This is to certify that the
dissertation entitled

SPILL-OVER FROM ‘THE JUNGLE’ INTO THE LARGER
COMMUNITY:
SLAUGHTERHOUSES AND INCREASED CRIME
RATES

presented by
Amy J. Fitzgerald

has been accepted towards fulﬁllment
of the requirements for the

Doctoral degree in Sociology

 

 

4W (goo/L

Major Professor’s Signature

/2£/ /6’L

Date

 

MSU is an Afﬁmiative Action/Equal Opportunity Institution

LIBRARY
Michigan State
”University

 

 

 

 

 

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

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2/05 p:/ClRC/DateDue.indd-p.1

SPILL-OVER FROM ‘THE JUNGLE’ INTO THE LARGER COMMUNITY:
SLAUGHTERHOUSES AND INCREASED CRIME RATES

By

Amy J. Fitzgerald

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

2006

ABSTRACT

SPILL-OVER FROM ‘THE JUNGLE’ INTO THE LARGER COMMUNITY:
SLAUGHTERHOUSES AND INCREASED CRIME RATES

By
Amy J. Fitzgerald

In his now infamous book, Upton Sinclair referred to the massive slaughterhouse
complex in Chicago as The Jungle (1906). One hundred years later studies are beginning
to document the negative effects of large slaughterhouses moving into rural America. Of
these effects, dramatic increases in crime rates have been the least readily explainable.
The increases in crime have been theorized as being linked to: the characteristics of
slaughterhouse workers (who are mostly young male immigrants); general social
disorganization resulting in the loosening of social bonds and social control in these
communities due to abrupt population changes; and the high turnover rates in the industry
and resultant unemployment. While discussed theoretically, none of these explanations
have been empirically tested. Further, none of the literature to date has addressed the
possibility of a link between the increased crime rates observed and the violent work
conducted in slaughterhouses. The purpose of this study is to begin to ﬁll these gaps. To
do so this study utilizes data from six secondary sources obtained for all of the non-
metropolitan counties in states in the US with ‘right-to-work’ laws (a total of 581
counties) annually from 1994 through 2002. Pooled time series cross section techniques
are used to analyze the data using both OLS and negative binomial regression.

The analysis indicates that counties with slaughterhouses have signiﬁcantly
higher rates of arrests for sex offenses, reports of index offenses, rape, robbery, burglary,

motOr vehicle theft and arson, and counties with high levels of slaughterhouse

 

employment additionally have signiﬁcantly higher rates of total arrests, arrests for violent
offenses, rape, other assaults, and disorderly conduct. This study also ﬁnds that the
reasons proposed in the literature for the increases in crime cannot account for all of the
increases. Controlling for the demographic, social disorganization, and unemployment
variables proposed in the literature, as well as the time-invariant variables speciﬁc to each
county by using ﬁxed-effects to estimate the models, this study documents signiﬁcant
effects of slaughterhouse employment on arrest and report scales developed in the study,
as well as on the total number of arrests, arrests for violent offenses, rape, and other
sexual offenses. Further, the effects of slaughterhouse employment are unique when
compared to employment levels in other industries that also have high immigrant worker
concentrations, high injury and illness rates, and entail routinzed labor but do not require
the slaughtering of animals. This study therefore provides evidence of a relationship
between slaughterhouse employment levels and increased crime that can be generalized
from the earlier studies of communities where large new slaughterhouse facilities have
opened to all non-metropolitan counties in right-to-work states. Additionally, this study
demonstrates that the theories proposed in the literature cannot explain all of the
increases in certain types of crimes experienced in these communities and that the effects
of slaughterhouse employment appear to be unique compared to other similar industries
that differ simply in that animals are not the raw materials of production. These ﬁndings
point to the need for further research investigating the possibility that the type of work
conducted in slaughterhouses (killing and dismembering animals) is at least partially
responsible for the increases in crime that cannot be explained by the theories put forth in

the literature and are not paralleled by comparison industries.

This research is dedicated to the humans and other animals whose lives have been
devastated and terminated, although in divergent ways,
by the industrialized animal slaughter industry.

ACKNOWLEDGMENTS

This research would not have been possible without the assistance of many
people. I would like to thank Noelle Hartner of the Department of Community,
Agriculture, Recreation, and Resource Studies at MSU; Shawn Nicholson at the MSU
library; and Darren Warner of the Library of Michigan for their assistance in acquiring
parts of the data used in this study. I am also indebted to several faculty members —
Aaron McCright, Angela Mertig, and Toby Ten Eyck - for their thoughtful feedback on
this project, as well as on others. I owe a deep debt of gratitude to Linda Kalof and Tom
Dietz for the amount of time and expertise they have devoted to this project and
additionally for their continual support and encouragement. Finally, I would like to thank
my family and friends, particularly Sean Demers and Sandra Fitzgerald, for listening to

me talk about a generally unpleasant topic and for providing emotional support.

TABLE OF CONTENTS

LIST OF TABLES ................................................................................. viii

LIST OF FIGURES ................................................................................... x

INTRODUCTION ............................................................................................................... 1

The Research Problem ...................................................................... 2

The Study ..................................................................................... 3

The Contribution ............................................................................. 9

The Format .................................................................................. 12
CHAPTER 1

LITERATUREREVIEW ........................................................................... 14

The History of the Slaughterhouse ...................................................... 15

The agricultural revolutions ....................................................... 15

Banishing the slaughterhouses from society ................................... 16

The Modern Meatpacking Industry ...................................................... 18

Meat as mass production .......................................................... 18

Unionism in the meatpacking industry ......................................... 19

The IBP revolution ..................................................................................... 23

The Effects of the Contemporary Meatpacking Industry ...................................... 27

Effects on the physical environment and human health ............................ 28

Slaughter and the worker ........................................................................... 31

Slaughter and the social impacts on the community .................................. 34

Theorized causes of increased crime in slaughterhouse communities ....... 37

Summary ................................................................................................................ 61
CHAPTER 2

RESEARCH METHODOLOGY ....................................................................................... 63

Study Design .......................................................................................................... 63

Data Collection ...................................................................................................... 69

Instrumentation and Measurement ......................................................................... 70

Independent Variables ............................................................................... 70

Control Variables ....................................................................................... 74

Dependent Variables .................................................................................. 80

Data Analysis ......................................................................................................... 81

The Longitudinal Dimension ..................................................................... 82

The Count Dimension of the Dependent Variables ................................... 88

Methodological Strengths and Limitations ............................................................ 91

Summary ................................................................................................................ 98
CHAPTER 3

DESCRIBING THE EFFECTS OF SLAUGHTERHOUSES ........................................... 99

The Slaughterhouse Variables ............................................................................... 99

vi

Bivariate Relationships ........................................................................................ 101

Summary Arrest Rate Variables .............................................................. 102
Speciﬁc Arrest Rate Variables ................................................................. 109
Crime Report Variables ........................................................................... 112
Increased Crime in Counties with Slaughterhouses? ........................................... 117
Examining the Differences between Counties with and
without Slaughterhouses .............................................................. 1 17
Examining the Effects of High Levels of Slaughterhouse
Employment ............................................................. 123
Summary .............................................................................................................. 125

CHAPTER 4
INTRODUCING CONTROL VARIABLES AND COMPARISON INDUSTRIES. ...127
Analyzing Slaughterhouses and Crime Rates Utilizing OLS Regression... . . ....128

Factor Analysis ....................................................................................... 128
Multiple Regression Analyses ................................................................. 131
Estimating the Effects of Levels of Slaughterhouse Employment ........ 138
Analyzing Slaughterhouses and Crime Rates Utilizing Negative
Binomial Regression ................................................................................ 141
Overall Effects of the Predictor Variables on the Crime Variables ................ 158
Comparing the Results from the Restricted Time Period with the Entire
Time Period .............................................................................................. 160
Estimating the Effects of Levels of Slaughterhouse Employment ................. 166
Summary .............................................................................................................. 169
CHAPTER 5
DISCUSSION AND CONCLUSIONS ........................................................................... 170
Slaughterhouses and Crime: Making Sense of the Study Results ................. 170
Bivariate Correlation Results ................................................................... 170
Testing Hypothesis 1 ................................................................................ 173
Testing Hypothesis 2 ................................................................................ 177
Testing Hypothesis 3 ................................................................................ 183
Theoretical Implications ...................................................................................... 186
Recommendations ................................................................................................ 1 99
Holding Corporations Responsible .......................................................... 199
Educating Consumers .............................................................................. 200
Further Research ...................................................................................... 201
NOTES ............................................................................................................................. 206
APPENDICES ................................................................................................................. 208
REFERENCES ................................................................................................................ 227

vii

Table 1.

Table 2.

Table 3.
Table 4.

Table 5.

Table 6.

Table 7.

Table 8.

Table 9.

Table 10.

Table 11.
Table 12.
Table 13.
Table 14.
Table 15.
Table 16.

Table 17.

LIST OF TABLES

Rural-Urban Continuum Codes, 2003 .......................................... 68
Changes in the Number of Slaughterhouse Facilities

from 1997 to 1998 ................................................................. 72
Slaughterhouse and Comparison Industries .................................... 73
Frequency Distribution of Slaughterhouses for County/Year ............. 100

Trends in Slaughterhouse Establishment and Employment
Variables, 1994-2002 ............................................................ 101

Descriptive Statistics and Zero Order Correlations, Independent .
Variables, Control Variables, and Summary Arrest Variables. . . . . . . .....103

Descriptive Statistics and Zero Order Correlations, Independent
Variables, Control Variables, and Individual Arrest Variables ............ 110

Descriptive Statistics and Zero Order Correlations, Independent
Variables, Control Variables, and Crime Report Variables ................ 114

Difference in Crime Rate Means between Counties with and
without Slaughterhouses ........................................................ 121

Difference in Crime Rate Means between Counties with
Slaughterhouse Employment of 1000+ and Counties with 0-999

Slaughterhouse Employees ...................................................... 124
Eigenvalues for the Arrest Rate Factor Analysis ............................ 129
Factor Loadings for Arrest Variables ......................................... 129
Eigenvalues for the Report Rate Factor Analysis ........................... 131
Factor Loadings for Report Variables ......................................... 131

Multiple Regression with Arrest Scale as the Dependent Variable. .......133
Multiple Regression with Report Scale as the Dependent Variable ...... 134

Results of TSCS OLS Equation at Varying Levels of
Slaughterhouse Employment, Keeping Control Variables Stable. . . . . ....139

viii

Table 18.

