statistical significance, and were
thus not noticeably different from the above study.
The above panel designs were an attempt to address the troubles
cross-sectional designs
had in determining the reciprocal relationship between crimes
and arrest, and these studies were
unable to replicate the findings of cross-sectional studies.
However, panel studies are not
without their own limitations. Namely, the panel studies and the
data used place restrictions on
the estimation of the lag structure between crime and arrests.
The studies by Logan (1975),
Greenberg et al. (1979), and Greenberg and Kessler (1982) use
yearly data. If the relationship
between arrests and crimes is distinguished by yearly lags, then
the techniques used are
satisfactory. However, there is reason to believe that the
reciprocal relationship between crimes
and arrests occurs at a smaller lag (see, for example, Bursik,
Grasmick, and Chamlin 1990;
Chamlin 1988; Chamlin, Grasmick, Bursik, and Cochran 1992).
Thus, the inability of panel
designs to provide evidence of a reciprocal relationship may be
due to model misspecification.
-
35
Chamlin (1988) addressed the identification of a lagged
relationship through the use of
auto regressive integrated moving average (ARIMA) techniques.
The use of ARIMA techniques
does not require one to make rigid assumptions about the
potential lag structure between arrests
and crimes. Rather, bivariate ARIMA analyses allow the data to
determine the lag structure.
This is beneficial because no theory specifies precisely the lag
at which crimes affect arrests or
vice versa (Greenberg and Kessler 1982; Loftin and McDowall
1982). To this end, Chamlin
(1988) uses monthly data for both Oklahoma City and Tulsa
spanning 1967 to 1980 for the
offense categories of robbery, burglary, larceny, and auto
theft. Chamlin found that arrests have
a negative effect (with a 1 month lag) on crime for robbery, but
not for the remaining crimes.
This is in partial support of the deterrence hypothesis.
However, there was no lagged
relationship evidenced for the effect of crime on arrests for
either city on any of the offense
types.
In a further examination of the lag structure between arrests
and crimes, Chamlin and
colleagues (1992) evaluated the reciprocal relationship between
arrests and crimes for the
individual offenses of robbery, burglary, larceny, and auto
theft in Oklahoma City. Each of the
series was examined at the monthly, quarterly, and semiannual
time aggregation. Bivariate
ARIMA analyses indicated: (1) crime had no significant lagged
effects on arrests; (2) monthly
levels of arrest had significant negative effects on robbery
crimes and auto theft at the one month
lag; and (3) quarterly arrests had a negative effect on larceny
crimes at the 1 quarter lag. Thus,
the results support a short term deterrent effect.
Bursik, Grasmick, and Chamlin (1990) argue that prior
examinations of the reciprocal
relationship between crimes and arrests are not only hindered by
the inflexibility in examining
the proper lag structure, but they are also confounded with
aggregation bias. Greenberg and
-
36
colleagues (1981) and Green