|UFDC Home||myUFDC Home | Help|
This item has the following downloads:
STRUCTURAL CORRELATES OF RACE-SPECIFIC DRUG SALES ARRESTS
OVER TIME: ARREST TRAJECTORIES FROM 1980-2001
SCOTT RICHARD MAGGARD
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
Scott R. Maggard
Sean Allen McCluskey
This dissertation is dedicated to the best friend I could ever ask for, Sean McCluskey. A
sociologist at heart, it was Sean who first convinced me to change my maj or from
psychology to sociology. Sean had the uncanny ability to see the world from what
seemed to be an infinite number of perspectives. I learned a great deal about life from
Sean and I hope that he learned as much from me as I did from him. I will never have
another friend like Sean; however I am forever grateful for the years that I was able to
know him and share so much with him.
First and foremost I thank Karen Parker, my mentor, chair, and friend. I first met
Karen as I began a teaching assistantship for research methods and teaching the
laboratory sessions for her course. Karen also introduced me to the theoretical and
methodological approaches discussed throughout this research. Her expertise and
willingness to share that expertise are second to none. My approach to research in general
is much attributed to the knowledge I gained through working closely with her. I am also
grateful to Aaron Griffin for his willingness to share her time, especially these past few
Coupled with Karen, Lonn Lanza Kaduce, Ron Akers, Alex Piquero, and Joe
Spillane comprised what I would argue to be the most solid and well-balanced committee
one could ask for. Lonn was always there to provide practical advice, whether needed
within the academy or elsewhere. From Ron Akers I learned much about both theory and
the assessment of theories. Alex provided methodological expertise, and seeing his use of
these techniques in the life-course arena led to the idea to apply these methods to drug
arrests in the first place. Finally, Joe provided not only his never ending insight into the
world of drug use and drug policy, but also added much needed humor and reflection on
life in general.
The remainder of the faculty and staff in the Department of Criminology, Law, and
Society all deserve thanks for their willingness to share their knowledge and experiences
with their students. Whether writing the dissertation to obtaining employment, they were
always approachable and willing to help.
I thank Bob Jones, who is the architect of the PROC TRAJ software plug-in that
was utilized in this research. I am grateful for the many hours Bob has sacrificed
answering emails and assisting me to better understand and use the PROC TRAJ
software. Bob's support for a software program for which he is not paid is unprecedented.
Craig Boylstein and Kim Raymond deserve much credit for alleviating stress and
providing much needed humor when needed. Craig was one of the first people I met in
the sociology program after moving to Gainesville in 1999, and we immediately became
I thank my parents for putting up with what sometimes seemed to them my
everlasting j ourney to remain a student as long as possible. They will likely get as much
satisfaction from the fact that I am finally exiting school as anyone. I would not have
been able to do this without them.
The many fellow graduate students with whom I became friends at the University
of Florida I also thank. John Reitzel, and Jennifer Matheny both provided much needed
humor and entertainment throughout graduate school.
Finally I thank my wife, Allison Chappell. Without Allison I likely would not be
finishing this document. She is my life, and her selflessness which made this possible I
will be forever grateful for. Allison has been there day and night, through rough times
and bright, pushing me along in order to complete this task.
TABLE OF CONTENTS
ACKNOWLEDGMENT S .............. .................... iv
LI ST OF T ABLE S ................. ................. viii............
LIST OF FIGURES .............. .................... ix
AB S TRAC T ......_ ................. ............_........x
1 INTRODUCTION ................. ...............1.......... ......
Introducti on ................. ...............1.................
Back ground ................. ...............1.................
Present Study ................. ...............3.......... ......
Specific Aim s .............. ...............6.....
Significance .................. ...............7.................
Layout of the Dissertation ................ ...............9................
2 THEORETICAL PERSPECTIVES............... ..............1
Introducti on ................. ...............10.................
Social Disorganization ................. ...............11........... ....
Concentrated Urban Disadvantage ................. ...............18........... ....
Unemployment and the Urban Underclass ................. .............................18
Poverty and Segregation ................. ............ ... ................22....
Racial Threat: Blalock's Power Threat Hypothesis ................. ........................27
Conclusion ................ ...............35.................
3 REVIEW OF THE LITERATURE .............. ...............38....
Introducti on ................ ... ....... .... ... .. ...............38...
Race, Drugs and Crime: A Historical Perspective............... ..............3
Race, Drugs and Crime: The New Era .............. ...............44....
Hypotheses and Expectations .................. ........ .............4
Social Disorganization and Drug Sales Arrests ................. .......................51
Concentrated Urban Disadvantage and Drug Sales Arrests .............. .... ........._..52
Racial Threat and Drug Sales Arrests .............. ...............53....
4 DATA AND METHODS .............. ...............55....
Introducti on ................. ...............55.................
Data Sources .............. ...............55...
Dependent Variable ................. ...............56.......... ......
Independent Variables ................ ...............58.................
Methodology ................. ...............63.................
Statistical Procedures............... ...............6
Ri sk F actor s ................. ...............67.......... ......
5 DESCRIPTIVE STATISTICS............... ...............6
Introducti on ................. ...............69.................
Descriptive Statistics .............. ...............69....
Independent Variables ................ ...............69.................
Dependent Variable ................. ...............73.................
Drug Sales Arrests Traj ectories ................. ....__. ....__. ...........7
Conclusion ........._.__........_. ...............78....
6 MULTIVARIATE RESULTS ............. .....__ ...............85..
Introducti on .............. ......_ ...............85...
White Drug Sales Arrests .............. ....... ...............86.
Accounting for Change: 1980-1990 .............. ...............86....
Accounting for Change: 1990-2000 .............. ...............89....
Black Drug Sales Arrests .............. .........__ ...............90..
Accounting for Change: 1980-1990 .............. ...............90....
Accounting for Change: 1990-2000 .............. ...............92....
Conclusion ............. ...... ...............92...
7 CONCLU SION................ ..............10
Discussion and Implications ............. .....__ ...............100..
Lim stations ............. ...... ._ ...............103..
Future Research .............. ...............105....
Final Thoughts ............. ...... ._ ...............106..
APPENDIX: SAMPLE CITIES .............. ...............108....
REFERENCES ............. ...... ...............115...
BIOGRAPHICAL SKETCH ............. ...... ...............129...
LIST OF TABLES
4-1. Concentrated Disadvantage Index Factor Loadings. ................. .................6
5-1. Descriptive Statistics (Means with Standard Deviations in Parentheses). .................81
5-2. Mean Percent Changes Across Time for Independent Variables. .............. ...............82
5-3. Means and (Standard Deviations) for Drug Sales Arrest Rates. ............. .................82
6-1. Parameter Estimates and (standard errors) Modeling Structural Changes from
1980-1990 on White Drug Sales Arrests. ................. .....___............. .....9
6-2. Parameter Estimates and (standard errors) Modeling Structural Changes from
1990-2000 on White Drug Sales Arrests. ................. .....___............. ......9
6-3. Parameter Estimates and (standard errors) Modeling Structural Changes from
1980-1990 on Black Drug Sales Arrests. ................ .....___.............. .....9
6-4. Parameter Estimates and (standard errors) Modeling Structural Changes from
1990-2000 on Black Drug Sales Arrests. ................ .....___.............. ......9
A- 1. Traj ectory Group Memb ershi p for White Drug S al es Arre sts ................. ...............10 8
A-2. Traj ectory Group Membership for Black Drug Sales Arrests ................. ...............1 12
LIST OF FIGURES
5-1. Traj ectories of White Drug Sales Arrests 1980-2001 ......._____ .... ....__ ...........83
5-2. Traj ectories of Black Drug Sales Arrests 1980-2001 ........._.__.... ...__ ............84
Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
STRUCTURAL CORRELATES OF RACE-SPECIFIC DRUG SALES ARRESTS
OVER TIME: ARREST TRAJECTORIES FROM 1980-2001
Scott R. Maggard
Chair: Karen F. Parker
Cochair: Lonn Lanza-Kaduce
Major Department: Sociology
The relationship between community structure and crime has received a great deal
of research attention in criminology over the past two decades. Rising out of the
traditions of Shaw and McKay, researchers have documented how structural changes in
communities are related to crime rates in those areas. While the maj ority of these studies
have focused on property and violent crimes, few studies have investigated the
relationship between social structure and race-specific drug arrests. Moreover, most
studies investigating structural correlates of crime have used decennial time periods and
typically employ change score techniques, thereby only allowing between-city
comparisons, while neglecting aI ithrin-city comparisons. Employing techniques used to
study the life-course of individual offenders over time, this research aims to classify the
long term behavior of drug sales arrests in large cities as distinct traj ectories over time.
Assuming that cities behave differently, this research will shed light on how structural
changes in cities affect changes in arrest traj ectories over time. Findings support the
hypotheses that cities do in fact behave over time in regards to drug sales arrest from
1980-2001. Moreover they vary significantly by race and while certain cities may have
experienced exponential growth in Black drug sales arrests, other cities witnessed similar
(yet less pronounced) growth in White drug sales arrests. Findings also provide support
for both social disorganization and concentrated disadvantage perspectives on urban
crime. The increase in concentrated disadvantage among Blacks and Whites from 1980-
1990 significantly impacted the likelihood of those cites being in higher drug sales arrest
traj ectories. Additionally, those cities which experienced dramatic increases in residential
mobility from 1980-1990 were more likely to be in higher drug sales arrests trajectories
as compared to the lowest. Overall these findings suggest that the structural changes in
large cities occurring from 1980-1990 had a more significant impact on drug sales arrest
rates than the changes occurring from 1990-2000.
The primary focus of this dissertation is to investigate the extent to which structural
factors in large urban areas contribute to the rise in drug sales arrests over time. Many
researchers have focused on the individual factors that lead to drug use and crime.
However an alternative approach is to discern to what extent ecological factors contribute
to the distribution of crime in urban areas, and specifically drug sales arrests. As large
urban centers have undergone a metamorphosis covering the past several decades, the
individual actors within these areas have been subj ect to "cognitive landscapes" of crime
and a rather dismal outlook for the future (Sampson and Wilson 1995).
Therefore, peering through an ecological or "macro" lens, it is believed that it is the
structure of these urban landscapes that permit crime to permeate the lives of the
residents, rather than simply being that the residents of these areas are somehow more
likely to commit crimes on their own. The goal of this dissertation is to a) assess to what
extent large urban areas "behave" over time (using drug sales arrests as a proxy of
behavior), and b) determine what structural factors predict the likelihood of a city
experiencing sharp increases in drug sales arrests over time.
In the tradition of the Chicago School, structural theories of crime and delinquency
focus on the community to understand criminal activity. Rather than focusing on the
individual, structural theories seek to understand how community structure may either
foster or inhibit delinquent behavior. The classic work of Shaw and McKay (1942) laid
much of the groundwork which led to our current conceptualization of community social
disorganization. Their argument was not that there is a direct relationship between
community factors such as economic status and delinquency, but that those areas
characterized by dire economic circumstances tended to have high rates of population
turnover and population heterogeneity. This, in turn, weakened informal social controls
within the community resulting in a more favorable environment for delinquency to
occur. Additionally, social disorganization theory also posits that the effects of declining
structural conditions in urban areas are invariant across race. In other words, as family
disruption and low economic status begin to prevail in an urban area, informal controls
break down and crime can result, regardless of race (Bursik 1988; Shaw and McKay
In addition to the social disorganization of urban communities, researchers have
posited that deindustrialization beginning in the 1970's has also contributed to
delinquency in urban areas. Wilson (1987) argues that the decline in manufacturing and
semiskilled j obs beginning in the 1970's has depleted j obs for minorities in those areas.
He argues that black j oblessness has led to concentrated disadvantage among blacks in
urban areas. Others have also argued that racial isolation, joblessness and racial
residential segregation have fostered delinquency in many communities (Massey et
al.1987; Massey and Denton 1988, 1993; Wilson 1987).
The concentrated disadvantage faced by blacks has led to a decline in marriageable
males, a rise in incarceration rates, family disruption, as well as a rise in crime rates
(Almgren et al. 1998; Krivo et al. 1998; Parker and McCall 1999; Sampson 1987). As the
residents of these communities struggled to survive with the loss of the unskilled and
semi-skilled jobs, many relied on the growing informal economy of inner cities for
income (Johnson et al. 1990). In addition, others have documented the dramatic surge of
young people who became involved in the illicit drug trade as being unemployed and
unskilled workers (Fagan 1992).
With urban areas already facing social disorganization and concentrated
disadvantage after deindustrialization, it could be argued that the rise in drug use and
sales only exacerbated these problems. In fact, the war on drugs began an era of targeting
the urban and disproportionately black population of America' s cities with specific drug
law enforcement tactics aimed at curbing the rising use and sales of crack cocaine
beginning in the early 1980's (Tonry 1995).
Drug offense arrests in the United States have varied over time for much of the past
century. However, recent research has demonstrated that drug arrests experienced an
unprecedented rise beginning in the early 1980's throughout much of the 1990's.
Blumstein (1995) has noted that the crack epidemic, which began in the 1980's, was
directly related to the sharp rise in homicide rates during the same decade. Moreover it
began to affect selected urban areas and later spread to other areas of the country as well
(Blumstein 1995). Additionally researchers have recently presented data that demonstrate
the way in which homicide rates changed during this time in different sections of the
country (Messner et al. 2005). They found that thirty five of the sixty eight large cities
which they examined behaved in the "epidemic-like cycle" described by others
(Blumstein 1995; Blumstein and Rosenfeld 1998; Golub and Johnson 1997; Messner et
While most data that researchers and activists alike rely on are based solely on
averages or means, little is known about the 11 ithrin-city variations of arrests over time.
That is, with such a wide margin between cities with the lowest drug arrest rates and
those with the highest (see Mosher 2001; Parker and Maggard 2005), these averages may
paint a picture that presents a false sense of urgency for the entire country. I contend that
it may be that different cities experience different rates of increases of drug arrests over
time. In fact, I argue that some cities may experience few problems with drug arrests,
while others may experience great troubles. Contributing to this issue is that as specific
drug "epidemics" or drug "scares" arise, they often develop and diminish in different
cities at different periods in time, as was observed with crack cocaine (Golub and
While much of the macro level research on communities and crime has focused on
between-city change over different time periods, very little is known about ~I ithrin-city
changes in criminal offending over time. That is, how do specific cities behave in terms
of arrest rates over a long period of time? This is what is known in the life-course
literature as a trajectory. While studies that utilize data for many cities may investigate
how the change in structural conditions contributes to the change in crime rates over
time, they fail to decipher which cities' crime rates changed and why they changed.
Until recently, the idea of "community careers in crime" was very rare (but see
Bottoms and Wiles 1986; Schuerman and Kobrin 1986). Only in the past year has
research surfaced linking the methods utilized in life-course criminology (to analyze the
offending traj ectories of individuals) to macro level analysis of communities and crime
(Griffiths and Chavez 2004; Weisburd et al. 2004).
Schuerman and Kobrin (1986) investigated the 20 year histories of Los Angeles
neighborhoods from a developmental perspective. They used cross-sectional and time
series analysis to discover that neighborhoods could be classified as one of three distinct
stages: emerging, transitional and enduring. They found that neighborhood deterioration
precedes rising crimes rates in the early stages but that rising crime rates precede further
neighborhood deterioration in the later stages. They conclude that efforts to prevent the
rapid escalation of crime in urban areas must be initiated in the emerging stages of the
cycle in order to be most successful (Schuerman and Kobrin 1986). In contrast, Bottoms
and Wiles (1986) focus their research in Britain, and they argue that the key to
understanding the criminal careers in neighborhoods is housing tenure. They find that key
factors include bureaucratic mechanisms and renting versus home ownership (Bottoms
and Wiles 1986).
Weisburd and his colleagues (2004) analyzed traj ectories of crime in
neighborhoods in Seattle over a 14 year period. They used a group based traj ectory
analysis to uncover distinct developmental trends within the city. They conclude that
while a city's crime rates may be declining over a period of time, certain segments within
the city may be rapidly declining or rapidly increasing at the same time, and may
contribute significantly to the overall crime picture in the city (Weisburd et al. 2004).
Griffiths and Chavez (2004) merge a group based traj ectory analysis and
exploratory spatial analysis to investigate homicide trends in Chicago communities from
1980-1995. They identified distinct traj ectories for total, street gun, and other weapon
homicides across 831 census tracts in the city. Key findings included evidence of a
weapon substitution effect for violent neighborhoods that are proximate to each other, in
addition to the fact that street gun specific homicides increased in areas bordering the
most violent areas of the city (Griffiths and Chavez 2004).