Table 19.

Table 20.

Table 21.

Table 22.

Table 23.

Table 24.

Table 25.

Table 26.

Table 27.

Table 28.

Table 29.

Table 30.

Table 31.

Table 32.

Negative Binomial Regression with Total Arrests as the Dependent
Variable, 1994-1997 ............................................................. 145

Negative Binomial Regression with Arrests for Violent Crimes
as the Dependent Variable, 1994-1997 ........................................ 146

Negative Binomial Regression with Arrests for Murder as the
Dependent Variable, 1994—1997 ............................................... 148

Negative Binomial Regression with Arrests for Rape as the
Dependent Variable, 1994-1997 ............................................... 149

Negative Binomial Regression with Arrests for Offenses against the
Family as the Dependent Variable, 1994-1997 .............................. 150

Negative Binomial Regression with Arrests for Sex Offenses as the
Dependent Variable, 1994-1997 ............................................... 151

Negative Binomial Regression with Arrests for Aggravated Assault
as the Dependent Variable, 1994-1997 ........................................ 152

Negative Binomial Regression with Reports of Index Offenses as
the Dependent Variable, 1994-1997 ........................................... 154

Negative Binomial Regression with Reports of Murder as the
Dependent Variable, 1994-2002 ............................................... 155

Negative Binomial Regression with Reports of Rape as the
Dependent Variable, 1994-1997 ............................................... 156

Negative Binomial Regression with Reports of Assault as the
Dependent Variable, 1994-1997 ............................................... 157

Effects of the Independent and Control Variables (Net of the other
Variables) on the Crime Variables of Interest, 1994-1997 ................. 161

Comparison of [RI values for the Effects of Slaughterhouse
Employment Prior to 1998 and for the Entire Study Time Period
(1994-2002) on the Dependent Variables .................................... 162

Effects of the Independent and Control Variables (Net of the other
Variables) on the Crime Variables of Interest, 1994-2002 ................. 164

Incidence Rates Obtained via the TSCS Negative Binomial
Regression Equation at Varying Levels of Slaughterhouse
Employment, Keeping Control Variables ..................................... 168

Figure 1.
Figure 2.
Figure 3.
Figure 4.
Figure 5.

Figure 6.

LIST OF FIGURES

Total Arrest Rates in Counties with and without Slaughterhouses........118
Index Report Rates in Counties with and without Slaughterhouses... . . ..l 19
Rape Arrest Rates in Counties with and without Slaughterhouses........119

Rape Report Rates in Counties with and without Slaughterhouses ....... 120
Log Scale Prediction Equation Values for Arrest and Report Scales ..... 140

Log Scale Prediction Equation Values for Total Arrests, Arrests
for Violent Offenses, Rape, and Sexual Assaults ............................ 168

INTRODUCTION

It was moral, spiritual, and physical degradation, a ‘jungle’ in which humans lived
barely above the level of animals. — Upton Sinclair, The Jungle, [1905]1946.

The above quotation, from Upton Sinclair’s introduction to the 1946 Viking Press Edition
of his famous book, The Jungle, explains that the title of the book was intended to be a
reﬂection of the working conditions in the stockyard slaughterhouses. Although
Sinclair’s purpose was to expose the public to the working conditions in slaughterhouses
to arouse socialist sympathies, the book also alluded to the effects that the degradation
within the slaughterhouses had on the workers and the broader community. For instance,
in explaining the numerous ﬁghts instigated by slaughterhouse workers after hours and
the reactions of the police thereupon, Sinclair tangentially noted a connection between
these ﬁghts and the killing and dismembering of animals all day at work:

He [the police ofﬁcer] has to be prompt — for these two-o’clock-
in-the-moming ﬁghts, if they once get out of hand, are like a
forest ﬁre, and may mean the whole reserves at the station.
The thing to do is to crack every ﬁghting head that you can
see, before there are so many ﬁghting heads that you cannot
crack any of them. There is but scant account kept of cracked
heads in back of the [stock] yards, for men who have to crack
the heads of animals all day seem to get into the habit, and to
practice on their friends, and even on their families, between
times. This makes it a cause for congratulation that by modern
methods a very few men can do the painfully necessary work
of head-cracking for the whole of the cultured world

(1946, 18-19, emphasis mine).

The Research Problem

In more recent times slaughterhouses have been relocated from urban areas such as
Chicago (the setting of Sinclair’s book) to rural communities. The impacts of the
opening of slaughterhouses in smaller communities have received particular attention of
late. Of course, some impacts are to be expected, since “social and environmental
impacts are inevitable with the arrival of a new industry” (Broadway 2000, 37); however,
research has found that slaughterhouses might have a particularly pernicious inﬂuence on
the surrounding communities, such as an increase in social disruption in the community,
as characterized by Sinclair above.

The forms of social disruption documented by research in slaughterhouse
communities thus far include housing shortages (due to the inﬂux of workers into the
community), increased demand for social assistance (due to a number of factors,
including the low wages paid by the industry, high injury and illness rates, and the high
employee turnover rate), and an increase in crime. Of these social impacts, the increase
in crime has often bafﬂed researchers particularly since in some cases the increases have
been dramatic. For example, while Finney County, Kansas experienced a 33% increase
in population after the construction of two slaughterhouse facilities they simultaneously
witnessed an astounding 130% increase in violent crimes (Broadway 1990).

Various explanations have been proposed for the increase in ofﬁcial crime
statistics in slaughterhouse communities. The increased crime has been theorized as being
linked to the characteristics of slaughterhouse workers, who are mostly single young

adult male immigrants (see Broadway 1990; Broadway 1994; Broadway 2000; Broadway

2001; Broadway and Stull 2005; Stull and Broadway 2004). Additionally, it has been
suggested that the increase in crime in slaughterhouse communities might be associated
with social disorganization and a resultant loosening of social bonds and social control in
slaughterhouse communities due to the abrupt population changes (Broadway 2000;
Markus 2005; Stull and Broadway 2004). Finally, it has also been suggested that the
increase in crime in slaughterhouse communities is due to the high turnover rates in the
industry and resultant unemployment (Broadway 1998; Eisnitz 1997; Schlosser
2005[2001]). While discussed theoretically, none of these explanations of the increased
crime rates have been veriﬁed empirically, prompting Broadway to remark, “[c]learly,
further research is needed to determine the reasons behind the increases in social
disorder” (1990, 342). Further, none of the literature to date has addressed the possibility
of a link between the violent work conducted in slaughterhouses and social disruption in
the surrounding communities — an effect which Sinclair seems to have been attuned to
one hundred years ago — which would at least partially account for the observed increase

in crime in these communities. The purpose of this study is to begin to ﬁll that gap.

The Study

As mentioned above, a number of case studies (reviewed in detail in the next chapter)
have documented increases in crime in slaughterhouse communities (Broadway 1990;
Broadway 1994; Broadway 2000; Broadway 2001; Broadway and Stull 2005; Gouveia
and Stull 1995; Grey 1995; Grey 1998; Horowitz and Miller 1999; Stull and Broadway

2004). Building upon those studies, this study utilizes data from six secondary sources to

examine the crime patterns (the United States Department of Agriculture, Economic

Research Service; Bureau of Labor Statistics; US Census Bureau, Small Area Income and

Poverty Estimates; US Census Bureau, Population Estimates; the FBI’s Uniform Crime

Reports; and the County Business Patterns). In doing so, this study tests the following

hypotheses at the county level:

Hypothesis 1:

Hypothesis 2:

Hypothesis 3:

General and speciﬁc crime rates1 in counties will increase as the number

of slaughterhouses and slaughterhouse employees increases.

Controlling for the variables proposed in the literature (the unemployment
rate, the number of people in poverty, net immigration, net migration, the
number of non-white and/or Hispanic residents, the number of young
males, the total number of males, and the population density of the
county), slaughterhouse presence and employment will be associated with
increased crime rates in counties, more so than industries that utilize the
same type of labor force, have high injury and illness rates, and entail

routinized labor, but do not involve killing and dismembering animals.

Controlling for the variables proposed in the literature (the unemployment
rate, the number of people in poverty, net immigration, net migration, the
number of non-white or Hispanic residents, the number of young males,

the total number of males, and the population density of the county), rape

and family violence rates in particular will increase in counties where

there is an increase in slaughterhouse employees.

A test of the ﬁrst hypothesis is needed because the studies conducted thus far have
only examined a few communities where extremely large slaughterhouses have recently
opened. Therefore, it is necessary to test whether or not results from those studies are
generalizable. Testing the second hypothesis is necessary to ascertain whether the
increase in crime in slaughterhouse communities can be explained by the variables
proposed in the literature, and if the effects are unique to slaughterhouses, or if
employment rates in similar industries characterized by high levels of immigrant
employment, high injury and illness rates, and routinized work would result in similar
increases in crime. The ﬁnal hypothesis is grounded in the theorizing of scholars such as
Adams (1991; 1995), Patterson (2002), and Spiegel (1996) positing a link between the
victimization of animals and the victimization of less powerful human groups, such as
children and women. Testing this hypothesis will demonstrate whether or not there is a
speciﬁc link between the amount of slaughterhouse employment in counties and the
victimization of less powerful groups of people, such as women and children, in the form
of rape and family violence.

The ﬁrst hypothesis is tested using t tests, which permits an assessment of
whether there is signiﬁcantly more crime in communities with slaughterhouses compared
to those without and whether or not there is signiﬁcantly more crime in communities with
high levels of slaughterhouse employment compared to those with little or no

slaughterhouse employment. The second and the third hypotheses are tested using two

techniques, which were selected due to the longitudinal dimension of the data and the
discrete count nature of the dependent crime variables (described in detail in Chapter 2).
The ﬁrst technique entails using factor analysis to create arrest and report crime scales.
Then these scales are used as the dependent variables in pooled ﬁxed effects time series
cross section (TSCS) ordinary least squares (OLS) regression. The second technique
entails analyzing individual dependent crime variables using ﬁxed effects TSCS negative
binomial regression, which takes the discrete count nature of the dependent variable into
account. The combination of these two techniques permits a more complete picture of
crime in slaughterhouse communities. ‘

Testing the second and third hypotheses also requires the use of several control
variables and comparison industries. The control variables were selected based upon
theorizing in the literature surrounding the causes of increased crime in slaughterhouse
communities, and include the following: the county unemployment rate, the number of
people in poverty in the county, net immigration into the county, net migration into the
county, the number of non-white and/or Hispanic residents in the county, the number of
young males in the county, the total number of males in the county, and the population
density of the county. Several comparison industries are also included in the analyses.
These industries were selected to match the slaughterhouse industry as closely as possible
on the following characteristics: injury rate, illness rate, the concentration of immigrant
employees, and de-skilled routinized labor. These criteria resulted in the selection of the
following ﬁve industries: Iron and Steel Forging, Truck Trailer Manufacturing, Motor
Vehicle Metal Stamping, Sign Manufacturing, and Industrial Laundering. All of these

variables are described in greater detail in Chapter 2.

Twenty two crime variables are utilized as dependent variables in this study.
These crime variables were selected for theoretical and methodological reasons. Many of
these variables are of theoretical interest because they are violent offenses which are
implicated by the hypothesis that violence from the slaughterhouses would spill-over into
the larger community. The other variables were identiﬁed by factor analysis as grouping
together with the variables of most theoretical interest. These variables include fourteen
arrest variables and eight report variables. Both arrest and report variables are utilized in
this study because there can be important differences between the two. For instance, the
arrests made for rape can be markedly less than the number of reports of rape. Two of
the arrest variables (total arrests and arrests for violent offenses) are termed general or
summary variables. The other twelve arrest variables are for speciﬁc or individual
offenses, including arrests for murder, rape, offenses against the family, sex offenses,
assault, robbery, burglary, forgery, possessing stolen property, vandalism, other assaults,
and disorderly conduct. Of the report variables, one is a summary variable (reports of
Index offenses, which is comprised of reports of murder, rape, robbery, aggravated
assault, burglary, larceny, and motor vehicle theﬁ). The other report variables included
are for the following offenses: reports of murder, rape, assault, robbery, burglary, motor
vehicle theﬂ, and arson.

The deﬁnition of slaughterhouses and employment therein utilized in this study is
the deﬁnition used by the North American Industry Classiﬁcation System (N AICS):
Animal (except poultry) slaughtering. This category includes establishments that
slaughter quadruped animals (mostly cows and pigs) and prepare meats. Prior to the shift

to the NAICS system of classiﬁcation in 1997 (the conversion to NAICS, however, was

not implemented in the slaughterhouse industry until the following year), the Standard
Industrial Classiﬁcation (SIC) system referred to this type of work as meat packing plant
employment. In 2002 (the most recent year of data used in this study), there were
148,551 people categorized as working in Animal (except poultry) slaughtering, and
there were 1,816 establishments (County Business Patterns, US Census Bureau). In these
establishments approximately 9.15 billion animals are killed annually (up from 3.36
billion in 1975) (Markus 2005).

As will be discussed in later chapters, changes in industry classiﬁcation posed a
challenge for this study. More speciﬁcally, in 1998 ‘custom slaughter’ facilities were
added to the slaughterhouse category. These facilities are smaller enterprises that either
slaughter animals for their own consumption or slaughter animals for individuals who
own them and then use the meat for his/her consumption. The potential implications of
the change in classiﬁcation are discussed in Chapter 2, and to the extent possible the
analyses were adjusted to take this change in classiﬁcation into account.

As alluded to above, the unit of analysis for this study is the county, which is
large enough to permit access to uniform data across years and cases, but small enough so
that each case should be sensitive to the effects of the slaughterhouses and presumably be
less affected by confounding variables than larger units, such as the state. Only non-
metropolitan counties not adjacent to metropolitan areas in right-to-work states (in right-
to-work states employees cannot be required to join or pay dues to a union and may
resign from the union at any time, but still enjoy the beneﬁts of the collective agreement)
are analyzed in this study in order to protect against the confounding effects of

urbanization and strong labor laws. Most slaughterhouse facilities have been moved to

non-metropolitan counties in right-to-work states in order to reduce labor costs. The

delimitations of this study are explored in more detail in Chapter 2.

The Contribution

This study makes contributions in both theory and method. Regarding the theoretical
state of the slaughterhouse literature, Brueggemann and Brown (2003, 338) have stated
that it has not been theoretically infused “because the extant work is primarily historical,
it neither delves deeply into the signiﬁcant developments of the last 40 years, nor
contributes much in the way of theory.” The literature that does address the recent
changes in the industry and the social effects has included theorizing surrounding the
hypothesized causes of the crime increases in these communities, but this study is the
ﬁrst to begin to test these theories. This study is also the ﬁrst to compare the effects of
slaughterhouses with other industries.

This study additionally moves us toward what Richard York refers to as ‘a
sociology of the slaughterhouse’, “a deeper understanding of how exploitation and
oppression (not to mention horriﬁc acts of cruelty) are perpetrated and justiﬁed” (York
2004, 263), and how they are connected. Studies that have documented a link between
the abuse of animals and the victimization of people have examined behaviors towards
animals that have been deemed unacceptable by society and termed abusive. What is
missing in this literature is an examination of systematic harmful behavior towards

animals that is sanctioned in our society. Such sanctioned behavior occurs on a massive

scale in slaughterhouses. Accordingly, criminologist Piers Beirne has proposed that

studying slaughterhouses would provide a useful case for examining a potential link
between socially sanctioned harmful behavior to animals and the victimization of people.
He states, “the multiple sites of violence condoned in slaughterhouses perfectly
exemplify these practices [socially sanctioned harm to animals]. Consider how these
might lead, or ‘progress,’ to extra-institutional violence” (Beirne 2004, 54). This study
constitutes the ﬁrst empirical step in examining the effects of the socially-sanctioned
violence undertaken in slaughterhouses. I theorize that labor which requires killing and
dismembering animals constitutes a different kind of alienation and has different effects
on workers and the community than work related to the manufacturing of inanimate
materials.

In addition to these theoretical contributions, this study also makes contributions
by way of research methods. Most notably, this study uses panel data for a nine-year
period. The advantages of this type of analysis are important. First, having data from
several years permits a fuller picture of what is transpiring in these counties than what
could be gleaned from a cross-sectional study, especially given the sometimes sporadic
changes in crime rates. For instance, a mass murder in a county in a given year would
signiﬁcantly impact the data and the results of a purely cross-sectional analysis. Second,
the ﬁxed effects techniques employed in this study permit the controlling of time-
invariant county effects. Therefore, factors that could affect crime rates, such as
geographic area and historical/cultural elements, are controlled for in this study.

Further, this study has greater breadth than previous studies of crime in
slaughterhouse communities. This study includes 581 counties. Previous studies of the

social consequences of slaughterhouses have been conducted on only a few counties

10

where very large slaughterhouses facilities have opened (such as F inney County,
Kansas). Therefore, this study permits an examination of whether or not the increases in
crime found in the few community studies conducted thus far are generalizable to other
counties where slaughterhouses have not recently opened, to counties with smaller
slaughterhouse facilities and less employment in the industry. Using data from several
similar industries, this study additionally permits a comparison of the effects of
employment in slaughterhouses on crime with the effects of other industries.

This study also examines a greater number of crimes than previous studies, which
have generally only reported increases in the total number of arrests. In addition to the
total number of arrests, this study examines thirteen other arrest variables and eight report
variables. This facilitates a better understanding of which crimes are affected by
slaughterhouse employment and a comparison of the effects on arrest rates versus report
rates.

Finally, this study uses two different techniques to derive the results. Factor
analysis of the crime variables is used to create scales, which are then used as dependent
variables in TSCS OLS regression, and TSCS negative binomial regression is used to
analyze the effects of the independent and control variables on individual crime variables.
The use of different techniques to derive the results provides increased evidence that the
results are robust.

In spite of the theoretical and methodological contributions made by this study,
due to the aggregation of the data, conclusions cannot be drawn about whether or not
being employed in slaughterhouses causes violent and other forms of criminal behavior,

or what such a causal mechanism might look like. Rather, this study is a ﬁrst step in

examining 1) if levels of slaughterhouse employment in communities are related to
crime; 2) whether or not this relationship is unique compared to industries similar to
slaughterhouses in characteristics except for the killing and dismemberment of animals;
and 3) whether slaughterhouse employment is still related to crime rates once the
explanatory variables proposed in the literature are controlled for. The ﬁndings of this
study indicate that for some types of crime in slaughterhouse communities the three
propositions above are supported. The ﬁndings presented here also demonstrate that
subsequent research is needed to understand if and why employment in slaughterhouses

is causing the increased crime rates observed in these communities.

The Format

Before discussing the ﬁndings of this study it is important to properly contextualize it
temporally and within the larger scholarship. Chapter 1 provides an historical
examination of the development of the modern slaughterhouse and details the research
that has been conducted thus far on slaughterhouse communities. The chapter situates
this study within the substantive literature on slaughterhouses and the broader literatures
of community sociology and criminology.

Chapter 2 details the study design. The variables used in this study are
Operationalized and discussed, linking this chapter to the literature review conducted in
Chapter 1. Chapter 2 also provides an important examination of the methods and
analysis techniques used in this study. Finally, attention is paid to the strengths and

limitations of the design of the study.

Chapters 3 and 4 present the results of the study. Chapter 3 details the results of
the descriptive analysis, providing insight into the bivariate relationships between the
variables and testing the ﬁrst research hypothesis — whether there are higher crime rates
in counties with slaughterhouses. Chapter 4 introduces the control variables and the
comparison industries using multiple regression models and TSCS techniques.