This research proj ect seeks to examine the criminal careers of cities using race-
specific drug arrests over two decades, identifying structural correlates to help explain
this phenomenon. I am focusing on the theoretical perspectives of social disorganization,
urban disadvantage and racial threat. This research seeks to address several related
research questions. First, I seek to determine if there are identifiable traj ectories of race
specific drug-sales arrest rates in large urban areas over time in order to determine
whether the changes in certain cities experienced different traj ectories for different races.
In addition, I seek to answer the question of whether certain structural characteristics of
cities contribute to explaining the likelihood of a given city being in a specific traj ectory.
That is, if I determine there to be 4 distinct traj ectories, do social disorganization and
other structural correlates explain the increase in arrests for the highest traj ectory and not
the lowest? Do they help to explain specific traj ectories for specific races? In other
words, are cities plagued with the highest increase in drug activity during this time period
(as measured by drug sales arrests) also the cities which have experienced the greatest
change in ecological factors? These factors may include the deindustrialization of urban
areas, an increase in residential mobility and racial residential segregation, increasing
disorganized family structures throughout the community and the changing ethnic
makeup of the area over time.
The focus of this research will be race-specific drug-sales arrests in the largest
cities in 1980 for the period 1980 through 2001. The justification for utilizing only drug-
sales arrests as the focus for this research is twofold. First, the arrest data dating back to
the 1980's are limited to only drug sales arrests. Second, much of the theoretical
reasoning driving this research revolves around the vanishing manufacturing and semi-
skilled jobs documented during this period in urban areas (Sampson 1987; Wilson 1987,
1991). This trend has also been documented at the micro level, as the ethnographic work
of Elij ah Anderson (1999) illustrates. In the Code of the Streets (1999) Anderson writes:
Deindustrialization and the growth of the global economy have led to a steady loss
of the unskilled and semiskilled manufacturing j obs that, with mixed results, had
sustained the urban working class since the start of the industrial revolutions. At the
same time "welfare reform" has led to a much weakened social safety net. For the
most desperate people, many of whom are not effectively adjusting to these
changes .. the underground economy of drugs and crime often emerges to pick up
the slack. (pp. 108) (Emphasis added)
The significance of this research is that it will enable me to differentiate those cities
that have experienced the greatest increases in drug sales arrests from those that have not,
disaggregated by race. Where prior research describes certain structural variables as
being significant to crime in urban areas, or to the change in crime in urban areas, this
research will enable me to identify cities that have experienced the most dramatic
changes in drug arrest rates. This will enable policy makers to identify specific urban
areas that are in more (or less) need of state or federal aid programs, community building
programs, j ob training programs, etc. For example, if it is found that the highest traj ectory
groups for both blacks and whites also see the largest impact from a decrease in
manufacturing jobs, then it can be said that policy makers should prioritize resources to
those cities rather than cities that may not feel such an impact.
While few studies have focused on structural correlates of drug arrest rates in large
cities (Mosher 2001; Parker and Maggard 2005), structural theories of crime and
deviance have been used to help us understand many other types of crime from property
crimes to homicides (Bursik and Grasmick 1993; Crutchfield 1989; Crutchfield and
Pitchford 1997; Krivo and Peterson 1996, 2000; Parker and McCall 1997, 1999; Sampson
1987; Sampson and Wilson 1995).
There are several theoretical reasons to assume that areas plagued by social
disorganization will witness an increase in drug sales arrests over time. The first is quite
obvious; as deindustrialization has accelerated beginning in the 1970's, many individuals
watched as their employer literally drifted away. It could be argued that this in and of
itself may be enough to warrant expecting an increase in drug sales (income generation).
However when this economic motivation is coupled with the criminogenic environment
that social disorganization and concentrated disadvantage theorists would predict, it
makes even more sense logically. That is, if one desperately needs money to support
her/his family, and the family resides in an environment with little informal controls,
where criminal behavior is somewhat accepted and there exists a lack of hope for the
future, the chances of joining the illegal drug trade should increase.
This research aims to merge the methodologies traditionally used to analyze
individual offending rates over time, with those that have been used to analyze structural
changes in urban areas. That is, I seek to utilize the TRAJ procedure in SAS to
investigate patterns of drug arrest rates over a 22 year period, creating clusters or
traj ectories of cities with similar patterns. In addition, I seek to use the change in
structural conditions between decades to determine their effect on a given trajectory.
This research will add to the literature in several ways. First, it will build on the
efforts of Griffiths and Chavez (2004) and Weisburd et al. (2004) of merging the TRAJ
procedure to macro level analysis. It will add the ability to analyze how the change in
certain structural conditions within cities affects drug arrest rates over time. It will also
enable me to untangle these changes in order to identify whether certain cities were
impacted more severely by these changes than others. Finally, it will evaluate the social
disorganization, urban disadvantage and racial threat perspectives in a new context. This
methodology provides a significant departure from traditional methods, in that statistical
outliers may be grouped together in their own traj ectories and will not have a significant
impact on cities that experience few changes in drug activity over time.
Layout of the Dissertation
The dissertation will consist of 7 chapters. Chapter 2 will provide the theoretical
background driving the research, including a review of the empirical support for each
theoretical perspective chosen. Chapter 3 will provide a review of the literature regarding
drugs and crime as well as present several hypotheses to be tested. Chapter 4 will
describe the data and statistical procedures utilized for testing the hypotheses. This will
include detailed information on the data, as well as background information relating to
the TRAJ procedures that will be performed in SAS. Chapter 5 will provide descriptive
statistics for both dependent and independent variables. It will also include the
traj ectories of drug sales arrests over time, providing unique traj ectories for White and
Black drug sales arrest rates. Chapter 6 will present how the various theoretically drawn
explanatory variables interact with the clusters, or traj ectories, and highlight how each
theoretical perspective may contribute to our understanding of the changes in drug arrests
over time. Finally chapter 7 will provide concluding remarks, policy implications of the
current research, and directions for future research.
While the Chicago School propelled many of the great symbolic interactionists
within sociology, it was also ripe with many of the brightest thinkers in the field in
examining the world of sociology through a macro lens, or ecological approach. It was a
shift that took the emphasis away from asking why certain individuals behave the way
they do, and instead asked, how do particular geographic areas influence rates of crimes
to behave the way they do? It was within the great city of Chicago that what has now
become known as Social Disorganization was born.
During the early 20th century, metropolitan areas of the United States were about to
embark on a maj or metamorphosis from sparsely populated farmlands to highly
populated industrial centers. The transformation not only brought greater prosperity and
opportunities to more people, but it also welcomed an influx of immigrants seeking
employment and new opportunities.
Many of the structural (or ecological) theories within sociology and criminology
today share their roots with the Chicago School tradition. Robert Park and Ernest Burgess
studied the social organization and characteristics of cities. Whether by coincidence or
not, they happened to reside in one of the fastest evolving metropolitan areas in the
country, so what better place to examine how these changes occur over time through a
sociological lens (Cullen and Agnew 2003).
It was in this context that Clifford Shaw and Henry McKay drew upon the theories
of Park and Burgess and their ideas of how a city is like an organism when they
conducted their classic study on juvenile delinquency in Chicago, Juvenile Delinquency
and Grban Area~s (1942). Burgess had noted that as cities such as Chicago transform into
the industrial meccas that they are today, they do so from the inside out. Burgess
documented specific "zones" within a city, beginning in the center and expanding out in
concentric rings or zones (Burgess 1925).
From a city's core (e.g. central business and industrial center), it naturally sprawled
outward over time, through this evolution of concentric rings or zones. Just outside the
central business district, Burgess noted what he coined the "zone in transition". It was
here that immigrants frequently first settled as they arrived to new areas seeking work
opportunities and affordable housing. Just outside the "zone in transition", lied the "zone
of working men' s homes", the "residential zone", and the "commuters' zone" (Burgess
Burgess observed that while most immigrants new to an urban area first settle
within the "zone in transition", most families do not stay in the zone for long periods of
time. In fact, the concentric zones are labeled as such since there appears to be a natural
progression outward for financial and family reasons. Once individuals established stable
employment (typically factory/manufacturing j obs), they would typically save their
money for a more prosperous life in one of the outer zones. Therefore, within the
transition zone there always exists residential mobility both in and out of the area. As
new immigrants excitedly move in seeking employment and housing, another family
leaves the zone in search of a more relaxed, family oriented life in the outer zones. These
outer zones are certainly where most live today, the suburbs or suburbia.
Using this logic, Shaw and McKay hypothesized that these urban centers (e.g.
zones in transition) should have the highest concentration of crime and delinquency, due
to the ever-changing landscape of the areas. They theorized that the persistent residential
mobility and poverty would contribute to these areas being more "disorganized"
compared to areas that were more affluent such as the outer zones within a city. In order
to test their hypotheses, Shaw and McKay set out to collect data on juvenile delinquents
throughout the city of Chicago and map which areas of the city experienced the highest
rates of delinquent behavior (Shaw and McKay 1942).
Shaw and McKay did not believe that any city necessarily had a distinct number of
zones (they are rather arbitrary), or that lines existed for the exact measuring of the zones.
However, they argued that organizing their research based on five zones, each about 2
miles apart, they could map how the organization and changes in these areas affected
juvenile delinquency rates over three periods of time. They collected data on recidivism,
truancy rates and referrals to juvenile court to map the periods of 1900-1906, 1917-1923,
and 1927-1933. Their methodology would enable them to compare different cohorts of
juveniles residing in the same geographic areas at different time periods.
Their results confirmed much of what they had predicted. In all three-time periods,
the central business districts consistently witnessed the highest rates of juvenile
delinquency, with the rates dropping significantly for each successive zone moving
outward from center. In other words, it was not the individuals who resided in these areas
that created high crime rates, but rather the organization (or disorganization) that made
these areas ripe for delinquent activity. Recall this is in stark contrast to earlier
criminological theories, which purported that the key to understanding individual
delinquency lies in individual traits as advanced by social scientists such as Cesare
Lombroso and E.A. Hooten for example (Akers and Sellers 2003; see also Lindesmith
and Levin 1937).
One criticism of Shaw and McKay's assertions lies in the fact that they never
explicitly described how or why the zone in transition would have the highest
concentration of juvenile delinquents (Akers and Sellers 2003; Cullen and Agnew 2003),
although it is assumed today that the true cause is due to the breakdown of informal
controls in these disorganized areas. In other words, an area that is disorganized may be
characterized as one experiencing a lack of the informal control mechanisms that close-
knit communities provide in order to deter and control juvenile delinquents. Further,
researchers have long argued that individuals living in high delinquency urban areas
experience a "duality of conduct norms", where their conduct norms are split between
those of mainstream society and those of delinquent subcultures (Kobrin 1951). In
addition it has been noted that it is extremely difficult for communities to establish
informal control mechanisms when many of its residents are "uninterested in
communities they hope to leave at the first opportunity" (Kornhauser 1978: 78).
While Shaw and McKay's ideas were widely cited, Bursik (1988) notes that they
fell out of favor for many years due partly to the shift to more micro level theories of
crime and delinquency such as differential association/social learning and control
theories. Furthermore, Bursik notes that while widely cited, it was generally in reference
to the economic composition of communities and their respective rates of delinquency,
according to Bursik:
Shaw and McKay did not posit a direct relationship between economic status and
rates of delinquency. Rather, areas characterized by economic deprivation tended to
high rates of population turnover (they were abandoned as soon as it was
economically feasible) and population heterogeneity (the rapid changes in
composition made it very difficult for those communities to mount concerted
resistance against the influx of new groups). These two processes in turn, were
assumed to increase the likelihood of social disorganization, a concept that is very
similar to Park and Burgess' s (1924: 766) formulation of social control as the
ability of a group to engage in the activity of self- regulation" (Bursik 1988: 520).
It was the reintroduction of social disorganization theory in the 1980's that led to
the vast body of research that exists today, a groundwork laid by a select few sociologists
and criminologists. While one reason that social disorganization was virtually abandoned
for several years is surely due to the trendy nature that often occurs in academic circles
regarding research agendas (see Sampson and Laub 2005), much of it was certainly
attributable to the difficulty in undertaking such projects (see Kubrin and Weitzer 2003).
To truly test Shaw and McKay's hypotheses regarding social disorganization, one
would need to conduct a multi-year study that took into account interviews with residents
as well as collect statistics relevant to the area. However, as technology has progressed,
sociologists and criminologists have utilized official crime data and demographic data
collected through the United States Census program to develop proxies, which help us
understand social disorganization and how it may be related to an area' s criminal activity.
Researchers have been able to extend social disorganization theory further than in the
past as well as clarify key conceptualizations.
Upon the revitalization of Social Disorganization theory in the 1980's, studies
began to emerge to gauge to what extent Shaw and McKay's hypotheses remained valid
over time. Upon an "extremely fortunate stroke of luck", Bursik and Webb (1982) were
able to obtain much of the original data utilized by Shaw and McKay from the basement
file cabinets of the Institute for Juvenile Research at the University of Chicago, as well as
data the Institute had compiled through 1970 (Bursik and Webb 1982:28). Utilizing this
newly discovered archive they were poised to examine the affects of four key variables
on the changes in "areal delinquency rates" in Chicago from the 1940's forward through
the 1970's. They measured changes in the community population, the percentage of
foreign-born whites, the percentage of non-whites, and levels of household density. Their
findings indicate that between 1940 and 1950, invasion and succession did not affect the
distributional patterns of delinquency within communities (as Shaw and McKay
However, during the later time periods (notably 1950-1960) community change
wa~s associated with patterns of "areal delinquency rates", which is contrary to Shaw and
McKay's original theses. Rather than state that Shaw and McKay were "wrong", they
argue that Shaw and McKay were writing and making their observations within a much
different historical context than the later years studied by Bursik and Webb. In fact, the
premise of the Burgess' original thoughts on invasion and succession rested on the
assumption that new immigrants would make a natural transition or assimilation into a
new area. Instead, during these later times, immigration and segregation policies changed
the dynamics of how an urban area would react to the invasion and succession of new
residents. They conclude that it is the fact that invasion and succession are occurring, as
opposed to who is involved in the occurrence of these changes. In other words, it matters
little of which race is attributed to the invasion/succession so much as the process itself
(Bursik and Webb 1982; Bursik and Grasmick 1993; Sampson 1987, 1995; Sampson and
Groves 1989; Sampson et al. 1997).
Sampson and others have noted the importance of "collective efficacy" in
maintaining control over youths in high crime areas. They have recognized that the
relationship between disorganization and crime rates is not a simplistic/direct
relationship, but rather complex. Sampson goes on to argue the social disorganization
framework views neighborhoods and communities as a complex system of friendship and
acquaintance relationships which are rooted in family life and ongoing socialization
processes within these urban areas (Bursik 1984, 1988; Bursik and Webb 1982; Sampson
et al. 1997).
Researchers have also noted the reciprocal effects of structural factors and urban
crime. Shihadeh and Steffensmeier (1994) found that reciprocal effects do exist among
measures of family disruption and crime rates in large cities. They note that while family
disruption has a positive significant affect on crime rates, violent crime also has a
positive significant affect on family disruption. They argue that this is due to both the
incarceration of young males as well as the unwillingness of young women to marry
criminally involved young men. Moreover family disruption was found to be among the
strongest predictors of juvenile violent crime due to the lack of informal controls that
strong family networks within communities are believed to provide. Similarly the
dramatic increase in incarceration has been argued to decrease social organization and
thus stimulate the breakdown of informal controls within urban communities Rose and
Clear 1998; Sampson 1987; Shihadeh and Steffensmeier 1994).
More recent research has investigated the relationship between social
disorganization within communities and specific types of homicides. Kubrin (2003) Einds
that the percentage of divorced males in an urban area is statistically significant with only
"general altercation killings", which are homicides that began as a general altercation,
leading the subj ects outside and resulted as homicides. Additionally she found that areas
with high residential mobility experienced a greater number of felony homicides as well
as it affecting the overall homicide rates in those areas, consistent with disorganization
theory. These findings lead us back to the issue of reciprocal effects discussed above.
Kubrin and Witzer (2003) argue that future studies of social disorganization need to
become more dynamic. For instance they argue that street killings within a neighborhood
may increase residential mobility, while domestic homicides may have little impact on
mobility. Since domestic homicides are between spouses or acquaintances residents are
expected to be less fearful than they may be of random or street homicides (such as drug
related homicides) (Kubrin 2003; Kubrin and Weitzer 2003).
Kubrin and Weitzer (2003) go on to argue that researchers evaluating social
disorganization theory need to improve not only the conceptualization of concepts but
also the methods used to test the effects of these structural covariates on crime rates. One
suggestion is to borrow methods from the psychological literature which is used to
identify life course turning points in offending in individuals. While they suggest growth
curve modeling techniques, this study utilizes a procedure that is typically used to model
individual offending, PROC TRAJ, and is described in Chapter four (Kubrin and Weitzer
Concentrated Urban Disadvantage
Unemployment and the Urban Underclass
In The Truly Disadvantaged, William Julius Wilson outlined a multi-faceted
framework describing what he believes has lead to the growth of the urban ghetto in
America, and more precisely, what he has termed the underclass. The underclass, he
states, are the results of "...changes have taken place in ghetto neighborhoods, and the
groups that have been left behind are collectively different from those that lived in these
neighborhoods in earlier years" (Wilson 1987: 8).
He argues that urban Blacks have been especially vulnerable to both geographic
and industrial changes in the economy, which have had the devastating effect of creating
high proportions of joblessness among Blacks. This is due to the history of discrimination
in America, as well as Blacks' migration to large metropolises, both of which have
resulted in weak labor force attachment. He argues that these are the results of a variety
of compounding factors that have contributed to the underclass. These factors include
shifts in the labor market from goods producing sectors to service oriented sectors, the
polarization of the labor market into low salary and high salary j obs, numerous
technological innovations, relocation of manufacturing outside large cities, as well as
economic recessions (Wilson 1991: 640).
All of these factors combined have resulted in a significant increase in the
concentration of poverty in urban areas, as well as increasing both the number of poor
single parent families and the number of families depending on welfare. Moreover, since
1970, inner city neighborhoods have experienced out migration of working and middle
class families, thus leaving a concentrated amount of poverty behind, with "extreme
poverty" (greater than 40%) increasing dramatically (Wilson 1987, 1991).
With the working and middle class black families moving out, Wilson contends
that critical social buffers have been removed from these communities. Past decades had
seen the working and middle class families bring stability to inner city neighborhoods,
through patronizing local stores and churches, sending their children to local schools, and
providing legitimate and meaningful visions to lower class Blacks of the chances of
When all of these factors compound over time, Wilson argues that weak labor force
attachment is inevitable. He goes on to elaborate by stating:
Thus neighborhoods that have few legitimate employment opportunities,
inadequate j ob information networks, and poor schools not only give rise to weak
labor force attachment but also raise the likelihood that people will turn to illegal or
deviant activities for income, thereby further weakening their attachment to the
legitimate labor market (Wilson 1991: 651).
So in addition to further weakening these attachments, crime in the area is
heightened as it manifests and perpetuates itself within the community, with residents
being influenced by the beliefs, behavior, and social perceptions of other disadvantaged
families. He refers to this as "concentration effects", or the "effects of living in an
overwhelmingly impoverished environment" (pp.651).
Sampson and Wilson (1995) have extended Wilson's original ideas with what they
have coined "cognitive landscapes". Cognitive landscapes emerge based on the context of
any given community. In other words, everyone has cognitive landscapes, however, those
landscapes will differ greatly across populations based primarily upon the organization
and structure of the community. Those who reside in impoverished areas such as the
areas described by Sampson and Wilson will form a system of values that is less likely to
condemn drug use and disorder than those of more affluent communities (Sampson and
Wilson 1995; see also Kobrin 1951).
Generally speaking, Wilson's arguments aren't usually tested alone. That is they
are more often incorporated into tests of theories such as social disorganization theory.
Studies focusing on these concepts have generally supported many of Wilson' s claims.
Jargowsky and Bane found during the 1970's alone, the number of people living in
"extreme poverty" (>40%) increased 30 percent and that the proportion of the poor
population residing in ghettos significantly varies by race. They found that in 1980, only
2% of the non-Hispanic poor lived in ghettos, compared to about 21% and 16% of the
Black and Hispanic poor, respectively. In addition they noted that almost one third of
Blacks living in metropolitan areas resided in the ghetto, and 65% of the ghetto poor were
Black. There is also evidence that this trend continued through the 1980's as well
(Jargowsky and Bane 1991; see also Nathan and Adams 1989).
They also added to Wilson' s argument by discovering that most of this increase
was in the Midwest and Northeast alone. In fact, only 10 cities accounted for 75% of the
rise in ghetto poverty throughout the 1970's. Even more startling was that nearly one
third was accounted for by New York City alone, and nearly one half by New York City
and Chicago combined (Jargowsky and Bane 1991).
Rosenbaum and Popkin placed low income Blacks in suburban areas and rented
apartments for them, while placing a control group in apartments in the city. Controlling
for personal characteristics such as family background/circumstances, motivation, as well
as education after beginning the program, the authors found that those who had been
placed in the suburban apartments were significantly more likely to find employment
than their inner city placed counterparts, citing more employment opportunities
(Rosenbaum and Popkin 1991).
Testing Wilson' s hypothesis, Sampson (1987) notes that the effect of black male
joblessness on black crime rates is mediated through its effects on family disruption. That
is, as the number of unemployed black males increases, this in turn increases the number
of female-headed households in these areas. He concludes that high black crime rates
appear to be caused by a combination of structural factors including unemployment,
economic deprivation, and family disruption. Sampson also finds support for Wilson' s
general thesis that structural unemployment is a more important factor in accounting for
Black family disruption as compared to White family disruption. These findings are
consistent with other works as well (Sampson 1987).
More specifically, the implications of Wilson' s work lie more in community level
explanations of crime. For instance, Sampson and his colleagues have found continued
support for his ideas regarding the underclass. In their study of neighborhood violent
crime and collective efficacy, it is clear how well Wilson's arguments mesh into this
scheme. Similar to Wilson's arguments regarding concentration of poverty, collective
efficacy demonstrates those conceptualizations well. For instance Sampson et al. (1997)
argue that low SES and concentrated disadvantage lead to a decrease in collective
efficacy, thereby leading to increased neighborhood crime, which clearly is in line with
Wilson's arguments as well as social disorganization theory in general (Sampson et al.
Researchers have explored these concepts utilizing indices intended to represent the
concentration of poverty among inner city residents. They have found that measures such
as male joblessness and racial economic inequality have significant affects on urban
crime. For example Krivo and Peterson (2000) find that concentrated disadvantage is a
significant predictor of race-specific homicides. However they also note that if blacks and
whites held similar economic positions in society and similar levels of disadvantage, the
impacts of theoretically relevant measures would have more equal impacts on crime rates
for both groups (Krivo and Peterson 2000:557).
While much of Wilson's thesis rests on the concept of the changing urban
landscape and the increase of j obless males, some research has shown that the
relationship between male j oblessness and crime may be more complex. It has been noted
that while male j oblessness is an important factor in the concentration of poverty in urban
areas, racial residential segregation may be key to understanding why this is so, and is
addressed in the following section (Krivo et al. 1998).
Poverty and Segregation
Exacerbating the urban disadvantage felt among African-Americans in large cities,
Massey and his colleagues argue that residential racial segregation is the primary culprit
for the growing ghettos and underclass in America. In their book, American Apartheid,
Massey and Denton outline the history of segregation in the United States, addressing
each step and offer empirical evidence to back their claims. As Wilson notes, poverty
rates grew rapidly during the 1970's, but more importantly, poverty became more
geographically concentrated (Wilson 1987; Bane and Jargowsky 1988; Massey and
Much of how Massey views Wilson's contentions can be summed up by the
following quote from Massey:
I agree with Wilson's main argument- that poverty concentration has increased in
U.S. cities, with pernicious consequences for minorities. I disagree, however, with
his hypothesis that this transformation was brought about by the exodus of middle-
class minority members from the ghetto and with his argument that industrial
restructuring, in and of itself, was responsible for concentrating urban poverty.
While these processes may have exacerbated poverty concentration, neither was
necessary for its creation. In the absence of racial segregation, the economic
dislocation of the 1970's would not have produced concentrated poverty or led to
the emergence of a socially and spatially isolated underclass (Massey 1990: 330).
The three primary reasons that Massey disagrees with Wilson's claims include, 1)
racial segregation in urban areas remain high and show little sign of declining, 2) as
education and income go up, the degree of black segregation does not decline, and 3)
although the degree of segregation between affluent Blacks and poor Blacks has
increased slightly during the 1970's, it is still lower than the difference between the
affluent and poor of other minority groups (Massey and Eggers 1990).
They go on to suggest that "instead of being caused by the departure of middle-
class blacks from the ghetto, however, these developments are explained statistically by a
strong interaction between the level of segregation and changes in the structure of the
income distribution". They go on to suggest that groups that experience high rates of
poverty in addition to high rates of segregation will experience the greatest levels of
disadvantage. Furthermore, concentration of poverty rose the most in urban areas that
suffered economic downturn and suffered high levels of racial segregation-such as New
York and Chicago (Massey 1990; Massey and Denton 1993).
Massey and his colleagues use hypothetical cities to illustrate how segregation
works with other factors. They argue that when racial segregation is imposed, some
Whites are better off, yet all Blacks are worse off. In addition, the problems inherent with
segregation compound as the level of segregation increases, thus as segregation increases,
so does the concentration of poverty as the area that minorities occupy becomes smaller
and smaller. Likewise as the concentration of poverty grows for the minority population,
it declines for the White population (Massey 1990).
Additionally, Massey and Denton argue that when class segregation is included in
the equation, the effects are compounded further. That is, when one considers the
differential impact of both race and class segregation, it heightens the impoverishment
felt by Blacks and actually improves the qualities for Whites. So, when racial segregation
exists in a class-segregated society, any economic "shock" (loss of manufacturing j obs,
etc.) has a profound impact on that group and in turn increases not only the overall
poverty rate, but also the concentration of poverty. This is essentially in line with
Wilson' s claims, notwithstanding the segregation quotient of course (Massey 1990;
Massey and Denton 1993).
In speaking of support for Wilson' s contentions, it is hard to separate them from
Massey et al.'s arguments. They are so intertwined that both are extremely valuable
contributions, and both receive a great deal of rigorous empirical testing. Typically,
neither Wilson's nor Massey et al.'s arguments are tested alone. That is, they are more
often incorporated into tests of social disorganization theory and general structural
analyses of crime.
Although many of the examples put forth by Massey et al. (1990, 1993) involve
hypothetical data, other researchers have found support for their conceptualization of
racial segregation. Massey and his colleagues measure segregation using the index of
dissimilarity, which represents the proportion of the minority population that would have
to change residence locations in order to achieve an even settlement pattern in any given
geographical area. This conceptualization has successfully been used in subsequent
macro-level analyses of crime for some time (Krivo et al. 1998; Massey 1990; Massey
and Denton 1993).
In his research on infant mortality rates in U.S. cities, LaVeist found that residential
segregation was the strongest predictor of Black infant mortality rates. Furthermore he
noted that while racial segregation increases mortality among Blacks, it decreases it
among Whites, providing further support for the claim that segregation not only hurts
Blacks but also benefits Whites (LaVeist 1989).
Other researchers have noted a dynamic relationship between racial segregation,
Black socioeconomic status (SES), and discrimination. They found that rising racial
segregation was positively correlated to heightened differences in Black-White
occupational differences, which as a result increased the levels of Black-White
segregation via negative correlations to Black income. Moreover, the decreasing Black
SES led to an increase in housing market discrimination, thus resulting in more
segregation. Obviously this is a complex relationship, but it clearly demonstrates the
power of this conceptual model of how racial segregation influences as well as is
influenced by many forces (Galster and Keeney 1988).
Several researchers have used racial segregation to study other topics such as
housing preferences and educational disadvantages. Supporting Massey et al.'s claim that
segregation leads not only to disadvantages for Blacks but also specific advantages to
Whites, Roscigno (1998) finds that racial segregation is significant in predicting students'
academic achievement. His results indicate that attending a Black segregated school
(>75% Black) is indicative of performing more poorly, while attending a White
segregated school (>95% White) is associated with better academic performance
Racial segregation has also proved to be a significant predictor of homicides.
Peterson and Krivo examined African-American homicide victimization for many large
U.S. cities, and the impact of segregation on the homicide rates. They found that with the
exception of intra family homicides, racial segregation was a significant predictor of both
acquaintance and stranger homicide cases. This coincides with the idea that as
segregation increases, social isolation also increases, which may explain the lack of
segregation predicting intra family homicide rates (Peterson and Krivo 1993).
While both the concentrated disadvantage and social disorganization perspectives
view their constructs as being racially invariant in their impacts on crime rates (see
Sampson and Wilson 1995), thus far results have been mixed as to whether this is the
case. Shiahadeh and Ousey (1998) have noted that while a lack of low skilled jobs affects
both black and white homicides, other measures such as the percentage of renters or the
prevalence of high school dropouts may only affect black or white homicide rates
exclusively. Therefore, segregation may provide insight into why certain structural
measures are significant predictors of black but not white crime rates (or vice versa).
Krivo and Peterson (1996) have shown that while both Blacks and Whites experience
extreme disadvantage, violence rates in predominantly Black neighborhoods remains to
be significantly higher than for White neighborhoods (Krivo and Peterson 1996;
Shihadeh and Ousey 1998).
Providing further support, Krivo et al. (1998) have noted that not only does
segregation exacerbate the disadvantage experienced by Blacks in urban areas, it also
increases opportunities for socioeconomic status and employment among Whites. They
note that disadvantaged Blacks experience significantly more isolation from other groups,
thus increasing the likelihood that their interaction and contact is with other similarly
disadvantaged residents. Furthermore, they find that "j obless males are less likely to have
contact with people in high status occupations", which operates to curtail the upward
mobility we may expect providing residents are exposed to diverse role models and
experiences (Krivo et al. 1998: 74).
Similarly, Parker and Pruitt found support for racial segregation having a positive
influence on Black homicides. Interestingly, they find that this is the case only in the
West, and not in the South. They attribute this to early migration patterns, as well as the
West being known for having less economic inequality, poverty, and social isolation
(Parker and Pruitt 2000).
Racial Threat: Blalock's Power Threat Hypothesis
In his 1967 book Toward' a 7Jheory of2~inority-Group Relations, Hubert Blalock,
introduced a concept he referred to as "power threat" hypothesis, which he argued may
help social scientists better understand how and when different forms of social control are
used by the maj ority against the minority. Over the past three decades, his propositions
have received much attention, however the results are mixed as to whether empirical
Generally, Blalock hypothesizes that as the size of a minority population grows,
those in power will sense a threat to their power. Upon the arrival of such a threat, he
argues that those in power will implement specific mechanisms of social control in order
to curtail the ensuing threat (Blalock 1967).
Specifically, Blalock argues for two distinct, uniquely behaving hypotheses
surrounding power threat. The first is the "political threat" hypothesis, which posits that
as the Black population increases over time, those in power (Whites) will perceive some
threat to their political stature and therefore utilize formal social control mechanisms
(arrest and punishment) to maintain their power. He goes on to argue that threats of this
nature appear in the form of a positive curvilinear relationship between an increasing
minority population size and increased discrimination in response to power threat, with
an increasing slope (Blalock 1967).
The second hypothesis Blalock introduces is that of the "economic threat"
hypothesis. He states that as competition over economic resources, such as job
availability, increases between Blacks and Whites, Whites will again utilize social control
mechanisms (arrest and punishment) to the advantage of those in power. He posits that
there will be a positive curvilinear relationship between discrimination resulting from
economic competition, and the slope will decrease (Blalock 1967).
Blalock also recognized the difficulty that his assertions posed in terms of
measurement and conceptualization. He states that:
Empirically determined relationships should then be a composite of these two
different forms. In many instances, such compositions will be approximately linear
in nature, so that adequate tests will require one to locate relatively "pure" instances
in which motives can be linked to behavior in a one-to-one fashion (Blalock 1967:
Clearly, the level of abstraction in his ideas creates a challenge in not only
measuring these concepts, but conceptualizing them to a specific problem or scenario.
His propositions have received support in explaining a wide variety of issues related to
race and crime. These include lynching (Reed 1972; Corzine et al. 1983, 1989), arrest and
incarceration rates (Liska and Chamlin 1984; Liska 1992; Myers 1990; Crawford,
Chiricos, and Kleck 1998; Parker et al. 2005), as well as interracial killing (Jacobs and
John Reed applied Blalock' s ideas to lynching rates in Mississippi, and measured
the proportion of the Black population to assess its impact on the rates. Reed found
support for the thesis in that the more concentrated the Black population, the higher the
lynching rate (considered informal social control). However, his study would later come
to be criticized for its limited scope and lack of controlling for external factors (Reed
1972; Corzine et al. 1983).