Chapter 5 provides a discussion of the results detailed in the previous two
chapters, highlighting the conclusions that can be drawn from this research, the
recommendations that can be made in light of the ﬁndings, and the types of future

research needed in this area.

It will be demonstrated in the following pages that the amount of slaughterhouse
employment in a county does have signiﬁcant positive effects on some types of crime
(most speciﬁcally, on the arrest and report rate scales developed in this study, as well as
on individual variables including total arrests, arrests for violent offenses, arrests for rape,
and arrests for sex offenses) — effects which are unique when compared to other similar
industries that do not entail killing and processing animals and are also signiﬁcant even
with all of the variables proposed in the literature controlled for. It is argued herein that
these results indicate that the possibility of a diffusion or spill-over of violence from the
slaughterhouse (aptly referred to as ‘the jungle’ by Sinclair one hundred years ago) into

the larger community can no longer be ignored.

l3

Chapter 1

LITERATURE REVIEW

Auschwitz begins wherever someone looks at a slaughterhouse and thinks: they ’re only
animals.
- Theodor Adorno

In this chapter I provide an overview of the literature on slaughterhouses, slaughterhouse
communities, and relevant aspects of other literatures, such as community sociology and
criminology. The aim of this review is to situate this study within the larger literature
and to elucidate the gaps in the literature which this study is designed to ﬁll. This chapter
begins with a brief examination of the social history of the slaughterhouse, with particular
attention paid to the attempts made to obscure the mass slaughter of animals from the
gaze of the public. Then the transition to the modern system of meatpacking is
examined, focusing on relatively recent changes in the structure of slaughterhouse labor
and the logistics of meat processing and production (despite these changes, Sinclair’s
description of the slaughterhouse as a jungle remains eerily apt). The effects of these
industry changes on the environment, the workers, and the community are subsequently
detailed. Special attention is paid to the social consequences of slaughterhouses,
especially documented increases in crime in the ethnographic community studies
conducted thus far. Finally, the theorized causes of the observed increases in crime
proposed in the literature are examined and critiqued, and an alternative theory is

proposed.

The History of the Slaughterhouse

The agricultural revolutions

It is necessary to ground an understanding of the effects of slaughterhouses on modern
society in an historical understanding of agricultural developments and the birth of the
modern slaughterhouse. Stull and Broadway (2004, chapter 1) explain that there have
been three revolutions in agriculture (which can be deﬁned as the production of crops,
livestock, or poultry). The ﬁrst revolution occurred in Southeast Asia 10,000 years ago
with the development of seed agriculture and the domestication of animals. The second
revolution was ushered in by industrialization in the late eighteenth century in Western
Europe. This revolution replaced subsistence agriculture with a new agricultural system
based upon surpluses and proﬁts. Simultaneously, urbanization created a commercial
market for food. The third revolution, which we are currently in the midst of, originated
in the United States at the beginning of the twentieth century and is characterized by
mechanization, chemical farming, and food manufacturing, adding value to agricultural
products through processing and packaging. Throughout these phases in agricultural
development great effort has been made to keep animal slaughter hidden from the public

gaze.

Banishing the slaughterhouses from society

During medieval and early modern times public authorities tried to keep the slaughter of
animals out of public places, attempting to move the activity outside of town walls
(Thomas 1983, 294) to keep the activity out of sight. Later, during the early nineteenth
century, arguments for relocating a livestock market outside of the city of London
focused not only on health and safety issues but also on the claim that its proximity to the
population of London posed a moral danger to the populace. Philo (1998, 63) cites a
shop owner who stated in response to a committee investigation in 1849 that “the chief
trades ‘encouraged by the existence of Smithﬁeld’ [the meat market] were ‘gin shops and

3”

public houses . This individual also discussed “the locality’s ‘degeneracy’ and a claim
about its ‘respectable’ citizens wanting to see the market’s removal’” (cited in Philo
1998, 63). Another man interviewed by the committee said that the violence against the
animals “educate[d] the men in the practice of violence and cruelty, so that they seem to
have no restraint on the use of it” (Philo 1998, 65). In the US, slaughterhouses were
also removed from the concentrated populations of the cities and the more densely
populated eastern states (Stull and Broadway 2004). Philo (1998) argues that this
sentiment is based upon the assumption that those who work with animals will become
debased, like the animals themselves. A distinctly different explanation is that such work
with animals entails and perhaps requires becoming desensitized to the suffering of both
animals and other humans.

Due to the immediacy of their involvement in acts which ‘offended sensibilities’,

butchers in particular were shamed. During the early modern period they were even

16

considered unsuitable for jury duty on capital cases because of their ‘cruel inclinations’
(Thomas 1983, 295). Some social thinkers during this time speciﬁcally stated that those
whose work entailed killing and butchering animals were morally compromised. For
instance, in 1748 David Hartley pointed to the “frequent hard-heartedness and cruelty
found amongst those persons whose occupations engaged them in destroying animal life”
(cited in Thomas 1983, 295) and Adam Smith referred to the work of butchers as “a
brutal and an odious business” (cited in Thomas 1983, 295). Vilifying those who labored
to produce the large amounts of meat demanded by the public (as opposed to those
producing small amounts for personal consumption) served to assuage the conscience of
the consumer and contribute to the rationalization of slaughter.

The historical practice of slaughtering animals out of the sight of the public,
which has today been perfected, also serves to keep the consumer’s hands and conscience
clean (Serpell 1986; Smith 2002, 50). Serpell (1986) describes this as concealment, one
of four categories of distancing devices used to mitigate guilt associated with the banning
of animals. The other three categories he delineates include detachment,
misrepresentation, and shifting the blame. The following statement made by Bataille in
his discussion of slaughterhouse architecture is illustrative of the concealment that occurs
in the siting of slaughterhouses:

Today, the slaughterhouse is cursed and quarantined like a boat
carrying cholera. In fact, the victims of this curse are not butchers
or animals, but the good people themselves, who, through this, are
only able to bear their own ugliness... The curse (which terriﬁes
only those who utter it) leads them to vegetate as far as possible
from the slaughterhouses. They exile themselves, by way of

antidote, in an amorphous world, where there is no longer anything
terrible (1997, 22).

I7

The hyperseparation of the public from the slaughter of the animals they consume
(detailed eloquently by Adams (1991)) is a marked characteristic of the third revolution
of agriculture. Despite the desire to separate slaughter from the eyes of the public
throughout history, it is not until the third revolution that this separation really became
institutionalized, which is likely related to the process of mechanization and mass

production.

The Modern Meatpacking Industry

Meat as mass production

Just prior to the emergence of the third agricultural revolution and throughout its
development, drastic changes took place in the process of slaughtering animals for food.
In the US these changes began in 1865 with the development of the Union Stock Yards in
Chicago, which made meatpacking a major industry in the country. By the following
year there were 100 miles of railway connecting the Union Stock Yards to supplies of
livestock (Cronon 1991; 2002).

In response to the growing population’s increased demand for meat and the
increased volume of livestock entering the Stock Yard, the conveyor belt was introduced
to increase production speed and efﬁciency. Importantly, this new conveyor system took
control of the speed of production away from the workers and put in the hands of the
supervisors (Patterson 2002, chapter 3; Stull and Broadway 2004, chapter 2). According

to some, meatpacking became the ﬁrst mass-production industry in the United States,

from which Henry Ford is said to have at least partially adapted his conception of
assembly-line production (Patterson 2002, chapter 3). According to Cronon (1991), the
industrialized packing plants represented an important break with nature: it distanced the
consumers from the animals, the act of killing, and the natural environment from which
the animals had grown. The industry continued to expand during this period as a result of
increasing demand and increased distribution possibilities (Patterson 2002: chapter 3).
Many of the workers in the stockyard lived in the back of the yards where a slum
developed. This slum was characterized by extreme poverty, crowded conditions,
delinquency, and environmental pollution. The stockyard community, which experienced
growth up until World War 11, became home to nearly 60,000 people, about half of whom
had emigrated from other countries (Jablonsky 1993). The miserable working and living

conditions fostered pro-union sentiments.

Unionism in the meatpacking industry

Industry expansion and harmful working and living conditions gave rise to labor unions.
During the ﬁrst two-thirds of the twentieth century labor unions became increasingly
powerful in the meatpacking industry, even as unions in other industries suffered.
Brueggemann and Brown (2003) have asserted that in spite of major obstacles to
organized labor at the time (such as deindustrialization and globalization, rising
afﬂuence, technology, de-skilling, and working-class factionalism), there was widespread
support for the meatpacking unions, strong rank-and-ﬁle activism, and a strong sense of

purpose among the workers. Beginning in the 19303, the United Packinghouse Workers

of America (UPWA) and the Amalgamated Meat Cutters (AMC) worked hard to
unionize slaughterhouse employees. Reportedly, by the early 19603 these two unions
represented more than 95% of the slaughterhouse employees outside of the Southern
states. As a result, meatpacking became one of the best-paid industrial occupations.
During this period the industry was dominated by the ‘old Big Four’ meatpacking
companies — Swiﬁ, Armour, Wilson, and Cudahy.

The power of the unions in the meatpacking industry began to wane in 1969
(Brueggemann and Brown 2003), and by the end of the 19903 only 60% of
slaughterhouse workers were unionized (Bacon 1999). Numerous reasons have been put
forth to account for the decline of the unions in the meatpacking industry. Brueggemann
and Brown (2003) have usefully grouped these factors into three categories: economic
restructuring, working class fractionalization, and employer ascendancy. These
categories are discussed below since a full understanding of contemporary developments
in the meatpacking industry requires attending to the decline in unionism.