A similar study was performed to further Reed's findings two decades later.
Corzine, Creech, and Corzine (1983) also used lynching rates in the South to test
Blalock' s hypothesis. Using data obtained from the NAACP on lynching rates from
1889-1931, in the eleven confederate states, Corzine et al. set out to unravel the entangled
relationship between the impact that percent Black had shown in past research to have on
crime in the South. They used Reed's lynching index, which he introduced in his
assessment in 1972 (Corzine, Creech, and Corzine 1983).
Their Eindings offer limited support to Blalock' s thesis, however they also illustrate
that percent Black effects were highly correlated with the specific geographic area. They
then separated two sub-regions of "deep South" and "bordering South", finding that the
power threat explanation fits better in the deep South region as compared to the bordering
South (Corzine, Creech, and Corzine 1983).
As promising as these two studies were for the prospects of the "power threat"
hypothesis, they would both come under heavy scrutiny in the late 1980's. With the
publication of a criticism of the two studies by Tolnay, Beck, and Massey, a heated
exchange occurred regarding many of the conceptual and measurement issues of both
studies. Tolnay et al. (1989) re-examined the data used in the Corzine et al. (1983) study.
They pointed to several discrepancies which existed in their data and measurements as
well as their analysis and subsequent interpretations (Tolnay, Beck and Massey 1989).
Their criticisms covered several areas, but perhaps their greatest concern was with
the deletion by Corzine et al. of counties with less than 5% total Black population, and
the inclusion of counties with greater than 80% Black of the total population. They argue
that three counties in particular, with high concentrations of Blacks in the population and
a unique structure of lynching during this time skew the results presented by Corzine et
al. They offer an analysis based on approximate data that closely resembled that used by
the Corzine et al. study (from the same data bank at NAACP), and they report to find
substantially different results. In addition to these data shortcomings, they point to the
inconsistency and inaccuracy of the NAACP data, in that they found several instances
where the wrong case was recorded, etc. (Tolnay, Beck and Massey 1989).
In addition to these criticisms, they also take issue with Reed's lynching index.
Relating back to the three outlying counties mentioned above, they argue that these
counties exacerbate Reed's index and can potentially inflate the findings that percent
Black is a significant predictor of lynching rates in the South. This is due in part to the
fact that two of the counties that were outliers in the Corzine et al. study, also resided in
the data utilized by Reed in his original study (Tolnay, Beck, and Massey 1989).
Additionally, they argue that Reed' s index doesn't represent anything that a rate
should. That is it (admittedly by Reed) doesn't represent the rate of lynching compared to
the population, but rather it represents the rate of lynching, as compared to the expected
rate of lynching. This, they contend, inevitably causes Reed's index to automatically
assume the slope predicted by Blalock, and that it is unable to differentiate between two
scenarios (Tolnay, Beck, and Massey 1989).
These criticisms prompted commentary from Corzine et al., Reed, as well as
Hubert Blalock. The response by Creech et al. to claims of "trncating data" by
eliminating certain counties, while keeping others with no conceptual or theoretical
justification, is simply that they had theoretical justifications for doing so due to the
scope of the power threat hypothesis. That is, the power threat hypothesis is not a general
theory of crime, criminal behavior, or discrimination. Instead it is a fairly narrow and
specific hypothesis predicting when discrimination is most likely to increase based on the
proportion of the minority population (Creech, Corzine, and Huff-Corzine 1989).
In response to the critics' claims that the Reed index was inappropriate, Creech et
al. admit that the index is not perfect (what index is?), and that they had considered others
at the time and settled on Reed's as it best captured the measure for testing Blalock' s
hypothesis. They argue that the alternative measure provided by Tolnay et al. ignores the
racial composition of the county, thus making it difficult to discuss any implications in
terms of power threat hypothesis (Creech et al. 1989).
John Reed's response is much less detailed. He does however declare his
disappointment with the data collected by the NAACP. However, in his defense of the
data, he argues, and rightly so, that data to test Blalock' s hypotheses rely on areas with
heavy concentrations of Blacks where lynching had occurred. If the data that were
available had not been used, there simply would not have been a test of the hypothesis,
because due to the nature of the hypothesis, Mississippi data was crucial to even having
the opportunity to apply Blalock to lynching rates in the South. He also points out a
quotation by Blalock regarding his original hypotheses and the difficulty in testing them.
He states: "inadequate data, the infrequency of lynching, and certain methodological
difficulties have prevented [us] from obtaining definitive results" (Reed 1989).
The response by Blalock is largely methodological, yet brief. He offers many
suggestions if choosing different denominators for creating ratios and measures of such
items. He ends his short reply with saying that: "In this instance we are involved with a
cross-level analysis problem, where the theory refers to the micro-level (perceived threats
and loosely coordinated lynching behaviors), whereas data have been aggregated at the
macro-level." This presents a problem similar to that of the ecological fallacy, he goes on
to state that when testing these ideas further, caution must be taken to carefully construct
a theoretical model that clearly specifies the assumptions in order to select a suitable
denominator for creating ratios (Blalock 1989:633).
Crawford, Chiricos, and Kleck use "racial threat" as a means to better understand
sentencing under Florida' s habitual offender laws, based on race. They conceptualize
racial threat as threatening to "mainstream America" as well as "political elites". They
argue that racial threat has become increasingly prevalent, especially with the media
frenzies depicting juvenile violence as "ghetto pathologies" as well as the images of
crack cocaine potentially spreading to previously "safe places" (Crawford, Chiricos, and
They utilize data from Florida during the period of 1992-1993, involving
sentencing of "habitual" offenders. The data file consisted of 9,690 "habitual eligible"
offenders to determine to what extent race contributed to being sentenced as such. In
order to single out race effects, they controlled for demographic characteristics of the
offender, legal attributes, as well as the county in which they were sentenced.
One limitation that they state up front is their lack of data on race of victim, which
is used in much of the racial threat literature. The most interesting finding perhaps in light
of racial threat is that of drug offenders. They argue that a racial threat argument may
help to explain the strong effect of race in the habitualization of offenders. Nearly all of
these were cocaine offenders, which is precisely what the media frenzy in the 1980's
focused on almost exclusively regarding drugs and crime. These findings were similar for
both drug possession as well as dealing/trafficking offenses (Crawford et al. 1998).
An interesting paradox arises in their research relating to racial threat. Where the
race variables showed the most strength were also the areas that would traditionally be
viewed as low in racial threat. That is, they were in communities below the median in
percent Black, racial income inequality, and so on. This seems at odds with racial threat
on the one hand, yet on the other, it may contribute to our understanding of racial threat
at the same time. While traditionally it was argued that racial threat occurred primarily in
the areas of highest concentration, it could be that these areas deemed "safe" and free of
crime and drugs, are even more punitive (especially considering the images in the media
portraying crack coming to a neighborhood near you) (Crawford et al. 1998).
Eitle, D'Alessio, and Stolzenberg (2002) propose three separate hypotheses. They
indicate a need to look at political racial threat, economic racial threat, as well as Black
crime racial threat. Using NIBRS data from South Carolina, they measure political racial
threat as the ratio of Black to White voters during the 1992 and 1994 elections in South
Carolina, economic threat as White to Black unemployment, and Black crime threat as
the influence of Black on White crime on Black arrest levels, where Black on White
crime is the percent of violent felonies reported to the police where the offender was
Black and the victim was White (Eitle et al. 2002).
Controlling for other contextual factors such as White/Black divorce ratios, violent
crimes in area, and population density, they found no support for either political threat
nor the economic threat hypotheses. However, they do Eind strong support for Black
crime threat. Furthermore, in counties with a higher percentage of Black on White crime,
Blacks are significantly more likely to be arrested. This finding was robust, even in the
model containing all of the control variables (Eitle et al. 2002).
The power threat hypothesis has evolved substantially since its early
conceptualizations by Blalock. Since its earliest tests by Reed and Corzine et al., the
perspective has been improved upon greatly. Many taking note of the methodological
issues associated with the early studies, clearly has had some impact. Eitle et al. stated
that their goals were to better synthesize the conceptualizations as to limit confusion,
since it' s often referred in several manners.
Through the use of NBRS data, Eitle and colleagues may have discovered a
unique and accurate way of measuring and testing the threat hypothesis, without
distorting or having to stretch too far the measurement issues that have surrounded this
approach in past studies (see also Stolzenberg et al. 2004). Clearly, having data on race of
both offenders, as well as victims, provides a dynamic way of conceptualizing racial
threat. The Crawford et al. study offers another unique direction in that those areas that
perceive great racial threat may in fact be those with the least concentration of Blacks,
and of crime in general (e.g., middle class suburbs, where the images of "crack babies"
More recent studies of racial threat have attempted to move beyond the myopic
measures of percent black of the population. More recent tests of Blalock' s hypotheses us
other measures such as racial inequality and black immigration patters as proxies of racial
threat (Stolzenberg et al. 2004; Parker et al. 2005). Findings indicate that contrary to
Blalock' s hypothesis, percent of the population that is black and racial inequality were
found to have a negative affect on black arrest rates. One explanation that has been
asserted is the "benign-neglect hypothesis" by Liska and Chamlin (1984). They argue
that in areas with a large black population, crime is more likely to be intra-racial. As a
result, there is less pressure on police to control crime because non-white victims are less
likely to report crime or, even when they do report crime, police may allocate fewer
resources to resolve the offense (Liska and Chamlin 1984).
In light of the preceding discussion of the prevailing macro-level theoretical
perspectives on crime and delinquency, the ways in which these theories have been and
will continue to be tested has not been consistent to say the least. For the purposes of this
study, it is better to explicitly state that I am attempting to account for the rise in race-
specific drug sales arrest rates over time, using structural factors of large urban areas.
That is, I am not necessarily setting out to explicitly "test" social disorganization theory.
To do so would require conducting large scale interviews, surveys and focus groups in
large urban areas to truly assess to what extent these urban areas are able to exercise
informal controls over its residents.
Rather, what I am attempting to do is to choose theoretically relevant structural
predictors as proxies for the larger concepts of social disorganization and other ecological
theories of crime. This is important as many of the frequently utilized measures of
structural factors as they relate to crime and delinquency are "shared" across theoretical
paradigms. For example Pratt and Cullen (2005) point out that:
Researchers have used measures of unemployment as proxies of guardianship
(routine activity theory) and economic hardship (economic/resource deprivation
theory), as an indicator in the breakdown in the viability of community
control/socialization (social disorganization theory), and even as a precursor to
frustration-induced anger (anomie/strain theory) (Pratt and Cullen 2005: 430-431).
My reasoning for choosing these three theories specifically is twofold. First an
analysis such as this has not yet been conducted, so it adds a significant contribution to
the research examining these factors on changing drug arrest rates over time. Second
these three theories not only allow for race-specific changes over time, but as Pratt and
Cullen note they are among the most consistently significant predictors of crime in urban
areas. As Messner et al. (2005) have suggested, the changes that cities in regards to both
drug arrests as well as homicides has changed significantly over the past several decades.
Additionally Levitt (2001) suggests that the declining crime rate during the 1990's is
partially attributable to the decline in crack markets in large cities. While these
approaches note the importance of accounting for changes over time, Levitt in particular
considers drug crimes to be an independent (explanatory) variable of crime rates.
Therefore the changes in urban areas during this same time period creates a dynamic
relationship where the changes in theoretically relevant macro-level indicators may be
related to the changes in drug sales arrest rates over time (Levitt 2004; Messner et al.
This is especially true for African-Americans in large urban areas. As Wilson
(1987) has argued that black families have been especially impacted by these industrial
shifts. The theories selected for this study allow not only for change in general, but
especially allow for modeling change for blacks specifically, across multiple decades for
multiple cities. While the discrepancies in Black and White crime rates have been well
documented in past research, no other crime does this difference stand out more so than
for drug crimes.
REVIEW OF THE LITERATURE
While variations of the structural theories presented above have received
considerable empirical testing, little attention has been paid to addressing structural
explanations of drug offenses. These theories have all been used to help us explain both
violent and property crimes (Crutchfield and Pitchford 1997; Liska et al. 1998; Krivo and
Peterson 1996, 1998, 2000; Parker and McCall 1997, 1999; Sampson 1987 Shihadeh and
Ousey 1998), and have only recently began to garner the attention of researchers
interested in drug crimes in large urban areas (see Mosher 2001; Parker and Maggard
2005). This chapter seeks to offer a historical overview of drug use and the laws created
to control their use in the United States. It is important to understand the historical
connections between drug use, public policy and crime in order to better apprehend why
our country is in the position that it is today. Finally, I will offer a brief discussion of the
merging of the theories and how tests of each theoretical construct are not mutually
exclusive from the other theories as well as provide several hypotheses to be tested.
Race, Drugs and Crime: A Historical Perspective
Any discussion of drugs and crime needs to account for the historical landscape
that surrounds drug use in this county as well as the way that we have chosen to deal with
these issues. We must ask ourselves, how did we get here? Here referring to the fact that
in 2004 there were an estimated 1, 745, 712 drug violation arrests in the United States
(U. S. Department of Justice 2004). While ordinances and legislation dates back more
than 100 years, we continue to find ourselves arresting millions of individuals year after
year. We must also take into account the roles that the media, doctors, minority groups,
as well as the pharmaceutical industry play in this expensive game that is illicit drug use.
Drug use in this country is best understood beginning in the 1800's. Opiates were
used in medicine in the early years and while addictive, offered medical uses that suited a
variety of needs. It was the discovery of heroin during this period that prompted doctors
to believe they had found an iteration of opiates that was nonaddictive and yielded few
side effects. This belief lasted only about five years, before it became apparent that heroin
did in fact possess the same addictive properties of other opiates (Conrad and Schneider
1980). Heroin and opiates were later characterized as recreational drugs, yet
pharmaceutical companies continued including opiates in a variety of medications,
despite the risks of addiction, among others. Although the drug industry continued to use
opiates in these formulations, the press was reluctant to criticize the pharmaceutical
companies for fear of losing advertising dollars and future revenues. Instead, the media
portrayed users of opiates in a negative manner, since criticizing these companies had
financial risks; the media began to portray users of opiates in a derogatory manner, for
"powerless groups could be criticized without such caution" (Tieman 1981: 242).
The power of the media may never be quite as transparent as is realized while
peering through the sociological lens at American drug use. During this same period
cocaine use was beginning to be attributed to African Americans in particular. The media
presented an image of "cocaine-crazed" black men who upon the ingestion of cocaine
experienced greatly increased sex drives and desires. These desires were said to increase
their likelihood of raping white women, which instilled fear among citizens. Further, it
was reported that as African Americans used cocaine, they experienced greater visual
clarity, which significantly increased their marksmanship with firearms. Perhaps the most
outrageous media reports rest on the assertions that cocaine made black men less
vulnerable to .32 caliber bullets. While this seems outrageous today, police agencies
actually began switching to .38 caliber handguns to combat the dangerous cocaine crazed
black men (Musto 1973: 7-8; McBride 1981: 105-124; Smith 1986). While these
derogatory portrayals of African Americans committing crimes while under the influence
of cocaine are preposterous, more utilitarian uses of the drug were also circulated in the
media. For instance, the M~edical News reported that its use in New Orleans made it
possible for workers to work seventy hour shifts loading and unloading steamboats, that
otherwise may not have been able to be accomplished without the use of stimulants
African Americans were not the only minority group to catch the attention of the
media and the government. During the late 1800's Chinese immigrants, many of whom
were working on the railroad systems found employment in California among other
locales. Their opium smoking behavior led to an 1875 ordinance passed in San Francisco
"aimed at the un-American habits of the 'yellow devils'" (Brecher 1972: 42-44). These
negative portrayals are not unlike other "dangerous classes" such as prostitutes and
vagrants (see Adler 1986 and 1992).
This stigma emerged to stick not only to minorities but also to whites that had
adopted opium-smoking habits (Lindesmith 1968: 215). These casual opium users were
described as "gamblers, thieves and prostitutes", who were not legitimately receiving
medicine from doctors. These images of casual drug use eventually led to the increasing
restrictions placed on opiates which coincided with the 1) rise in patriotism in America,
2) an increasing fear of minorities and 3) what had become an estimated addict
population of 250,000 (Musto 1973: 5, 33).
This was a critical time in American history. While we wished to decrease opium
use among residents and immigrants alike, our immigration doors were growing as we
expanded and acquired new territories. It was important for the United States to remain
vigilant in its abhorrence for drug use with the Chinese government and it was quickly
realized that we would not be able to persuade other nations to curb drug use if we did
not enact specific legislation at home (Duster 1970: 14; Musto 1973: 36).
As one of the first measures to control drug use at the federal level, the Harrison
Act, passed by Congress in 1914 paved the way for the future of drug control in the
United States (Musto 1973; Spillane 2004). While supported by many, the Harrison Act
was not free of criticisms by supporters and critics alike, including then Speaker of the
House Oscar W. Underwood "...who had shepherded the Harrison Act through the house
the previous week, described the prohibition amendment as a 'tyrannous scheme to
establish virtue and morality by law"' (Musto 1973: 67). Sentiment such as Mr.
Underwood's notwithstanding, the next four decades would see attention focused on
enforcement and drug users were characterized as antisocial or psychopathical addicts
who chose addiction that "normal" people would never choose (Tieman 1981).
This trend continued for much of the coming century with the next maj or event
being the formation of the Federal Bureau of Narcotics (what is now the Drug
Enforcement Administration) in 1930 and the appointment of its over-zealous leader
Harry J. Anslinger. Anslinger abhorred drug use and demanded that his agents focus on
street peddlers of drugs in American cities (Tieman 1981; Carroll 2004). Anslinger was
able to persuade citizens of the United States that drug sellers and users alike were
criminals, and should be treated as such, and from the 1930's forward, drug use was
strictly a law enforcement issue versus a public health issue. Anslinger and his
subordinates were able to perpetuate the idea of strong linkages between drug use and
violent crime through telling fabricated "horror stories" (Carroll 2004).
While Anslinger directed most of the Federal Bureau of Narcotics' attention on
"hard drugs" (opiates and cocaine), he paid little attention to marihuana and did not
consider it a serious problem, at least not one that required the intervention of the federal
government (Carroll 2004). Anslinger even testified to Congress that marihuana did not
cause addiction and could be handled at the local level if problems persisted with the
substance. However Anslinger' s true political identity would emerge as rumors began
circulating on Capitol Hill that a new bill could dissolve the FBN and shift enforcement
of drug laws to other government agencies. At this point, marihuana was the obvious
vehicle he needed to instill hysteria and leverage the idea that marihuana was a "menace"
that needed to be dealt with and this eventually led to the Marihuana Tax Act of 1937.
Anslinger even began referring to marihuana as hashish (an extract from female
marihuana flowers), since it seemed to be so closely associated with the word assassin
(Musto 1973; Carroll 2004).
Anslinger' s role in shaping federal drug control policy would remain steady for
many years to come (Tieman 1981; Carroll 2004). While little scientific evidence was
presented linking marihuana use and drug use to criminal behavior, Anslinger offered
many "commonsensical" accounts by which addicts would resort to crime (Galliher and
Walker 1977:375). Until the end of the 1940's, addicts remained relatively uninvolved in
criminal behavior, as costs for drugs were low and many were able to use physicians for
their personal needs (O'Donnell 1969).
Perhaps the "line in the sand" drawn to make clear that drug use would be a
criminal justice issue versus a public health issue occurred through the Boggs
Amendment of 1951 and the subsequent Narcotic Control Act in 1956. These laws would
establish mandatory minimum sentences for drug offenses and were meant to emphasize
the seriousness of such offenses and how law enforcement would address them.
Anslinger was able to re-emphasize his beliefs that addicts were merely "parasites" as he
The person is generally a criminal or on the road to criminality before he becomes
addicted. Once addicted he has the greatest reason in the world for continuing his
life of crime (Anslinger and Tompkins 1953:170).
It was not monetary reasons that Anslinger argued would lead to the criminal career, but
rather that using drugs would erode the "moral fiber" of individuals in general. This
pattern of connecting drug use to criminal behavior continues to this day (Tieman 1981).
Anslinger' s tactic of demonizing marihuana to pave the way toward criminalizing
its users was carried forward to future eras and specific drugs over time. Indeed the media
would also play an important role in shaping the legislation that led to the "drug scares"
of LSD in the 1960's, PCP in the 1970's, crack in the 1980's, and MDMA or Ecstasy in
the 1990's and 2000 (Goode 2005). Were many of the horror stories true regarding the
toxicity of drugs such as PCP? Sure they were. PCP is a dangerous drug, but what the
media can contribute to is the misrepresentation of the extent of a drug' s use. As with
other "hard drugs", most users are not casual marihuana users one day who decide to
become hard-core drug users the next, but rather they have already become enveloped in
a hard-core drug scene to begin with.
Race, Drugs and Crime: The New Era
While the focus of my research is on drug sales arrests from 1980-2001, it was
important to point out how the shifts and oscillations in America's responses to drug use
have affected the very arrest rates that I am investigating. With each of the
aforementioned "drug scares", mass hysteria and over reaction has accompanied virtually
each one (Reinarman and Levine 1997). The onset of crack in American society was a
maj or event, and politicians were able to use this "drug scare" as an avenue to create
specific laws aimed at controlling its use, not unlike the Harrison Act in 1914 and other
drug legislation discussed above.
In 1982, President Ronald Reagan officially declared a "War On Drugs" (Sacher
1997). Shortly after this, Congress enacted the Sentencing Reform Act of 1984 (SRA),
which reportedly had the sole intent of establishing strict guidelines for federal
sentencing. This was an effort to promote fairness and uniformity in federal cases by
establishing mandatory minimum sentences for specific offenses. Following the 1984
SRA bill, the Anti Drug Abuse Acts of 1986 and 1988 substantially altered the way in
which we would come to treat drug offenders brought to federal court on charges of both
possession and trafficking of cocaine (Sacher 1997; Zedeck 2000).
During the period that I am focusing on, crack cocaine is the most salient drug
related event that occurred which resulted in unprecedented drug arrests in the United
States. Drug arrests rose from 18,521 to 88,641 respectively from 1980 to 1988 in New
York City alone (Belenko et al. 1991; New York City Police Department 1989). The
importance of studying arrest rates disaggregated by race are clear as Blumstein (1993)
notes that although White arrest rates remained relatively stable from the 1970's through
much of the 1980's, non-white drug arrests steadily climbed from 1980-1985, growing
exponentially at a rate of 15%-20% per year until peaking in 1989 (Blumstein 1993).
A similar portrait was seen at the national level where the number of drug arrests
in 1980 was just under 600,000 and by 1990 had risen to roughly 1.1 million total drug
arrests (representing nearly a 92% increase over a 10 year period). More striking however
are the racial shifts that occurred during this period. In 1980 African-Americans made up
about 25% of all drug arrests in the United States and by 1990 this figure had increased to
41% of all drug arrests (Goode 2005; U. S. Department of Justice 1996).
While it has been argued that the rising drug arrests during the 1980's was directly
attributable to crack cocaine, it is important to look at who was using crack cocaine and
the potential size of those markets. The demographic characteristics of crack users
sharply differ from their powder cocaine-snorting counterparts. Those who abuse crack
cocaine, much like daily heroin users, are likely to be urban poor. Moreover, they are
likely to have been hard-core drug users in the past, for most crack users do not start off
with crack cocaine. Rather, they are either hard-core heroin addicts or cocaine abusers,
looking for a better, more convenient high. It is important to note that even during its
most popular year, only a small percentage of those who used cocaine, used crack (Golub
and Johnson 1996; Reinarman and Levine 1997).
While crack cocaine is primarily prepared and distributed by African Americans, it
is also used and abused by Whites at even higher numbers. Although data do suggest that
a higher proportion of the Black population use the drug when compared to Whites.
However, though Whites make up the maj ority of crack users, Blacks continue to make
up more than 80 percent of those convicted for distribution of the drug (Gray 1998;
Each time a relatively unknown drug gains popularity, there are undoubtedly
outrageous exaggerations of its use and it becomes the next "scary drug of the year"
(Akers 1990). Typically these reactions are also accompanied to knee jerk hypotheses
regarding the new drug and its relationship to violence and crime in general (Goode
2004). In fact, one could certainly argue that as new drugs begin to garner the attention of
the mass media, their very role is "socially constructing" how the drug will come to be
viewed (Hartman and Golub 1999). The role of language and the mass media in shaping
attitudes toward both people as well as inanimate obj ects is often underestimated (see
also Berger and Luckman 1966).
One of the more prominent theories in regards to drugs and crime/violence is Paul
Goldstein' s tri-partite framework. His conceptual model of the relationship between
drugs and crime/violence consists of three mechanisms by which drugs and crime are
related. First, the psychopharnacological he argues is crime and violence that results
from the actual physical effects of the drug on the user. For example, taking a certain
drug may cause a person to lack judgment which may lead to crime and/or violence.
What he calls the economic compulsive mechanism is when individuals must commit
crimes in order to finance expensive drug habits. Finally, systemic crime/violence is the
result of any number of circumstances related to drug use and drug markets. For example
the black markets that drugs are typically peddled in are often territorial and controlled by
gangs which may lead to violence between rival gangs. In another example offered by
Goldstein and his colleagues, individuals may "punish" someone who sells them "bogus"
drugs in order to teach them a lesson (Goldstein 1985; Goldstein et al. 1991).
Goldstein et al. tested these concepts and found that alcohol was the substance most
correlated to the psychopharnacological model. In terms of the systemic model, heroin
markets were the most likely culprit in determining crime and violence for the areas in
which they studied. Finally cocaine was the strongest substance related to the economic
compulsive mechanism, however the numbers were quite small. Clearly, using the
reasoning that was used to create the new laws aimed at crack cocaine, one would expect
crack and/or cocaine in general to dominate all three categories in research such as this
(Goldstein et al. 1991).
History has shown us that officials often overreact to the introduction of new drugs,
especially when they are primarily associated with minorities. However, in line with
Goldstein's ideas, Blumstein (1995) has argued that drug markets such as the crack
cocaine market greatly contributed to the diffusion of firearms into the hands of youth. It
is further argued that the volatility of crack markets, coupled with the lower inhibitions of
inner city youths would lead to the increase of gun crimes and homicides within
American cities. While not immediate, Blumstein argued that as crack markets spread
across the country, a short lag would occur and we would Eind that homicides and gun
related violence would follow in those urban areas (Blumstein 1995).
In support of Blumstein's hypotheses, crack and cocaine have been linked to
different types of crime in large cities. Baumer (1994) found that cocaine use by arrestees
was a significant predictor of robbery rates and a more modest predictor of homicide
rates in large cities, even while controlling for structural characteristics such as poverty
and divorce rates. Of course whether the correlation is due to the dynamics of the drug
market or the drug itself is unclear. Similarly Ousey and Lee (2004) found that changing
drug markets are linked to the rise in both White and Black homicide rates in large cities.
However the relationship for is quite larger for Blacks than Whites, with structural factors
such as inequality having a much stronger impact on Black homicide rates than for White
homicide rates (Baumer 1994; Ousey and Lee 2004).
In another study, Cork (1999) also found support for Blumstein' s theory. More
specifically he found that the relationship between crack markets and their linkage to the
availability of guns is quite realistic. While the relationship is dynamic and complex, he
also notes that the increases in gun related homicides committed by juveniles, typically
occurred shortly after the emergence of crack markets in those areas. Also worth noting is
the fact that while slightly older homicide offenders diffused in a similar manner, they
were not as pronounced as with younger perpetrators, as youth have been said to lack the
empathy or concern for strangers as older individuals may (Cornell 1993; Cork 1999).
These studies are good examples of the way in which researchers have approached
the problem of drugs in the literature. It has been more common to use crack and cocaine,
or even heroin or alcohol as explanatory variables in determining crime rates in general.
That is, little research has focused solely on explaining the rates of drug arrests as much
as they have focused on property and violent arrest rates. Given the structural changes
across cities in America, as well as the increase in drug arrests, clearly this is worthy of
further research (see Mosher 2001; Parker and Maggard 2005).
Much of the research examining the crime drop that occurred across America in the
1990's has also noted the important role that crack cocaine markets in particular played in
this decline (Levitt 2004). This is interesting because while we do not want to unjustly
punish the behavior of one racial group compared to others, we also have an obligation to
protect the citizens in urban areas where much of these crimes are centered. It is these
reasons that we expect not only for drug arrests to have risen so sharply from 1980-2001,
but also expect Blacks to have incurred much more dramatic rises proportionally
compared to Whites.
Hypotheses and Expectations
Having reviewed prior theoretical research as well as reviewing the historical
aspects of drug use and arrests, I have generated several theoretically relevant hypotheses
and expectations in this study. The first set of hypotheses surrounds the trends in race
specific drug arrests over time while the second set of hypotheses relates specifically to
the structural factors discussed earlier and how they may be directly related to the trends
in drug arrests that appear in the data.
As noted earlier drug arrests have seen many fluctuations over the past century.
However no era has reflected as much change as the period beginning around 1980 and
extending to this day in some respects. The implications for race are even more
exaggerated, as African-Americans have been disproportionately targeted and arrested as
a result (Tonry 1995). Prior research on drug arrests trends have traditionally been at the
decennial level in regards to changes over time. Researchers have neglected to attempt to
group similarly behaving cities over time, as they relate to racially disaggregated drug
arrests. Therefore by applying the TRAJ procedure to such data, it is possible to model
whether particular cities and regions behaved as expected over these time periods. As
Blumstein and Rosenfeld (1998) have noted, the changes in homicide rates have differed
for "coastal cities" versus cities that are more inland, which they argue is consistent with
the patterns Golub and Johnson (1997) found to surround the crack epidemic (Blumstein
and Rosenfeld 1998; Golub and Johnson 1997). Given this past research on homicides
and the crack epidemic specifically, I draw the following two hypotheses in regards to the
expected traj ectories for drug sales arrests over time:
H1: I hypothesize that race specific drug sales arrest rates will significantly vary
over time, forming identifiable traj ectories distinguishing cities which experienced
many drug arrests from those that did not.
H2: The witnessed traj ectories will differ significantly by race, indicating that black
drug sales arrests accelerated at a much greater pace from 1980-2001 than did
white drug sales arrests.
These expectations arise from reviewing past work examining the relationship
between a city's structural characteristics and drug arrests over time. For example,
Mosher (2001) notes that the rise in drug arrest rates across cities is profound. This effect
is exacerbated by using the mean examine changes over time based on a pool of cities
that differ greatly in terms of arrests and structural characteristics. Parker and Maggard
(2005) report that black total drug arrests rose by about 211% (versus 106% increase
among whites) from 1980-1990 with similar patterns witnessed for possession arrests, yet
for drug sales arrests the racial gap narrowed with a 236% increase in black drug sales
arrests versus 193% increase for whites (Parker and Maggard 2005). Mosher' s finding
coupled with the differences noted by Parker and Maggard lead one to believe that it is
feasible that the mean increases witnessed in past research could have been inflated by
outliers, especially for drug sales arrests. In other words, most cities may not have
experienced "epidemics" of drug sales and use during key years, yet a few of these cities
may have experienced unprecedented problems surrounding drugs and crime which
resulted in highly skewed descriptive statistics. This is especially plausible given
Jargowsky and Bane' s findings that only 10% of the cities they studied accounted for
75% of the rise in ghetto poverty throughout the 1970's, indicating a few radically
changing cities may paint much of the picture of the nation for any given topic
(Jargowsky and Bane 1991).
The second set of hypotheses stem from theoretically drawn expectations of the
structural conditions of urban areas and their impact on fitting into specific traj ectories as
mentioned above. Since I draw upon multiple structural theories of urban crime in
constructing these hypotheses, some are not necessarily "theory specific". However the
directions that I predict clearly rest on particular theories and their expected relationships
to urban crime. For example, Pratt and Cullen (2005) note that the expected relationship
between unemployment and crime may be positive direction for social disorganization or
anomie based theories, while predicting an inverse relationship under the umbrella of
routine activities theory (Pratt and Cullen 2005).
Social Disorganization and Drug Sales Arrests
Based on Shaw and McKay's original hypotheses as well as past research, I expect
the rising disorganization of urban areas to result in both an increase in both white and
black drug sales arrests over time (attributable to the racially invariant predictions of
social disorganization theory). Recall that social disorganizations theory posits that the
disorganization of an area results in a breakdown in informal control mechanisms. This in
turn is expected to result in higher crime rates, regardless of who resides in an area. It is
this racially invariant nature of social disorganization theory that leads to the following
H3: Cities that experience an increase in family disruption (via the percentage of
divorced males) for both blacks and whites will increase the likelihood of those
cities belonging to a higher arrest traj ectory as opposed to the lowest arrest
traj ectory for drug sales arrests.