The ﬁrst explanation for the decline of unionism within the meatpacking industry
points to the economic restructuring in the industry caused by postindustrialism and
globalization. The effects of postindustrialism and globalization on the agricultural
industry in general have been examined (for instance, see Bonanno, Busch, Friedland,
Gouveia, and Mingione 1994), and some have had profound impacts on the meatpacking
industry. These effects have included job de-skilling as a result of the automation of
production, a reduction in the dependence of employers on experienced manufacturing
workers, reduced pay, a more transient and difﬁcult workforce to organize, and the

lessened ability of unions to win concessions through collective action. As Brueggemann

20

nd Brown (2003) point out, however, the major threat of economic restructuring to
rganized labor in the meatpacking industry has been the redistribution of jobs within the
’nited States instead of internationally, which makes the industry somewhat unique.
[eatpacking plants have shiﬁed geographically from mainly urban areas in the North
ith strong traditions of unionism to more Southerly rural areas where opposition to
lions is more normative, in contrast to the general trend of manufacturing companies
oving from the global north to the global south.
A second perspective on the decline of unions in the meatpacking industry points

working class fractionalization, fostered by the increasing representation of women
:1 men of color in the industry and the animosity between these groups when layoffs are
plemented during economic downturns. As the industry grew and exhausted the pool
reserve white male workers it was increasingly necessary to recruit people of color and
men. For instance, from 1890 to 1919 the white male labor pool dried up — at least
tially due to World War I — and the proportion of women working in the industry
reased from 2.7% to 10.5% (Horowitz 1997). Minority workers were notably
'uited during periods of strikes to work as scabs (Halpern 1997), which further
amed worker factionalism.

The women and people of color who were able to gain employment in the
.tpacking industry before the Depression, however, were subsequently subjected to
st and sexist discharge practices during the economic crisis, resulting in many of them
g laid off from their jobs to make room for white male employees. Later, in the
)3 through the 19605, African-American women in the United Packinghouse Workers

merica union developed coalitions to address discrimination based upon race and sex

21

the industry. They fought to desegregate departments and locker rooms in the
mpanies, but encountered resistance from the management, the male employees, the
lite female employees (Fehn 1998), and even from their unions (Horowitz 1997),
istrating the type of fractionalization that has affected the industry.

As the slaughter of food animals has become increasingly removed from the
olic gaze the work has become somewhat mystiﬁed in society, its realities ignored and
)pressed, and the work increasingly relegated to those who have little power and few
ployment opportunities. Racial and ethnic minorities are now the numerical majority
ployed in slaughterhouses. In 2003, for instance, 4.1% of meat, poultry, and ﬁsh
icessing workers were Asian, 12.7% were Black, and 41.5% were Hispanic (US
nsus Bureau 2003). Labor within the industry also remains gendered (Horowitz 1997).
of 2003, only slightly more than a quarter (26.6%) of meat, poultry, and ﬁsh
;sessing workers were women (US Census Bureau 2003). The diverse and ﬂuctuating
nographic composition of the industry has posed numerous challenges to labor
anizing.

The ﬁnal contributing factor to the decline of unionism in the meatpacking
ustry that Brueggemann and Brown (2003) delineate is the ascendancy of the
yloyer companies, which refers to the increasing power of capitalists relative to that of
working class, a shift which began after World War II and continues today. The ‘old
Four’ companies (Swiﬁ, Armour, Wilson, and Cudahy) that had dominated the
itpacking industry (and had been well penetrated by the unions) lost ground during the
3nd half of the twentieth century to smaller companies. The shares of sales controlled

:he Big Four companies declined from 52% in 1950 to 25% in 1972. The original Big

22

 

3'3?

I-4-‘

 

 

 

11' companies eventually gave way to a new Big Four — Iowa Beef Processes (IBP),
iAgra, Excel and Beef America (Brueggemann and Brown 2003). By the year 2000,
new Big Four companies controlled more than 81% of beef slaughter in the United
tes (Stull and Broadway 2004, 15). This monopoly has made the increasing
endancy of the employer companies virtually inevitable. One company in particular,
’, has been powerful enough to forever alter the ways in which the meatpacking

npanies do business.

a IBP revolution

’ (which was purchased by Tyson Foods in 2001) is the largest red meat provider
bally (Broadway and Stull 2005; Hake and King 2002; Olsson 2002),2 and has been
ticularly powerful in reshaping the industry and undermining the labor unions
ueggemann and Brown 2003; Stull and Broadway 1990). The company has taken
ee steps that have substantially altered the meatpacking industry: developed new
hnologies, changed the geography of production, and obtained cheaper labor. Their
:t step has been to aggressively pursue increased automation, new technologies, and
DCIUCIIOII processes that would reduce labor costs and increase proﬁts. Before
icribing some of the technologies introduced it is necessary to note that although gross
at consumption is up, the industry itself is not especially lucrative.

Despite the fact that per-capita consumption of beef has dropped since the 19703,
: gross amount of meat consumed by the entire population has continued to rise.

cording to USDA Agricultural Statistics summarized by Stull and Broadway (2004,

23

 

er 1), per capita beef consumption in the US. peaked in the late 19705 at
>ximately 126 pounds per year and was down to 99.3 pounds per year in 2000.
ever, pork consumption has remained fairly constant, and there has been an

losive growth” in the consumption of chicken. In 2002, meat and poultry

umption in the US. reached its highest level — 219 lbs per person (Markus 2005).
increase in overall meat consumption, despite the publicized associated health and
ironmental consequences of meat eating, has been facilitated by the low cost of meat
.ich has been fostered by the mechanization of meat processing, the increasing
nomies of scale, and the continued decimation of organized labor in the industry).

ten adjusted for inﬂation the price of meat has actually dropped, reaching the lowest
ce in 50 years in the 19905. The increased demand, coupled with the decline in
ces/proﬁtability, has resulted in faster production, or increased chain speeds (Stull and
oadway 2004, chapter 1). Since meatpacking is not an exceptionally proﬁtable industry
for each $100 in sales, $93 goes to production costs — these meatpacking companies
we become extremely competitive, continually seeking to reduce their costs and

lcrease their production (Stull and Broadway 1990). IBP has been at the forefront of
movating in order to increase proﬁts.

Illustrative of IBP’s innovating is the development of ‘boxed beef , which has
educed both labor and shipping costs. Instead of hanging and transporting sides of meat,
he fat and bone are removed and the meat is vacuum packed and boxed up. Working
with boxed beef makes distribution more efficient, cheaper, and reduces the skills
required by labor (Brueggemann and Brown 2003; Stull and Broadway 1990). Proﬁts

have also been increased by increasing the speed of the ‘chain’ (Eisnitz 1997; Stull and

24

 

lvvay 1990), or the rate at which the animals are stunned, killed, and
:mbered/processed; and by maximizing economies of scale — creating plants that can
hter greater numbers of animals (Broadway and Ward 1990). This trend is

:nced by the increase in the number of large slaughterhouses. Between 1974 and

', the number of slaughterhouses employing more than 1000 workers doubled, while
lumber of plants employing fewer than 1000 workers dropped signiﬁcantly

iadway and Stull 2005).

IBP further altered the industry by changing the geography of production
reggemann and Brown 2003). In the United States and Canada slaughterhouses were
ginally located in densely populated urban areas, close to their markets and railways
d to transport livestock. However, the importation of live animals to the
rghterhouses in urban centers resulted in a loss of value because of “shrinkage” (cows
e 5% of their weight in only 3 hours of shipping), bruising, and crippling during
pping (Stull and Broadway 1990). With improvements in refrigeration and the
pularity of boxed beef, the slaughterhouses could be relocated to communities with
raller populations, near the feedlots, to reduce production costs.3 (The location of hog
ocessing has not undergone as dramatic a regional shift as cattle processing because the
nning of pigs is still tightly tied to the production of corn in the western corn belt
Broadway 1995)). Most of the towns where IBP purchased or constructed plants had
opulations of less than 25,000 (Broadway 1998). This relocation also reduced
roduction costs by reducing the costs of labor, permitting the industry to move from the
ndustrial urban contexts (that had given rise to the unions) to ‘right-to-work’ states,

vhich had historically outlawed unionized plants, thus providing secure sources of

25

 

)er, non-unionized labor (Broadway 1998; Broadway and Stull 2005; Hake and King
). Today, however, these laws have been weakened, instead giving employees the
in to join unions, pay dues, and quit the unions at any time (National Right to Work
LI Defence Foundation 2005), which nonetheless hampers union organizing
adway and Stull 2005), keeping labor costs suppressed. This geographic shift in
luction also undermined the previous master labor agreements — agreements that
med across companies and established industry-wide wage and beneﬁt scales
reggemann and Brown 2003; Stull and Broadway 2004, chapter 1). This geographic
t has not only had negative consequences for organized labor, it has also “been a
Led blessing for small towns where packing plants have located” (Broadway 2000, 37),
vill be expanded upon shortly.

Finally, and related to the previous two steps that IBP took, the company sought
v sources of cheap labor. The technological innovations they implemented made it
ssible to employ a less skilled workforce, and by locating in small communities where
:re was not a reservoir of labor available to meet their needs they were able to take
vantage of the recruitment of immigrant workers for less pay (Brueggemann and
'own 2003), actually facilitating the entrance of immigrants into communities that had
I recent history of immigrant settlement (Gozdziak and Bump 2004). For instance, IBP
)ened a new meatpacking plant in Garden City, Kansas in 1980, and by 1985 the
)pulation had grown by 33%. The majority of the new residents were Southeast Asian
:fugees and Latinos, many of whom were from Mexico (Stull and Broadway 1990).

Cotenninous with the changes in technology, the shift to rural areas in right-to-

v'OI‘k states, and the recruitment of new sources of labor, the wages in the industry

26

eclined, and production increased (Skaggs 1986). Wages in the meatpacking industry,
*hich had once been the highest of manufacturing industries, dropped to 20% below
:neral manufacturing work by 1990 (Stull and Broadway 2004, chapter 5). The
creasing ethnic and racial diversity of the workforce in the industry has led to increased
actionalization and made union organizing even more difﬁcult logistically due to
lguistic and cultural differences (Brueggemann and Brown 2003), making labor
ganizing and wage increases less likely. It has even been asserted that the INS is being
:d as a tool in union busting by the industry (Bacon 1999). These steps taken by IBP
increase their proﬁts have not only resulted in reduced union protection in their own
nts, but have also placed pressure on their competitors to reduce their production and
or costs. IBP’s actions have resulted in an industry-wide shift in the production

cess, the location of production, and the characteristics of the workforce.