H4: Those cities that experience increases in residential mobility will increase the
likelihood of those cities belonging to a higher arrest traj ectory as opposed to the
lowest arrest trajectory for drug sales arrests (invariant across race).
Concentrated Urban Disadvantage and Drug Sales Arrests
Wilson (1987) argues that the deindustrialization of large urban areas has had an
unprecedented affect on the lives of African-American residents in those cities. Massey
and Denton have extended these ideas to include the role that racial segregation has in
concentrating poverty in particular areas. Research has shown that this is the case in
many instances for both violent and property crimes. However there is good reason to
believe that the decreasing availability of low-skilled jobs would lead to an increase in
drug sales arrests, especially for African-Americans. Elij ah Anderson' s ethnographic
research on urban areas does an excellent j ob describing how the illicit drug trade and
poverty are intertwined in a very complex way. Anderson writes:
In the impoverished inner-city neighborhood, the drug trade is everywhere, and it
becomes ever more difficult to separate the drug culture from the experience of
poverty. The neighborhood is sprinkled with crack dens located in abandoned
buildings or in someone's home. On corner after corner, young men peddle drugs
the way a newsboy peddles papers.
As Anderson illustrates the difficulty in separating urban poverty and the illicit drug
trade, the urban disadvantage literature presented above leads me to several hypotheses.
With black neighborhoods becoming increasingly segregated and the concentration of
poverty increasing, it should be no surprise that some individuals were coerced, so to
speak to sell illicit drugs. The lure of high profits, designer clothing, and the ability to
support one's family proved too tempting to pass by. Additionally the open-air nature of
crack markets left inner city drug sellers especially visible to law enforcement agencies.
These factors lead to the following two hypotheses:
HS: An increase in residential racial segregation (measured as the Index of
Dissimilarity) will increase the likelihood of belonging to a higher arrest traj ectory
as opposed to the lowest arrest traj ectory for black drug sales arrests.
H6: An increase in concentrated disadvantage (e.g., poverty, income inequality,
male j oblessness, and percent children not living with both parents) will increase
the probability of cities belonging to a higher arrest traj ectory versus the lowest
arrest traj ectory for both Black and White drug sales arrests.
Racial Threat and Drug Sales Arrests
Hubert Blalock' s power threat hypothesis states that as the minority population size
grows, those in the dominant group will begin to feel threatened and be more likely to
utilize formal social control mechanisms (e.g. arrests). Moreover as those in power begin
to utilize formal controls more liberally, arrest rates for the minority group should rise
over time. Drawing both on Blalock' s ideas, I am left with the following hypothesis:
H17: Cities that experience a significant rise in the proportion of the population
identifying themselves as black, will be more likely to belong to a higher traj ectory
group versus the lowest traj ectory group for Black drug sales arrest rates.
These variables are also among the list of strong structural predictors that Pratt and
Cullen argue are necessary for any study seeking to link macro-level structural conditions
of urban areas to crime rates. In their meta-analysis, they discovered that out of thirty-one
commonly used measures of structural covariates of crime, many of those listed above
remain in the "top ten", especially considering both the strength and stability of the
predictors across various studies (Pratt and Cullen 2005). Additionally as Pratt and
Cullen have noted, some of these concepts are not theory specific. In fact one could argue
that there is a movement away from testing specific theories and instead grouping
commonly used structural covariates to gauge the overall importance of structural
landscape of urban areas on crime rates (see Parker 2004).
Having presented several hypotheses for testing, the next several chapters will
intend to shed light on the importance of not only being able to identify traj ectories of
drug arrests over time, but also which structural covariates may predict being in a higher
versus lower traj ectory. Chapter four will present the data used to test these hypotheses
with the remaining chapters devoted to presenting the results of the analyses.
DATA AND METHODS
As discussed in prior chapters, structural theories have been used to explain many
types of crimes. Drawing upon the most widely and rigorously tested measures, I
describe these measures in the following sections. Following the meta-analyses of Pratt
and Cullen (2005), I have selected what have proven to be the most reliable and
theoretically relevant structural predictors over time. In addition I have chosen predictors
that Pratt and Cullen argue have had few (if any) serious methodological issues
throughout their empirical history (Pratt and Cullen 2005).
Since this research utilizes city-level analyses, all measures are operationalized as
aggregates at the city level. Both dependent and independent variables have been
constructed to represent the boundaries of American cities. This is consistent with past
research linking structural conditions to urban crime.
The unit of analysis is U.S. cities, beginning with those with a population greater
than 100,000 residents in 1980. While some debate has occurred regarding the best unit
of analysis (see Messner 1982; Bailey 1984; Messner and Golden 1992) for studying
structural covariates of crime rates, the city makes the most sense for a study of this kind.
First as Chamlin (1989) notes, aggregating police data across jurisdictions to the Standard
Metropolitan Statistical Area (SMSA) increases the risk of inflating these statistics (Liska
1989). Second, using city level data allows us to investigate the concentration effects
noted by Wilson (1987), where he contends that the effects of these structural factors
were due to the changes in the spatial and industrial structure of such areas (Wilson
1987). Finally Land and his colleagues (1990) argue that it makes little difference which
unit of analysis is selected (SMSA, City, etc.), providing proper statistical techniques are
employed to address the research question (Land et al. 1990).
This sampling strategy generates 179 cities in 1980 and these cities were the basis
for selecting cities for the entire sample in order to establish traj ectories for a nested
sample of cities. There are five maj or sources of data for this research. For the dependent
variables, the data are Uniform Crime Report arrest counts obtained from Chilton and
Weber (2000). The second and third major data sources are the 1980, 1990, and 2000
Census ofPopulation and Housing; Both the Characteristics of the Population and the
Social andEconomic Characteristics volumes (U.S. Bureau of the Census 1983; 1994;
2004). The fourth data source consists of the Bureau of Justice Statistics' Census of State
Adult Correctional Facilities ( 1979) and the Census of State and Federal Adult
Correctional Facilities (1990 and 2000). Finally, the Uniform Crime Report serves as the
source of information on the number of police officers in each city for 1980, 1990, and
2000. These data sources are essential to this research because they provide comparable
indicators during the time periods of interest (1980-2001 ).
This study utilizes race-specific drug-sales arrests based on continuous years from
1980-2001. I acknowledge the problem of missing data for one or more months of data
per year across the dependent variables. In those situations in which data were reported
for at least 9 or more months, mean substitutions were imputed to correct for missing data
on the remaining months. Arrest data for those cities reporting less than 9 months of
arrest statistics (i.e., 8 months or less) were considered missing. I also constructed a
dummy variable to indicate whether this mean substitution was performed in order to
determine whether it had significant impact on the data and results. The dependent
variable is computed as the race-specifie drug-sales arrest rate (arrests per 100,000
population) from 1980-2001. This was calculated as follows:
~ ~,,,,,= i RaceSpe cificDrugSalesA arrests "0,0
In addition to missing data from cities that failed to report all twelve months of
each year, it was also found that not all 179 cities reported drug sales arrests for each
consecutive year. It was decided that for those cities missing 3 or less years of data, mean
substitutions would be generated. Consequently I constructed a dummy variable for the
cities/years in which the mean substitutions were performed in order to determine if these
imputations made any significant impact on the data. Neither these nor the month
weighted dummy variables mentioned above proved to be significant, confirming the
integrity of the data.
Although mean substitutions were performed as described above, attrition is
inevitable when attempting to use data for 22 consecutive years for so many different
cities. After all missing data was disposed of and the mean substitutions were complete, a
Einal sample of 132 cities was retained as the final city list. This list is contained in
It is important to acknowledge an on-going debate surrounding the use of drug
arrest measures. There are two perspectives to consider--one being that a measure of
drug arrest rates reflects enforcement patterns or official responses to crime (see Mosher
2001; Quinney 1979) and the other is that measures of drug arrest rates indicate actual
drug behavior (Cohen, Felson and Land 1980; Rosenfeld and Decker 1999), particularly
at the city level (Rosenfeld 1986). It is not my intention in this dissertation to attempt to
decipher the true meaning of these measures. However I believe that these measures of
drug arrest rates provide a useful indicator of the amount/volume of race-specific drug
activity in a given area and thus can serve to gauge (at least the official accountability of)
drug activity in urban cities. It is my contention that no matter what these numbers may
actually mean, if a city experiences a dramatically increasing traj ectory of race specific
drug-sales arrests, this should indicate that the city has serious problems that must be
addressed (whether it be over-zealous law enforcement or increasing numbers of drug
dealers/users). Furthermore, in light of alternative indicators (e.g., DUF/ADAM, DAWN
and/or medical examiner' s data) of drug activity, prior research has reported high internal
reliability among data sources and that these data sources yield similar estimates of drug
use when compared to drug arrests for cities (Baumer 1994; Baumer et al. 1998;
Rosenfeld and Decker 1993; Warner and Coomer 2003). Given that the alternative data
sources are limited in sample size, drug arrests are a good proxy of drug activities across
multiple large urban cities over time.
I utilize a number of race-specific measures to serve as proxies of the larger
theoretical constructs of social disorganization, urban disadvantage and racial threat
theories, as well as their corresponding change scores in order to assess their impact on
drug sales arrests over time. These include race-specific measures of poverty, income
inequality via the GINI index, racial residential segregation, the percentage of jobless
males age 16+, and the percentage of children under the age of 18 not living with both
Poverty is defined as the percentage of black (white) persons living below the
poverty level and is calculated as follows:
Pove2~ =# PersonsBe lowPover2ln~pltyon ine
The numerator represents the race specific number of persons living below the poverty
rate and the denominator represents the total population of each race, respectively.
Income Inequality is measured using the Gini Index of Income Concentration and
is calculated as described by Coulter (1989:38). I calculate the Gini coefficients for 1980,
1990 and 2000 using race specific family income levels, as reported by the census. This
measure represents the proportion of the population within different income categories.
The index ranges from 0 to 1, with 0 representing perfect equality and 1 representing
perfect inequality (Coulter 1989).
Racial residential segregation is measured by the Index ofDissimilarity (D). As
described by Duncan and Duncan (1955) and Massey and Denton (1988) the index
represents the percentage of either population that would have to relocate in order for a
given area to achieve equal distributions (e.g. no segregation). The formula is expressed
D = -~~ *100
where b, and w~ are the number of blacks and whites in census tract j, and B and W are
the number of blacks and whites in the city of interest.
The measure of male joblessness is computed as the number of employed males
age 16+ in a racial group divided by the total number of males in that racial group that are
16 years of age and older, and multiplied by 100. The product is then subtracted from one
to represent the percentage of males in each racial group not employed. This is expressed
# Malesl 6 + Employed
Percentage2\ a~lesNotEmployed = 1-: *2ls 2Ge~sfg: 100
The final measure of urban disadvantage is the percentage of children under the age
of 18 that are not living with both parents. This measure is expressed as:
PercentKiu'fldsotithiothraents # Children Underl 8Nodl ir ing rS\itilrothParen ts 100
Two measures of social disorganization in cities are conceptualized as family
disruption and residential mobility. The first measure of family disruption is the race-
specific percentage of divorced males within the male population aged 15 and older. This
measures is expressed as:
~- ; r#Malesl 5 +InAge
Residential mobility is operationalized as the percentage of residents that reported
they were not living in the same residence for the previous five years and is expressed as:
# PersonsNotLivinglnSa~mePlace 5YearsoQoplt
The measure of racial threat represents the percentage of the population identifying
themselves as black. This measure is expressed as follows:
Percen lack # Re sidentsldenitfiedBI~c~Q~plack 100
Several control variables are utilized in this research. Based on their meta-analysis
of structural covariates of crime rates, Pratt and Cullen note that failing to control for
constructs such as the incarceration rate can lead to unreliable results (Pratt and Cullen
2005). The first is percentage of the population with Hispanic origins and is expressed as:
Pecet~sp.i # Re sidentsldenitfiedHisp anicoQopltn
The second control variable is the police presence rate (number of police officers
per 100,000 residents) and is expressed as:
,~,,,,;(Swrn~~c~f# eSwornPoplticOfficers ,00
The incarceration rate refers to the number of inmates in a given state per 100,000
residents) which is consistent with past research. Some researchers emphasize the
importance in lagging incarceration measures when attempting to explain crime. For
instance, the importance of lagging arises when attempting to explain current crime rates
with measures of the incarcerated population. This can be problematic since incarceration
data reflect the period through December 3 1 of the year the data represents. Given this,
lagging allows researchers to determine the effect of the incarceration rate in 1979 on
crime rates in 1980, for example. Since I am using the change in the incarceration rate
from one decade to the next as a control measure, lagging this measure makes little sense
(see Levitt 1996 and 2001). The rate is expressed as:
..,,,,,,,=(Tt~ Topu~talopulationO/State 0
In addition to the above mentioned control variables I have constructed several
"dummy" variables which represent regions that cities belong to. For instance it has been
found that cities in regions such as the South, Northeast, and West may experience
differential rates of crimes, so it is important to consider this as well.
As past research has noted, several of these structural covariates are highly
correlated with one another which results in multicollinearity (see Land et al. 1990), so in
order to deal with these issues I utilize principal components analysis, with varimax
rotation, in order to construct indices which are representative of the theoretical
constructs they are meant to measure, while maintaining the integrity of the data by
eliminating the partialing fallacy (Gordon 1967) often experienced when analyzing
aggregate data (Gordon 1967; Land et al. 1990).
The indices or component scores are calculated by multiplying the raw variables
with the weights obtained in the principal components analysis, and summing them
together to form one measure. This technique offer advantages over other techniques
such as scales or estimates. This results in a summary of the information contained in the
raw variables while avoiding many of the assumptions that scales or estimates may
contain (Kim and Mueller 1985).
Race specific indices for concentrated disadvantage are constructed, and their
corresponding component analyses are presented below in Table 4-1. While some
research has included racial residential segregation in similar indices, I have chosen to
preserve the segregation index in its original form in order to assess what affects the
change in segregation has on race specific drug sales arrests over time. This is consistent
with past research (Krivo et al. 1998; Krivo and Peterson 2000).
While the indices are race-specific, they are also calculated for each of the three
periods 1980, 1990, and 2000. By calculating the indices for each period it is possible to
generate change scores for not only the remaining independent variables in the models,
but also this index in order to assess what role the change in concentrated disadvantage
had on drug sales arrests specifically.
Table 4-1. Concentrated Disadvantage Index Factor Loadings.
1980 1990 2000
White Black White Black White Black
GINI Index .820 .698 .690 .880 .617 .871
% Below Poverty .915 .898 .906 .932 .925 .926
% Jobless Males .643 .706 .791 .766 .819 .772
% Kids One Parent .763 .867 .757 .649 .798 .752
Median Income -.865 -.894 -.820 -.932 -.813 -.928
Eigenvalue 3.26 3.34 3.17 3.52 3.21 3.64
% Variance 54.3% 55.7% 52.8% 58.7% 53.4% 60.7%
To further address the issue of multicollinearity and the partialing fallacy discussed
above, basic regression diagnostics were performed. I utilize the widely accepted
variance inflation factor, within OLS regression in order to determine to what extent the
standard errors of regression coefficients are inflated due to the possibility they share
variance with other predictor variables in the model. None of the variance inflation
factors in my diagnostics exceed four, which is a widely accepted cutoff (Fisher and
Mason 1981: 109; Messner and Golden 1992; Sampson 1987), therefore I feel confident
that my indices are unique and collinearity with other predictor variables is not an issue.
Statistical procedures for this study (PROC TRAJ) draw upon methods that have
been developed to analyze the developmental traj ectories of individuals over time. More
specifically, this procedure is based on a semi-parametric, group based modeling strategy
(Nagin 1999; Jones et al. 2001). The theory underlying this strategy relies on the
framework introduced primarily by Daniel Nagin and his colleagues (see Bushway et.al.
2001; Nagin and Land 1993; Land, McCall and Nagin 1996; Roeder, Lynch, and Nagin
1999; Nagin and Tremblay 1999; Nagin 1999). Essentially this technique relies on
mixture models for modeling unobserved heterogeneity within a given population and
generating offending (or arrest rate) traj ectories over time.