The Effects of the Contemporary Meatpacking Industry

3e drastic changes in the industry, pioneered mostly by IBP, have attracted the

ition of scholars who have begun to document their effects. The literature on rural
:packing began in the early 19905, and has grown since (Grey 1999). Most of the
lI‘Ch has been conducted by anthropologist Donald Stull and geographer Michael
dway. Funded by a Ford Foundation grant, they led a team of social scientists who
IIIICCI meatpacking towns in the Midwest. They appear to be the ﬁrst social scientists

idy the modern meatpacking industry and its impacts on workers and communities

27

III and Broadway 2004). The proliﬁc group of scholars has published more than 50
cles (Horowitz 2005); however, the literature on slaughterhouses remains ‘scattered
l disjointed’, with publications often narrow in focus, examining one segment of the
ustry, one community, a particular plant, or the history of unionization in the industry
ull and Broadway 2004).

As a result of his research in the area, Broadway (1994) has outlined the
lowing ten likely impacts slaughterhouses moving into an area: increases in minority
irkers, increase in low-paying jobs, offensive odors, increases in demand for low-cost
using, strains on local infrastructure, increases in crime, increases in persons utilizing
:ial services, increases in homeless individuals, strains on health-care providers, and
:reased language and cultural differences. These impacts can be grouped into three
tegories: the impact on the physical environment and human health, the impact on the
)rkers, and the social impacts on communities (these effects are not, of course, mutually
elusive). Each of these areas is examined below; however, because this study is
tuated mainly within the latter segment of the literature, it will be examined more

oroughly.

ffects on the physical environment and human health

arallel to the development of the modern, high-volume, decentralized slaughterhouses

)cated closer to the supply of livestock, a shift occurred from raising livestock on small

) medium size family farms to producing livestock on massive farms, colloquially

eferred to as factory farms and referred to in the literature as Concentrated or Conﬁned

28

imal Feeding Operations (CAFOs). Between 1982 and 1997 the number of CAFOs in
United States dropped from 435,000 to 213,000. The reason for the decline is that
aller operations went out of business as larger operations continued to swallow them
(Stull and Broadway 2004). Thus, while there has been an increase in the number of
Lmals being raised or produced, there has been a decrease in the number of facilities
iere these activities take place. Because of the increasing physical proximity between
: raising and the slaughtering of the animals, concerns about the physical
vironmental consequences of raising livestock have tended to encompass the
iughtering industry as well.
Negative impacts upon the air and water quality have been documented in regions

iere growth in the livestock industry has occurred (Caldwell 1998). Stull and
roadway (2004, chapter 1) explain that much of the problem is caused by the amount of
anure that these large operations produce. The nitrogen and phosphorous contained in
anure can be extremely dangerous to the environment and human health when they
iter the water systems in large quantities. This is not a new concern. In the 19th
entury London’s Smithﬁeld Flesh Market was closed as a live animal market because of
.e unregulated disposal of offal and blood which was believed to be a cause of cholera
rtbreak in the area (Kalof forthcoming). Today, environmental justice movements to
:ep CAFOs out of communities have emerged. These movements have challenged the
ghts of large corporations to come into their communities, (which are predominantly
rm] and economically disadvantaged), establish large CAFOs, and put pressure on local
irmers to establish CAFOs in order to compete with the larger corporations (Stull and

roadway 2004, chapter 9). In addition, The Farmers Union has appeared before the

29

.8. Senate, arguing that livestock concentration (spearheaded largely by the large
eatpacking companies) is negatively affecting their livelihoods and as one member
ited, it is ‘sucking the lifeblood out of rural communities’ (National Farmers Union
ews, 2002). Some communities, however, have been able to successfully remove the
AF OS from their areas (see DeLind (1998) for instance).

In addition to the dangers posed to human health in the form of pollution caused
' the industry, there is also the danger posed by food poisoning. An increased demand
r meat, coupled with the decline in prices/proﬁtability,4 has resulted in faster
aduction, or increased chain speeds (Stull and Broadway 2004, chapter 1). The Health
d Safety Director of the United Food and Commercial Workers union has reported that
ain speeds increased between 50% and 80% in the 10 years between approximately
82 to 1992 (Stull and Broadway 1995, 68). To put this in perspective, in the early
705, the fastest line killed 179 cattle an hour; today the fastest kills 400 per hour on
:h line. In Europe, however, only approximately 60 cattle are killed an hour (Markus
05). It is claimed that the increasing speed of production in the United States makes
: contamination of the meat during processing more likely (Eisnitz 1997, chapters 4
i 13; Stull and Broadway 2004). Pathogens such as Campylobacter, Salmonella, and
:herichia coli 0157:H7 have been documented entering the food supply (Stull and
)adway 2004, chapter 1). While it is unproven that food poisoning is caused by meat
itamination during processing, there has been a sharp increase in food poisoning
lIhS that corresponds roughly to the increase in the chain speeds and CAFOs. In the
year span from 1984 to 1994, deaths from food poisoning more than quadrupled from

00 to 9,000 cases (Eisnitz 1997).

30

 

ughter and the worker

teasing chain speeds are not only a potential source of meat contamination, they also
ISIIIUIC a safety hazard for the approximately 150,000 workers employed in the
atpacking industry. The industry became the most dangerous in the US by the 19805
‘oadway and Stull 2005; Olsson 2002, 12; Stull and Broadway 1990) -— more

igerous than other notoriously dangerous industries, such as mining. The reported
ury rate in the industry has fallen since the early 19905, which can be at least partially
ributed to advances in ergonomics and the desire of the companies to reduce the costs
worker compensation and of ﬁnes for safety violations (Broadway and Stull 2005).
IWCVCI’, at the close of the decade the reported injury and illness rate remained quite
gh: In 1999, the reported rate was 26.7 injuries/illnesses per 100 full-time workers,
fee times the average for industries manufacturing other commodities (Stull and
“oadway 2004, 75).

The type of work undertaken in slaughterhouses lends itself to more injuries and
nesses in general. The use of sharp knives in the dismembering/processing of the
limals coupled with the fact that too frequently animals are improperly stunned and
gain consciousness (Eisnitz 1997) results in the potential for many accidents in the
orkplace. Additionally, much of the work that is done in slaughterhouses is repetitive,
hich can lead to muscle strain and cumulative trauma disorder, such as carpal tunnel

mdrome (Stull and Broadway 1990, 15).

31

The safety of slaughterhouse workers has increasingly come to the public’s
ention, as illustrated by the following statement made in a recent editorial in The New
rk Times: “What is most alarming at the slaughterhouse is not what happens to the
imals — they have already met their fate. It is what happens to the humans who work
:re” (author unknown 2005). The competition in the industry and narrowing proﬁt
trgins (Stull and Broadway 1995) have resulted in cranking up the speed of the line,
rich has compromised both the safety of the food and the safety of the workers
aducing the food (Olsson 2002). A faster chain speed translates into more
Iportunities for accidents and increased repetitive movements.

Additionally, the meatpacking industry has an exceptionally high employee
mover rate: “[W]hen a plant starts up it is not uncommon for employee turnover
nong line workers to exceed 200% in the ﬁrst year of operation” (Broadway 2000, 39).

Lexington, Nebraska, within the ﬁrst 21 months after the plant opened the turnover
te was 250%, or 12% each month. An Excel plant opened in Dodge City, Kansas
:perienced a 30% monthly turnover rate and an IBP plant opened in F inney County,
ansas saw a monthly turnover rate of 60% (Gouveia and Stull 1997). The high turnover
.te is said to actually beneﬁt the industry (Grey 1999; Grey and Woodrick 2002; Stull
1d Broadway 1990), in spite of the fact that it results in less experienced workers and
lore accidents, because it keeps the costs of wages and beneﬁts down. The high
imover rate has been attributed to the dangerous working conditions and the physically
emanding nature of the work (Stull and Broadway 1990), but the effect that the
motional toll of slaughtering and dismembering animals might have on the turnover rate

as not been mentioned in the literature.

32

Ascertaining the exact turnover rate is difﬁcult: “Industry spokespersons do all
y can to avoid revealing turnover rates, but everyone agrees that employee turnover is
her than virtually any other industry” (Stull and Broadway 2004, 80). Illustrative of
5 high turnover rate is the fact that only 48 out of 15,000 hourly workers at IBP
:eived retirement beneﬁts between 1974 and 1986 (Stull and Broadway 1995, 70), and
>orted1y one-third of slaughterhouse workers quit within the ﬁrst 30 days (Stull and
oadway 2004, 80).

The high turnover, whatever the exact ﬁgure is, negatively impacts worker safety
cause it results in a high number of inexperienced laborers working in slaughterhouses,
nich compromises not only their own safety but also the safety of those around them.
re effects of the lack of experience on safety is exacerbated by the fact that the vast
ajority of slaughterhouse workers are immigrants (Olsson 2002), who because of
nguage barriers may not receive adequate training and who might be reluctant to report
ifety violations. Simultaneously, making the situation of the workers even more critical,
ispections by the Occupational Safety and Health Administration (OSHA) have declined
iarkedly, drOpping to a record low by the late 19905 (Olsson 2002), a trend started
uring the Reagan administration when the number of OSHA enforcement workers and
ispections were reduced (Claybrook, cited in Stull and Broadway 1995, 65).

In light of the impediments to worker safety in the industry, the following
uggestions have been made: longer and improved worker training; more adequately
itaffed work crews; varying the jobs in order to reduce muscle strain; longer recovery

)eriods for injured workers; and above all else, slowing down the speed of the chain

33

tull and Broadway 2004, chapter 6; Stull and Broadway 1995). Signiﬁcant changes

nain to be seen.

aughter and the social impacts on the community

examining the issues of health and safety among slaughterhouse workers a few

:holars have noted that there are important social consequences in communities where
aughterhouse facilities are sited, such as a shortage of housing, increased demand for
)cial assistance, and an increase in crime (Broadway 2000; Stull and Broadway 2004) —
7 general social disruption. The increased demand for housing and social assistance can
3 explained by the inﬂux of people into the community looking for work in the
aughterhouse(s). The increase in crime rates, however, is the least readily explainable
fthese social problems. Upton Sinclair (1946, 18-19) seemed to have hypothesized a
mnection between crime and the type of work undertaken in slaughterhouses: “men

ho have to crack the heads of animals all day seem to get into the habit, and to practice
1 their ﬁiends, and even on their families, between times”. This potential connection

15 not been addressed in the contemporary literature on slaughterhouses. I will expand
Jon this possibility shortly, but ﬁrst it is necessary to provide an overview of the

udies that have documented crime increases in slaughterhouse communities and

rbsequently examine the causes that have been hypothesized in the literature.

34

udies documenting increased crime in slaughterhouse communities

their recent book entitled Slaughterhouse Blues (2004), Donald Stull and Michael
oadway report that in Finney County, Kansas there was a 130% increase in violent
lmes within ﬁve years after two slaughterhouses opened, which can only be partly
counted for by the 33% increase in population (Broadway 2000). Property crimes in

3 county also increased, and the incidence of child abuse increased by three times and
15 50% higher than the state average (Gouveia and Stull 1995). Increases in crime in
tughterhouse towns have also been observed in Nebraska (Broadway 1994): In
:xington, monthly police booking increased 63% over a three year period (Gouveia and
ull 1995). Crime rate increases have also been documented in Iowa: crimes increased
Pen'y (Broadway 1994) and in Storm Lake, where the number of burglaries in the ﬁrst
ne months of 1992 was four times that of the previous year (Grey 1995) and by 1994
rious crimes reported were 2.5 times greater than in other Iowa cities of similar size
lrey 1998). Crimes also increased with slaughterhouses in Oklahoma: in Guymon total
rests increased 38% (Stull and Broadway 2004). Finally, increases in drug-related
iminality have also been documented in at least one poultry-processing town —
:orgetown, Delaware (Horowitz and Miller 1999).

Increases in crime rates after the opening of a slaughterhouse have also been
lSCl'VCd in at least one Canadian community to date: the town of Brooks, Alberta,
;perienced a 15% increase in population within approximately 5 years of plant
:pansion but also witnessed a 70% increase in reported crime (Broadway 2001; Stull

Id Broadway 2004, 123-124). The town of High River, Alberta, which hosted a new

35

laughterhouse, has not experienced the negative impacts that Brooks has, presumably
ecause it is close enough to Calgary so that many of the workers live there in order to
ave access to more affordable housing (Broadway 2001).

Broadway and Stull have also noted changes in the crime rates in such
immunities over time, most speciﬁcally in Garden City, Kansas. A meatpacking plant
pened in Garden City in 1980. Increases in violent crime were apparent immediately
ter the plant opened, but the crime rate surged in 1985 and 1986 after the rapid
)pulation growth (Broadway 1990). The reported crime rate peaked in 1994 and in
998 it fell to its lowest ﬁgure in a decade. Subsequently, the crime rate increased until
)01 and then fell in 2002 and 2003. Despite these declines in report rates in 1998, 2002,

Id 2003, the arrest rate actually increased (Broadway and Stull 2005).

creases in family violence

'oadway and Stull (Broadway 1990; Broadway 2000, 40; Stull and Broadway 2004,

'3) have noted that most of the increase in violent crime that they have witnessed in
mmunities where slaughterhouses are sited is the result of an increase in domestic
)lence. Gouveia and Stull (1995) state that a newspaper report in Lexington, Nebraska
inted to domestic violence (along with property crime) as accounting for the greatest
)portion of the increasing crime rate. Additionally, Finney County Kansas reportedly
perienced a tripling in the incidence of child abuse from 1980 to 1985 (Broadway

90). However, within the literature there does not appear to be much of a discussion of

iy these forms of violence in particular would increase in these communities, with the

36

xception of Broadway’s (1990; 2000) statements that increased alcohol consumption is a
ontributing factor to the increase in domestic violence cases. (Broadway (2001) also
oints to alcohol consumption as being a causal factor in the increases in general crime.)
his study examines rates of family violence (inclusive of both partner and child abuse),

1 counties with and without slaughterhouses in order to ascertain if this increase in

unily violence is signiﬁcant and if it is unique to slaughterhouse communities. This

udy also examines the rape rates in order to ascertain if perhaps the increase in violent
'imes in slaughterhouse communities is the result of the victimization of particularly
Jlnerable populations — women and children. Prior to discussing this alternative
ypothesis, however, it is necessary to examine the causes that have been proposed in the

terature and will be tested in this study.

heorized causes of increased crime in slaughterhouse communities

re explanations proposed in the literature for the increase in crime rates in

aughterhouse communities can be grouped into three categories: explanations based on
e demographic characteristics of the workforce, explanations based on population

ioms and theorized social disorganization, and explanations that point to unemployment
a cause. These categories are not intended to be mutually exclusive; rather, they are
eful in organizing and understanding the causes proposed in the literature. Each
tegory is discussed below, paying particular attention to what has been proposed in the
aughterhouse literature (these theories remain untested there), and what has been found

ing the theorized variables in general studies of crime.

37

 

Crime as a result of the demographic characteristics of the workforce

The transition to the use of immigrant labor has been a profound development in the
meatpacking industry in particular (Iowa State University Extension 1998). In fact,
available year-round employment in the food-processing industry is said to be the most
signiﬁcant draw of immigrants to the rural Midwest. This draw has resulted in

signiﬁcant demographic changes in the region, marking new migratory streams (Hagan
2004). Illustrative of these changes is the fact that in 1980 there were 1.2 million
iispanics in the 10 Midwestern states, and by 1992 the number had reached 1.8 million
Dalla, Ellis, and Cramer 2005). Martin, Taylor, and Fix (1996) note, however, that
rany of these settlers are not international immigrants, but rather are Latinos and Asians
roving from other parts of the United States.

This signiﬁcant change has been met with hostility in some areas. Immigrants
ho relocate to communities to work in slaughterhouses are often scapegoated by the
rneral public, the media, government ofﬁcials, and the meatpacking industry itself, in an
:empt to explain away the resultant social disruption in communities where

.ughterhouses have been sited.

One change commonly noted by residents is the rapid increase in
crime. Some speculate that crime naturally increases as the number
of people increases. Others feel the immigrants are lawless and
deﬁant to authority. Prejudice and racism against immigrant
newcomers fuels cultural clashes. Violent acts including cross
burnings, racial slurs, and physical conﬂict have been reported in
Iowa, Nebraska and Kansas (Dalla, Ellis, and Cramer 2005, 168).

h xenophobic explanations for increases in crime have not been limited to the general

‘i c: “politicians, pundits, and pseudo-academics have suggested that the newest

38

immigrants are a crime-prone group, a claim seized on by the media” (Lee, Martinez, and
Rosenﬁeld 2001 , 567). For example, the media coverage of INS raids in slaughterhouses
in Iowa portrayed the Latino immigrants as criminals and paid particular attention to an
alleged connection between the immigrants and drug trafficking, reinforcing negative
stereotypes (Grey and Woodrick 2002, n3; 373). Immigrant labor has also served as
scapegoat in Nebraska, where slaughterhouses have recently become highly concentrated:
I group of Nebraska police ofﬁcers and the state’s Congressional delegation contacted
he INS Commissioner in Washington with concerns over the increased crime rates,
vhich they attributed to the increase in immigrants in their communities (Bacon 1999).
ind in Buena Vista County, Iowa, an assumed link between immigration and crime
ecame the central issue of the 1994 election for the county attorney position (Grey
998). The challenger to the 16 year incumbent made the slaughterhouse industry’s
ring practices a central theme of his campaign, and accused IBP of ‘social pollution’.
e won the primary.

Some meatpacking companies, attempting to calm fears about the ‘type of people’
3y will bring into communities, have even made promises before opening plants that
:y will employ locals and not recruit from outside of the community. Of course, after
: local labor pool has been exhausted (which is expedited by the high turnover rates),
ployees are recruited from outside of the area (Gouveia and Stull 1997).5

Although not yet tested speciﬁcally in slaughterhouse communities, general
dies of cities have not supported the contention that immigration increases crime. In
ir study of the impacts of immigration on homicide rates in three cities, Lee, Martinez,

, Rosenﬁeld (2001) found that generally immigration does not increase homicide rates

39

 

for Latinos and African Americans. In another study, Nielsen, Lee, and Martinez (2005)
found that in general, the percent of recent immigrants had negative effects or no effects
at all on most motive-speciﬁc homicides for Latinos and blacks in Miami and San Diego.
The only exception was in Miami for the Latino group, where greater percentages of
recent immigrants had a positive effect on the number of intimate homicides. They
conclude, “[t]hese results may indicate that the neighborhoods into which newcomers
settle are in fact revitalized or stabilized by the presence of immigrants, arguments
consistent with the immigrant revitalization perspective rather than with the premise that
recent immigrants disorganize communities” (Nielsen, Lee, and Martinez 2005, 863).
The notion that immigration leads to increases in crime is consistent with the social
lisorganization theory, since it is believed that population heterogeneity and population
nﬂuxes result in the weakening of social institutions and resultant crime. That theory
vill be discussed shortly.

In the academic literature, attention has been paid to other demographic
iaracteristics of workers recruited to work in slaughterhouses as an explanation for the
creases in crime: “Increases in drug and alcohol abuse and an overall increase in crime

meatpacking towns has been attributed to the recruitment [into the meatpacking
iustry] of young, adult, single males, since this demographic group has the highest
:idence of committing crimes and alcohol consumption” (Broadway 2000, 40). Thus,
the academic literature, the age and gender of the workers has been posited as posing
increased criminogenic risk (Broadway 1990; Broadway 1994; Broadway 2000;
iadway 2001; Broadway and Stull 2005; Stull and Broadway 2004). The connection

ween the sex ratio, age structure, and crime, however, has not been as clear cut in

40

 

general studies of crime as has been assumed here, and these factors might not be as
applicable to slaughterhouse communities as assumed either (see Chapters 4 and 5).