This method allows researchers to analyze longitudinal data over long periods of
time to identify distinct groups or clusters of offending (or arrest rates as used in this
study) over time. That is, it estimates the probability of belonging to a particular group as
estimating the probability (pk) of belonging to group (k) based on ( Lk), drug sales arrest
rates over time. This enables us to predict the likelihood of belonging to group k, while
also allowing certain parameters (risk factors) to vary across all groups or traj ectories,
which can be used to explain membership is each group respectively. This technique is
especially helpful in making sense of more complex issues, such as offending patterns
over time (Nagin and Tremblay 2005).
The method assumes that in any given population within longitudinal data, there
exists a mixture of groups, or traj ectories. In order to determine the number of groups that
best fit my data, I will rely on the Bayesian Information Criteria (BIC) since conventional
likelihood tests are unable to determine whether a more complex model (more groups) is
statistically superior to a simpler model (fewer groups) (D'Unger et al. 1998). Typically,
researchers seem to find that four or five groups are sufficient, however, Weisburd and
his colleagues (2004) determined that with their Seattle community level data, nineteen
groups or traj ectories were appropriate. However it is important to note that Weisburd et
al. observed an unusual number of data points over an extended period of time and the
estimation of their model took many hours for the model to converge (Weisburd et al.
2004). A recent debate has emerged surrounding the reification of such groups, and it
should be pointed out that the goal is to identify the simplest model that best delineates
each group (Nagin and Tremblay 2005).
The technique that I employ consists of a SAS plug in, downloadable from its
author, Bobby L. Jones. Jones developed the programming and commands in order to
allow SAS to estimate developmental traj ectories while studying with Daniel Nagin at
Carnegie Melon University. Since its inception, Jones' downloadable plug-in has made it
possible for a broad range of researchers to utilize a statistical procedure previously
utilized by select statisticians. As a result of this expanded availability, there has been a
sharp increase in the use of these techniques in the analysis of longitudinal data. Alex
Piquero reports that more than 50 academic peer-reviewed articles have been published
recently as a result of the availability and increased discussion of these techniques, across
a range of disciplines from psychology to criminology (Piquero 2004).
The specific model allows choice in whether data require censored normal model
(CNORM), zero-inflated Poisson (ZIP), and Bernoulli (LOGISTIC) distributions of
longitudinal data (Jones et al. 2001). I am utilizing the zero-inflated Poisson (ZIP) model
in order to compensate for the highly skewed nature of drug arrest rates across cities over
time as well as those that may have a high proportion of zero values. This decision is
based on my own analysis of the distributions of the dependent variable (arrest rates from
1980-2001), as well as close consultation with Bobby Jones, the software' s author and
statistician. The ZIP model takes the following form:
Pr(r; = y, C, = k, W =I w,)= ] pifk,)~]nl~ gk~l yk gkO~- fk)
v, =o v,,> v. Fit
Nagin and Tremblay (2005) have recently addressed concerns of what could be
interpreted as misuse or misinterpretation of the results of traj ectory analysis. They
caution that researchers should recognize the utility of the technique as a tool to simplify
complex data, however they must avoid the temptation to treat the resulting traj ectories as
gospel. Suppose Atlanta is found to be within traj ectory six, the highest drug sales arrest
traj ectory, caution must be emphasized to avoid literal interpretation of these suggestions.
That is, because Atlanta may be included in traj ectory six along with several other cities,
it is highly likely that none of the cities will display the exact traj ectory as the group, if
analyzed individually (Nagin and Tremblay 2005). However it is a useful tool
nonetheless as it allows researchers to group somewhat like offenders (cities) over time
based on their cumulative behavior over time. Because of this, it is beneficial to have the
most complete set of longitudinal data possible, for having more cases over longer time
periods logically allows more accurate and smooth groupings.
In a rej oinder to the aforementioned article above, Sampson and Laub (2005) argue
that TRAJ may in fact be yet another methodological fad. They argue that much like
LSREL and HLM dominated studies for some time, TRAJ may run the risk of becoming
a bandwagon that researchers are boarding while ignoring the risks involved. Nagin and
Tremblay's (2005) response to their rej oinder seems to make clear their intentions in
addressing common myths about PROC TRAJ in general. I agree with Nagin and
Tremblay in that TRAJ is a useful tool for simplifying complex data (such as longitudinal
offending rates), in order to better grasp the pattern of offending that may occur.
Sampson and Laub's concerns of abuses of such a technique are based on assumptions of
what might occur, based on what has happened in the past (as with LSREL and HLM),
and may be a bit premature (Nagin and Tremblay 2005; Sampson and Laub 2005).
In addition to the ability of this procedure' s ability to help researchers identify
developmental traj ectories over time, it also provides the ability to incorporate certain
risk factors in predicting group membership. That is, by incorporating risk factors,
researchers are able to perform multivariate analyses using time stable independent
variables, in order to discern how each covariate affects the probability of belonging to a
Risk factors that are believed to affect group membership based on drug sales
arrests are change scores linked to the structural covariates outlined above. In other
words, assume there exist six distinct traj ectories of drug sales arrests for both whites and
blacks over time for the sample of 132 cities. Does the change in social
disorganization/urban disadvantage indicators affect membership in each group? Will
increasing amounts of deindustrialization and urban disadvantage of urban areas predict
membership in the highest traj ectory of drug sales arrests versus the lowest?
There are several ways to model change over time, and disagreement still exist as
to which method is best suited for longitudinal analyses (Allison 1990; Firebaugh & Beck
1994; Hausman, Hall, & Griliches 1984; Kessler & Greenberg 1981). Researchers have
relied on several methods to model change over time using quantitative data. These
include the change score method, also called the gain score or difference score method,
the cross lag method and the residual change method. However a strong case has been
made that the change score method is superior, therefore I rely on this method for
calculating the change in race specific measures in order to determine their relationship to
drug arrest traj ectories. Change scores are calculated as the percentage change between
decades and is represented as:
Chang~e = [(VlurzeAtTime2 -Va~lueAtTimel)]/Va~lueAtTimel
For this study risk factors are based on the independent variables described earlier
in this chapter. More specifically I will use the changes in the independent variables
across decades to assess to what extent they predict membership in any given traj ectory. I
will calculate change scores as described above to model the change from 1980 to 1990
and 1990 to 2000. For the purposes of the PROC TRAJ software, these measures are
"time stable covariates" and their effects on group membership are modeled with a
generalized logit function as follows:
Pr(C1= kl Z = z2)= kx(,+ ~l
~exp(0; +7;'z )
In the following chapters several analyses are presented. First, descriptive statistics
of both the dependent and independent variables are presented and discussed in Chapter
five. This includes discussion of the structural change over time experienced by large
cities in the sample and discussion of the basic traj ectory groups. Chapter six will present
the multivariate results using the above measures to determine to what extent (if any)
these macro-level structural covariates have on predicting the life-course, or traj ectory
membership of race specific drug sales arrests over time.
As described in Chapter 4, the final sample consisted of 132 cities. In order to
assure nested models across time these cities are the bases of all analysis contained
throughout this research. Following are discussions of both the descriptive statistics of
the independent and dependent variables as well as the resulting drug sales arrest
traj ectories. Chapter 6 will focus on the multivariate models used to model structural
change in large cities and its effects on race-specific drug sales arrests over time.
Descriptive statistics for all independent measures are presented in Table 5-1 and
Table 5-2. Referring to Table 5-1, racial differences within cities become clear, across all
three decades. For example the median annual income for White families was over $6000
more than for Black families in 1980. Moreover, Blacks average just over 65% the
median family income of Whites, as presented by the ratio of Black to White median
family income in Table 5-1. This trend continues in 1990 and 2000. In 1990 the median
family income for Blacks is about $13,500 less than for Whites, and the ratio of Black to
White income drops slightly compared to 1980. In 2000, the gap widens while Blacks
average almost $19,000 less median family income compared to Whites.
The percentage of families living below the poverty line changed only slightly from
1980 to 1990, with Blacks remaining relatively unchanged and Whites increasing nearly
a half percent. In 2000 the percentage of Whites living below poverty increases another 1
percent while Blacks living below poverty decreased by about 1 percent. So while the
income gap in terms of median family income was increasing from 1980, the percentage
of families living below poverty began to decrease for Blacks. This however still leaves
the proportion of Blacks living in poverty more than double the rate for Whites.
The percentage of unemployed males aged 16 and older changed little from 1980 to
1990. However from 1990 to 2000 the percentage increased about 3% for both Whites
and Blacks alike. In 2000 about 33% of White males aged 16 and over were not
employed while about 43% of Blacks males fell into this category. This increase
maintained the 10% difference in regards to the number of unemployed Black males in
large cities compared to Whites.
Recall from Chapter four that I constructed a concentrated disadvantage index and
its corresponding change scores to both address the issue of multicollinearity as well as
capture the race specific changes in concentrated disadvantage across decades. The
change scores are presented in Table 5-2. Referring to this table we find that Whites
experienced an increase of about 74% in the concentrated disadvantage index from 1980-
1990, compared to an 85% increase among Blacks for the same period. The changes for
the period 1990-2000 are more modest, yet still denote increases for both races. During
this period disadvantage increased about 37% for Whites and around 40% for Blacks.
Referring to Table 5-1, in 1980 the percentage of white divorced males age 15 and
older was around 7% and by 1990 around 9% of White males age 15 and older were
divorced. As seen in Table 5-2, the mean change was about a 28% increase from 1980 to
1990. During the period from 1990 to 2000, the mean percentage of divorced white males
remained relatively unchanged, rising from 9% in 1990 to 9.9% in 2000. However as
seen in Table 5-2 the mean change score represents an increase of about 13.7%.
For Blacks, the mean percentage of divorced males age 15 and older increased from
8.46% in 1980 to 10.5% in 1990, with the mean change score being 27.2%. This change
is seen again with the percentage of divorced Black males rising from 10.5% in 1990 to
11.25% in 2000 and a mean change score representing about a 13% increase.
Racial differences again become apparent referring to Table 5-1 and the measure of
the percentage of children under the age of 18 not living with both parents. In 1980 about
23% of White children did not live with both parents as compared to almost 53% of
Black children. In 1990 the percentage increases to 26.5% for White children while rising
to about 62% for Blacks. The steady increase seen between 1980 and 1990 for both
Whites and Blacks did not occur for the period between 1990 and 2000. In 2000, with
roughly 26% of Whites not living with both parents, virtually no change occurred across
the decade. There was however a slight decrease in the percentage of Black children not
living with both parents as this measure declines to about 60%. Despite the slight decline,
proportionally there were still more than twice the percentage of Black children versus
White children not living with both parents in 2000 (60.43% versus 26.36%).
In Table 5-1 and Table 5-2 data on racial residential segregation are also presented.
Recall the Index of Dissimilarity is a measure of the evenness of the distribution of
groups across census tracts. The measure ranges form 0 to 100, with 0 being perfectly
integrated and 100 being completely segregated. If a city has a segregation score of 60,
this would indicate that about 60% of one racial group would have to change locations in
order to achieve equal distributions of racial groups across racial groups.
The mean score for Black-White in 1980 across all cities in the sample was 55.34.
This is considered a moderately high score, indicating at least 55% of either Blacks or
Whites would need to change residences in order to be more evenly distributed across
census tracts. In 1990 the mean drops to 48.61, and as seen in Table 5-2 the mean change
score is -13%, representing a mean decrease of about 13%. In 2000 the mean score again
declines to a mean of 44.85. The mean change score indicates a decrease of about -6.5%,
so again across decades the means continue to decline in regards to residential
segregation, indicating that overall cities are becoming less segregated as we progress
into the 21st century. It is important to note however that scores of 30 or so indicate low
levels of segregation, so the means discussed here continue to represent moderate levels
of segregation despite the declining levels from 1980 through 2000.
In regards to racial threat, the percentage of the population identifying themselves
as Black is presented in Table 5-1 and Table 5-2. In 1980 the percentage of the
population that was Black was about 16%. By 1990 this number increased to about
17.26%, with a mean change score indicating an increase of about 33%. For 2000 the
percentage again increased to about 19%, and a mean change of about 21%.
As discussed earlier several control variables are utilized in this research. Again,
these measures are depicted in Table 5-1 with mean change scores in Table 5-2. The first
control measure is the percentage of the population that is Hispanic. In 1980 the mean
percentage of the population identifying themselves as Hispanic was about 1 1%. By 1990
this number increased to almost 15% with a mean change score representing a 42%
increase across cities. In 2000 the figure increased once again to 20.5%, indicating a
mean change of about 76.5% from 1990.
The second control measure is the police rate. Recall this measure represents the
number of sworn police officers per 100,000 residents within cities. In 1980 the mean
was about 189 police officers per 100,000 residents. In 1990, this measure increased to
about 194 per 100,000, with a mean change of almost 3%. By 2000 the mean police rate
was about 209 police officers per 100,000 residents, with a mean change of 8% from
Finally the incarceration rate represents the Einal control measure. While this
measure is a state level measure, controlling for the widely publicized exponential
increase in incarceration is important to any study attempting to account for the changes
in crime rates over time. Like the police rate discussed above, this measure represents the
number of incarcerated individuals in a given state per 100,000 residents. In 1980 the
mean incarceration rate was about 142 prisoners per 100,000 residents. By 1990 this
number increased to about 505 inmates per 100,000 residents across cities. The dramatic
increase from 1980 to 1990 is represented by a mean change score of a 277% increase in
the incarceration rate. In 2000 the incarceration rate grew once again to about 764
inmates per 100,000 residents. This represented a mean change from 1990 to 2000 of
about 53%. While these increases are dramatic, they are not all that surprising,
considering the impact of the crack epidemic on perceptions of "get tough" policies for
Means and standard deviations for the dependent variable, race-specifie drug sales
arrest rates, are presented in Table 5-3. Upon first glance at the table, several things jump
out immediately. The first is the racial discrepancies in drug sales arrest rates. Recall this
measure is a rate indicating the number of race-specific arrests per 100,000 residents of
that same race within a city.
In 1980 the mean rate for Whites was about 48.5 arrests per 100,000 White
residents. For Blacks this number was about 134 arrests per 100,000 Black residents.
While this study aims to account for the differential changes in drug arrests across races,
it is important to note the significance of this first data item. Looking at the means just
presented, it becomes apparent that racial discrepancies did not simply begin in 1980.
However as looking further down Table 5-3 reveals the discrepancies climb
exponentially over time.
The other item that jumps out immediately from the table is that the drug sales
arrest rates for both Whites and Blacks appears to have peaked in 1989, as has been
argued to be the peak of the crack epidemic in most cities across the United States (Golub
and Johnson 1997). In 1989 White drug sales arrests peaked at about 157 arrests per
100,000 White residents while Black arrests peaked at about 689 arrests per 100,000
While the rates decline for both Whites and Blacks over the 22 year time period
studied here, clearly they never return to where the began, and in 2001 they are still
significantly higher than 1980. In 2001 the mean White rate was about 115 arrests per
100,000 White residents while the mean Black drug arrest rate was about 447 arrests per
100,000 Black residents.
Drug Sales Arrests Trajectories
Having presented the descriptive statistics for drug sales arrests in Table 5-3, the
remaining section of this chapter is reserved for presenting the traj ectories of race-
specifie drug arrests over time. Recall the previous discussion of the PROC TRAJ
procedure in Chapter 4. Figure 5-1 presents the traj ectories for White drug sales arrest
rates and Figure 5-2 contains traj ectories for Black drug sales arrest rates. These
traj ectories are examples of how cities behave, in regards to drug arrest rates over a
period of 22 years (1980-2001).
Referring to Figure 5-1, the bottom trajectory represents about 45% of the sample
or 59 cities. Note that this traj ectory exhibited very little changes over time, and also
serves as the reference group for the multivariate analyses forthcoming. Group two began
slightly higher than group one, remaining relatively flat until doubling from about 1988
to 1990, reaching a peak average of just over 100 drug sales arrests per 100,000 White
residents, and then tapering off but remaining significantly higher through the remainder
of the study period.
Group three comprised almost 17% of the sample or 22 cities. Group three begins
the study period averaging roughly 50 drug sales arrests per 100,000 White residents.
Around 1984, group three rapidly escalates to its peak in 1989, averaging around 239
drug sales arrests per 100,000 White residents. Like group two, group three also peaked
around 1989 and slowly tapered off, however like group two, group three also remained
significantly higher throughout the study period compared to where it began averaging
just under 200 arrests per 100,000 White residents in 2001.
Finally group four is the highest traj ectory group in the sample for White drug sales
arrests. Group four makes up about 12% of the sample or 16 cities. As is apparent in
Figure 6-1, group four clearly stands out from 1980 onward. At the beginning of the
study period, group four cities averaged over 100 arrests per 100,000 White residents,
then began a rapid progression to peak in 1989 at around 450 White drug sales arrests per
100,000 White residents. While tapering off after its peak in 1989, group four still
averaged over 350 drug sales arrests per 100,000 White residents in 2001, so while these
cities declined somewhat after 1989, they continued to report unprecedented White drug
sales arrest rates though 2001.
Overall, we find that White drug sales arrests peaked in 1989. This is consistent
with past research documenting the peak of the crack epidemic in this country. It is also
worth noting once again that nearly one half of the sample (59 cities) did not experience
the kind of rise in drug sales arrests over the 22-year period that one might expect by
simply examining means and average of all cities over time. This finding in and of itself
provides support for using a methodological approach such as PROC TRAJ in order to
distinguish cities that experienced little or no problems with drug arrests from cities that
experienced maj or problems. The BIC score was used for determining the proper number
of traj ectory groups. The BIC score for the four group model was significantly lower than
the three group model by 760. 107, indicating that four groups is the best fit.
The next section will provide a brief discussion of Black drug sales arrest
traj ectories as presented in Figure 5-2. Finally the chapter will conclude by revisiting
hypotheses one and two put forth in Chapter 3 as they relate to the existence and behavior
of drug sales arrest traj ectories.
The second traj ectory represents Black drug sales arrests for the same sample of
132 cities and is presented in Figure 5-2. The first thing that jumps out looking at Figure
5-2 is that the maximum average rate on the left hand scale exceeds 2000. That is, the
highest trajectory averaged over 2000 Black drug sales arrests per 100,000 Black
residents at its peak in those cities. Recall the White maximum above appears extreme,
but even the highest traj ectory group peaks at just under 500 White drug sales arrests per
100,000 White residents. This initial finding alone sheds some light on the importance of
disaggregating crime rates by race and utilizing race specific indicators while attempting
to account for crime rates in large cities.
About 25% of the cities in the Black model remained in the lowest traj ectory (34
cities), compared to about 44% (59 cities) for the White models. This provides further
evidence for the fact that more cities experience at least some changes in drug sales
arrests for Blacks compared to Whites. The lowest traj ectory for Blacks reaches a peak of
about 100 drug sales arrests per 100,000 Black residents, then tapers slightly and remains
flat to end in 2001 averaging about 76 drug sales arrests per 100,000 Black residents.
The second traj ectory in Figure 5-2 begins around the same levels as traj ectory
one but steadily rises to peak in 1995, averaging just under 400 Black drug sales arrests
per 100,000 Black residents. After 1995 it began to decline slightly and in 2001 averaged
around 250 Black drug sales arrests per 100,000 Black residents. This group comprised
about 3 8% of the sample or 50 cities.
The third traj ectory comprises almost 29% of the sample or 3 8 cities. Traj ectory
group three begins slightly higher than the first two groups in 1980 but climbs more
rapidly, peaking around 1989. At its peak this group averaged about 1100 Black drug
sales per 100,000 Black residents. While declining after its peak in 1989, traj ectory group
three tapers off slightly, still averaging over 600 Black drug sales arrests per 100,000
Black residents through the 1990's and into 2001. These 38 cities alone continued to
average more drug sales arrests per 100,000 Black residents than even the 16 cities
comprising the highest White traj ectory group, at its peak in 1989
Finally, looking at traj ectory group four in Figure 5-2, it becomes apparent that
some cities experienced astronomical increases in Black drug sales arrests over time.
Group four made up only about 7.6% of the sample or 10 cities. These 10 cities began the
period above the other three groups, averaging about 360 Black drug sales arrests per
100,000 Black resident sin 1980. From 1986 to 1989 these averages leapt from 600 to
more than 2000 Black drug sales arrests per 100,000 Black residents. While this rise
declined slightly after peaking in 1989, group four continues to mark unprecedented
Black drug sales arrest rates throughout the study period. In 2001 group four cities, on
average, made 1800 Black drug sales arrests per 100,000 Black residents.
Both groups three and four stand out not only because they rise so rapidly until
peaking in 1989 but also because both groups end in 2001 still averaging about 660 and
nearly 1800 Black drug sales arrests per 100,00 Black residents, respectively. These two
groups alone comprise around 3 5% of the sample, or 48 cities, and continued to average
higher Black drug sales arrest rates than even the highest White group at its peak in 1989.
As with the White models discussed above, the BIC score was used to assess
model fit. The BIC for the four group model was 257.43 less than the three group BIC,
indicating that the four traj ectory group model is the best fit.
Having now both presented traj ectory models for both White and Black drug sales
arrests over time, recall that several related hypotheses were put forth in Chapter 3.
Addressing Hypothesis 1, the evidence presented in Figure 5-1 and Figure 5-2 provides
clear and convincing evidence that cities do in fact 'behave' over time in regards to race
specific drug sales arrests across large cities. Additionally it was shown that nearly half
and one fourth the sample, experienced little changes in either White or Black drug sales
arrest rates over time, respectively.
Hypothesis 2 presented the argument that traj ectories would vary significantly by
race, with Black drug sales arrest rates rising more rapidly as compared to Whites. Again,
referring back to Figures 5-1 and 5-2, it is apparent that cities did in fact differ
significantly by race, with Black drug sales arrest rates rising more sharply as compared
to White drug sales arrest rates. Moreover this provides evidence that certain cities
experienced significant differences in drug arrests by race. That is, the highest arrest
traj ectory for White drug sales arrest rates consisted of 16 cities, while the highest group
among Black rates contained only 10 cities. However these 10 cities outperformed the
highest White arrest traj ectory by as much as four times the average number of arrests per
White (Black) residents.
Appendix A contains a list of all cities in the sample as well as their corresponding
group membership for both the White and Black traj ectory models. Comparing the lists
presented in Appendix A, we can see that six of the ten cities that fall in group four in the
Black traj ectories are also in group four in the White Traj ectories. Two are in California
(San Francisco and Bakersfield), two are in New Jersey (Newark and Jersey City), and
the remaining two are Louisville, Kentucky and Allentown Pennsylvania. Perhaps
Louisville, Kentucky is the most surprising of the six, whereas the other five cities may
be expected to have had drug problems as the crack epidemic was shown to have began
on the East and West coats as well as the Northeast in general.
Finally, having presented the traj ectories for both White and Black drug sales arrest
rates, Chapter 6 will provide results of the multivariate models in order to determine
whether changes in key structural covariates over time, affect the probability that a city
may be in a given traj ectory group versus being in the lowest traj ectory.
Table 5-1. Descriptive Statistics (Means with Standard Deviations in Parentheses).
1980 1990 2000
Race-Specific White Black White Black White Black
Median Family Income $21,450 $14,784 $39,576 $26,071 $55,084 $36,109
(3402) (4243) (8675) (8959) (13947) (11411)
Percent Families 9.87% 24.10% 10.46% 24.56% 11.55% 23.53%
Below Poverty (3.62) (8.42) (4.35) (9.43) (4.67) (8.95
GINI Index for Family .35 .39 .41 .43 .39 .43
Income (.027) (.051) (.034) (.049) (.033) (.042)
Percent Jobless Males 30.53% 39.8% 30.22% 40.25% 33.39% 43.28%
16+ (14.27) (13.99) (6.86) (11.29) (6.85) (10.10)
Percent Divorced 7.09% 8.46% 8.89% 10.5% 9.9% 11.25%
Males 15+ (1.82) (2.19) (1.98) (4.80) (2.02) (3.47
Percent Kids Not 23.40% 52.56% 26.53% 62.38% 26.36% 60.43%
Living with Both (4.58) (11.85) (5.60) (12.16) (5.99) (12.22) o
Index of Dissimilarity- 55.34 48.61 44.85
Racial Segregation (18.54) (18.47) (17.02)
Residential Mobility 47.04% 57.26% 51.92%
(27.88) (6.70) (5.56)
Percent Black 16.04% 17.26 18.99
(16.13) (16.88) (18.01)
Percent Hispanic 10.70% 14.83% 20.52%
(12.30) (15.53) (18.29)
Police Rate 189.20 194.45 208.74
(73.89) (79.14) (88.69)
Incarceration Rate 142.9 505.11 764.51
(47.37) (113.08) (213.43)
Percent Divorced Males 15+
Non Race-Specific Measures
Index of Dissimilarity-
Table 5-3. Means and (Standard Deviations) for Drug Sales Arrest Rates.
Year White Black
1980 48.52 (59.57 134.43 (192.88)
1981 54.34 (71.86) 151.39 (209.65)
1982 54.59 (57.76) 157.38 (172.56)
1983 61.41 (62.02) 180.29 (235.38)
1984 56.01 (60.09) 163.24 (173.94)
1985 72.80 (78.07) 190.81 (217.04)
1986 76.85 (87.47) 233.59 (275.14)
1987 85.91 (103.27) 324.51 (348.30)
1988 96.28 (115.24) 447.95 (568.93)
1989 145.14 (157.63) 668.89 (762.28)
1990 138.66 (150.62) 559.33 (615.19)
1991 130.31 (145.39) 573.36 (589.09)
1992 130.05 (146.61) 528.04 (524.98)
1993 127.02 (141.63) 494.83 (469.62)
1994 130.95 (144.12) 533.59 (593.21)
1995 131.43 (143.15) 491.04 (525.39)
1996 123.47 (142.85) 480.46 (517.96)
1997 127.34 (136.67) 467.22 (512.27)
1998 130.08 (147.09) 504.03 (559.44)
1999 120.88 (124.91) 479.01 (553.86)
2000 118.16 (132.75) 447.66 (558.64)
2001 115.73 (127.59) 446.63 (567.00)
Table 5-2. Mean Percent Changes Across Time for Independent Variables.
Race-Specific Measures White Black White
White Drug Sales Arrest Rates Vs. Time
80 82 84 86 88 90 92 94 96 98 00
Figure 5-1. Traj ectories of White Drug Sales Arrests 1980-2001.
Black Drug Sales Arrest Rates vs. Time
82 84 86 88 90 92 94 96 98 00
Figure 5-2. Trajectories of Black Drug Sales Arrests 1980-2001.
Having presented descriptive statistics for all variables in Chapter 5, this chapter is
devoted to the multivariate analyses of traj ectory groups for both White and Black drug
sales arrests. Specifically these analyses are aimed at identifying what, if any, effects the
changes in the structural characteristics of large cities from 1980 to 1990 and 1990 to
2000 had on the probability of group membership in developmental traj ectories. In other
words if a given city experienced a rapid exponential growth in drug sales arrests, were
particular structural covariates responsible in distinguishing the higher trajectory groups
versus the lower groups, as hypothesized in Chapter 3?
In order to test these hypotheses, the PROC TRAJ procedure allows the
introduction of risk factors into the traj ectory models, as discussed in Chapter 4. With the
introduction of risk factors, the procedure then estimates a logit model for each traj ectory
group, allowing each risk factor to predict the probability of group membership in each
respective traj ectory versus the lowest traj ectory (reference group). The parameter
estimates presented below can be interpreted as the natural log of the odds ratio of
belonging to a given group versus the reference group, for each risk factor (Jones and
It is also important note that one of the nicest things about these models (in my
opinion), is that they allow traj ectory groups to vary across the risk factors. In other
words it is not simply the values of each dependent variable contributing to group
membership, but once risk factors are introduced, the groups may vary based on the
influence of those risk factors. This is important since some critics have argued that
researchers overstate claims that a given individual (or city in this case) actually belongs
to a particular traj ectory group. However as Nagin and Tremblay point out, traj ectory
group membership is not static, and the fluidity of the model specifications allowing
groups to vary based on both dependent variable values as well as risk factors makes the
technique that much more useful in making sense out of otherwise very complex
longitudinal data (Nagin and Tremblay 2005).
White Drug Sales Arrests
Recall the White drug sales arrest traj ectories presented in Chapter 5. The model
fitting specification comparing BIC scores of more complex models compared to simpler
models, showed that the four-group traj ectory was the best fit for the arrest rates for 132
cities, including risk factors. I have established that trends in city drug sales arrest rates
not only differ over time but by race as well, supporting hypotheses land 2 from Chapter
Recognizing that cities did in fact behave differently for White drug sales arrest
rates raises the question of whether changes in the structural factors of those cities had
any affect on their behavior. The next several sections provide multivariate results
examining the effects of the structural changes from 1980-1990 and 1990-2000
independently in order to determine whether the changes from one decade to the next
were more salient than the other.
Accounting for Change: 1980-1990
The first multivariate model examines the impact of the change in key structural
covariates from 1980-1990 and their affects on group membership in the four traj ectory
groups discussed above. Tables provides parameter estimates for each risk factor, with
each said parameter representing the natural log of the odds ratio of belonging to a given
traj ectory versus belonging to the lowest traj ectory (group 1). Each group in the table
represents a separate logit model comparing that group to the reference group (group 1).
Table 6-1 provides parameter estimates for the full model representing the changes
in structural conditions from 1980-1990 and their corresponding affects on group
membership. Beginning with traj ectory group two note the significance of the increase in
concentrated disadvantage among Whites on the probability of cities being in group two
versus group one. Additionally, an increase in the index of dissimilarity (segregation)
decreases the probability of group two membership, compared to group one. The control
measure police presence loses significance once the theoretical measures are introduced
into the model.
Regarding the probability of being in arrest traj ectory three, again the increase in
concentrated disadvantage among Whites is a significant predictor of membership in
group three versus group one. Moreover the increase in residential mobility in these cities
from 1980-1990 increases the likelihood of being in group three versus group one,
controlling for other factors. Control variables that are significant for model three show
that cities being in both the South and Northeast were less likely to be in group three
versus groups one.
Finally, looking at traj ectory group four, the change in concentrated disadvantage
among Whites is no longer significant. However those cities that experienced an increase
in the percentage of divorced males from 1980-1990 were less likely to be in traj ectory
group four versus group one. As with the model three results, an increase in residential
mobility from 1980-1990 significantly increases the probability of cities being in
traj ectory group four versus group one. Furthermore, as a cities' black population
increases proportionally, the likelihood of being in traj ectory group four versus group one
for White drug sales arrests decreases.
The only statistically significant control variable for trajectory group four
probabilities is the change in the incarceration rate from 1980-1990. This is consistent
with arguments of other researchers (see Levitt 2004). The change in incarceration was
also among the most significant changing independent measures. Recall from Chapter 5
that from 1980-1990 the mean change in the number of prisoners per 100,000 residents
across states was 277%.
In the models below, the changes from 1980-1990 in both residential mobility and
concentrated disadvantage among Whites were significant predictors of cities being in a
higher arrest traj ectory versus the lowest. The mean change in concentrated disadvantage
was an increase of about 74% and the mean change in residential mobility from 1980-
1990 was about 58% across the 132 cities in the sample.
It appears that the changing landscape across cities from 1980-1990 for constructs
such as the concentrated disadvantage among Whites and the percentage of residents not
living in the same home as Hyve years prior could cause those areas to become more
conducive to drug selling. While these changes are only from 1980-1990, the next section
will address the changes from 1990-2000, again for White drug sales arrests. Following
the changes from 1990-2000 for White drug sales arrests, the same format will be used to
address the changes in Black drug sales arrests for the same periods.
Accounting for Change: 1990-2000
Table 6-2 represents the full model, with parameter estimates indicating the
probability that structural changes from 1990-2000 had a significant affect on traj ectory
group membership. Beginning with group two, only two measures are significant, both
control variables. The rise in the incarceration rate from 1990-2000 decreased the
likelihood of a city falling in traj ectory group two versus group one. Also, those cities in
the Northeast were less likely to be in traj ectory group two versus group one during this
period, controlling for all other factors.
Group three models show that the rise in concentrated disadvantage among Whites
decreased the probability of cities being in traj ectory group three versus group one.
Additionally, as the proportion of a city's Black population increased, its likelihood of
being in traj ectory group three versus group one decreased. Finally, cities in the South
during this period, were less likely to be in traj ectory group three versus group one,
controlling for all other measures.
Similar to the model for group three, the likelihood of a city being in traj ectory
group four versus group one decreased as concentrated disadvantage among Whites
increased from 1990-2000. Also like group three, as the proportion of the population
identifying themselves as Black increased from 1990-2000, the likelihood of those cities
being in traj ectory group four versus group one decreased. The remaining measures were
found to be insignificant. These results suggest that the structural changes from 1980-
1990 had more of an impact on cities' arrest rates for drug sales arrests, at least among