The notion that most of the individuals who come to slaughterhouse communities
0 Work in slaughterhouses are single males is not necessarily accurate. It is true that
generally immigration for work follows the following steps: 5010 men are recruited or
ome to an area for work; later their families follow; and subsequently unauthorized
nmigrants might follow, using social networks with individuals already settled in the
rea to ﬁnd employment (Dalla, Ellis, and Cramer 2005; Martin, Taylor, and Fix 1996).
his pattern certainly characterizes migrant farmer communities. However, in many
1568 the immigrants moving to ﬁll jobs in meatpacking plants are not migrant farm
orkers, and there are fewer solo males and more families in meatpacking towns than in
igrant farm worker towns because unlike migrant farm work, meatpacking jobs offer
ar-round employment and enough money to make supporting a family more feasible
Iartin, Taylor, and Fix 1996).

The effect of age on crime is also not straight forward. The number of individuals
ween the ages of 15 and 29 has been found to be signiﬁcantly related to motor vehicle
ft and homicide rates nationally (Cohen and Land 1987). However, at least two other
:lies have come to different conclusions. One study of nonmetropolitan counties found
: while the number of people between the ages of 15 and 29 was a powerful predictor
obbery and motor vehicle theft, it was not signiﬁcantly related to homicide and
glary (Lee and Ousey 2001), and another study of census tracts in three cities found
the percentage of males between 18 and 24 was not signiﬁcantly related to homicides

veen 1985 and 1995 (Lee, Martinez, and Rosenﬁeld 2001). What is clear is that

41

“[w]hether these increases [in crime] come from an inﬂux of young, single males — the
group with the highest crime rate — as Broadway suggests for Garden City... awaits
detailed study” (Stull 1994, 115-116). In fact, all of the contributing factors theorized in

the literature awaited empirical investigation.

Crime as the result of population booms and social disorganization

It has also been hypothesized that the sheer increase in population in these communities
:ould foster social disorganization and result in increased crime. This hypothesis, put
brth in studies of boomtowns (communities that experience rapid growth as the result of
:conomic developments, such as the case of large-scale energy resource development in
he western United States),6 has been proposed in studies of slaughterhouse communities
Broadway 2000; Broadway and Stull 2005; Markus 2005; Stull and Broadway 2004).
imply put, this perspective begins by assuming that pre-boom communities are stable
1d characterized by social cohesiveness, where social control is made possible by ‘high
ensity of acquaintanceship’ (F reudenberg 1986). In areas that experience a population
ﬂux, newcomers bring new values that conﬂict with those of current residents, and may
srupt established networks and support systems (Broadway 1990), perhaps resulting in
‘eduction in informal social control, personal disorganization (manifested as mental
eakdowns, suicide, deviance, and disorder) and social isolation (believed to be a key
:tor in explaining child maltreatment in boomtowns (Broadway 2000, 40)).
Broadway and Stull (2005) explain that although Garden City Kansas did not

:et the threshold of 15% annual growth in order to be classiﬁed as a boomtown, they

42

found the boomtown model useful in examining the changes taking place there and in
other slaughterhouse communities. The application of this theorizing to the context of
slaughterhouse communities is exempliﬁed by the following statement: “A rapidly rising
population, coupled with the high population mobility characteristic of packinghouse
towns, contributed to the steady rise in property and violent crimes [in Garden City,
Kansas] during the 19805. Domestic violence was behind most of the increase in violent
:rimes” (Stull and Broadway 2004, 103). The ability of the variables proposed by the
heory to explain the increase in crime rates in slaughterhouse communities has not been
ested empirically. Before moving on to the last theorized cause of increased crime in
laughterhouse communities, however, it is necessary to more fully explicate the theory
nd the variables it implicates, and how well the theory has actually fared when applied
) energy boomtowns.

Stull and Broadway (2004, chapter 7) have pointed out that while ofﬁcial crime
.tes in Garden City, Kansas, where slaughterhouses were sited in the 19805, rose
roughout the 19805 and into the following decade, they actually began to decline during
e second half of the 19905. Stull and Broadway claim that this could conﬁrm that
cial disruption in boomtowns is temporary and declines once community ties are re-
:ablished. Such a decline in social disruption in boomtowns over time is congruent
th what Smith, Krannich, and Hunter (2001) found in their longitudinal examination of
stem boomtowns impacted by energy resource development. However, Stull and
)adway (2004, chapter 7) suggest an alternative explanation for the leveling off of the
ne rates in slaughterhouse communities: the increasing proportion of non-English

aking citizens who move into the area might not be reporting crimes due to both

43

 

cultural and linguistic barriers. They explain that this alternative explanation is
congruent with the crime data which show that the actual number of crimes reported was
nearly the same in 1990 and in 1999, but what does change is a 22.5% increase in the
population, thus reducing the per capita crime rate. The barriers to crime reporting Stull
and Broadway outline would also presumably have mitigated crime reporting from the
beginning of the immigrant inﬂux into the community to some degree; however, these
barriers have become more signiﬁcant as increasingly workers are recruited from
different countries, a greater proportion of them do not speak English, and translation
services for their native language are not yet available in the area.

The social disorganization theory that Broadway and Stull invoke can be traced
Jack to the Chicago School of sociology and the work of Shaw and McKay. The general
typothesis derived from their work is that low economic status, ethnic heterogeneity, and
iigh residential mobility results in social disorganization. One particular manifestation of
ocial disorganization considered important is increases in crime rates (Krannich, Berry,
nd Greider 1989). In comparison to other criminological theories, this theory differs in
rat instead of focusing on types of people to explain crime, social disorganization
muses on types of places and neighborhoods that make crime more likely. According to
ie theory, structural factors lead to the weakening of social institutions, which
ibsequently result in an inability to control social behavior, resulting in increased crime
tes (Kubrin and Weitzer 2003).

This theory dominated criminology in the ﬁrst half of the 20‘h Century, but then

11 out of favor. It was subsequently revitalized in the 19805, and some argue the theory

more pertinent now than ever:

44

Given increasing deindustrialization of central cities, heightened
middle-class mobility, growing segregation and isolation of the poor,
and the growth of immigrant populations in most American cities —
with implications for disrupting or revitalizing social networks,
community cohesion, neighborhood subcultures, and social control —
the theory's relevance is perhaps even stronger today than when it
was ﬁrst proposed many decades ago (Kubrin and Weitzer 2003, 397).
The more current models of the theory are based on a ‘systemic model of control’, which
emphasizes the role of social networks and social capital in facilitating social control
(Kawachi, Kennedy, and Wilkinson 1999; Triplett, Gainey, and Sun 2003). Sampson and
Groves (1989) have usefully delineated the exogenous and endogenous variables
implicated in the theory. The exogenous sources of social disorganization include socio-
economic status, residential mobility, population heterogeneity, family disruption, and
urbanization. These variables are believed to weaken community-based institutions
fTriplett, Gainey, and Sun 2003).7 The endogenous, or mediating variables, include the
ibility of communities to supervise and control teenage groups (this was considered the
iost important intervening variable in Shaw and McKay’s original model); informal
lCEII friendship networks; and the rate of participation in formal and volunteer-based
cal groups. In their own research Sampson and Groves (1989) found that communities
at had sparse friendship networks, unsupervised teenage groups, and low organizational
rticipation had disproportionately higher rates of crime. Additionally, Kawachi,
nnedy, and Wilkinson (1999) found a strong correlation between social capital and
lent and property crimes.
Research testing the social disorganization theory in general has had mixed

lItS, with some of the factors demonstrating more explanatory power than others. For

11106, as mentioned earlier, Lee, Martinez, and Rosenﬁeld (2001) examined the effects

45

of immigration on homicide. They state, “As a major agent of social change...
immigration should inﬂate levels of crime due to its disorganizing inﬂuence on
community institutions” (Lee, Martinez, and Rosenﬁeld 2001, 563); however, they found
in their study of three cities that, controlling for other factors, immigration generally did
not increase the rate of homicide among Latinos and African Americans. Racial
heterogeneity has also failed to signiﬁcantly explain geographic variations in automobile
theft (Rice and Smith 2002). Lee, Martinez, and Rosenﬁeld (2001) also found that
family disruption (measured by the number of female-headed households), another
exogenous variable, did not have a signiﬁcant effect on homicide rates. They did ﬁnd,
however, that economic deprivation (operationalized as the percentage in poverty),
unemployment, and residential instability were consistently related in the expected
direction to victimization in the three cities. Another study, in contrast, found only a
marginally signiﬁcant effect of residential mobility and an insigniﬁcant effect of
socioeconomic status on stranger violence, mugging and robbery, and total victimization
rate (Sampson and Groves 1989). The authors of the study point out, however, that
socioeconomic status does have a signiﬁcant effect on endogenous variables, such as
ﬁiendship networks and unsupervised youths, which in turn may have an impact on crime
rates.

Another study assessing the impact of neighborhood context on rates of intimate
violence and violence between family, friends, and acquaintances found that the presence
of multiple-unit dwellings and transient populations were not signiﬁcantly related to
intimate violence, and that contrary to theoretical assumptions, it was negatively

correlated with violence among family, friends, and acquaintances. The study did ﬁnd

46

that neighborhoods with high levels of poverty, unemployed males, and high numbers of
female headed households had higher rates of intimate violence than other
neighborhoods. Additionally, the neighborhood context explained less of the variance in
the rate of partner abuse than for violence involving others known to the perpetrator
(family, friends, and acquaintances) (Miles-Doan 1998).

Therefore, in examining different types of crime, some of the social
disorganization variables tested have not performed as expected. Nielson, Lee, and
Martinez (2005) contend that although social disorganization theory, as a global theory of
crime, has been helpful in explaining the distribution of violent crime rates nationally,
there are more local concerns that need to be addressed that might explain some of the
variation and deviation from what the theory would lead one to expect. The type of
industries in a community might be one such factor. Others have critiqued the current
social disorganization theorizing for failing to attend to possible subcultural elements that
might facilitate criminality, for assuming a constant level of opportunities to commit
crimes, and for overlooking the motivations to commit crime and how they may vary
over time (Rice and Smith 2002). Additionally, the applicability of the theory in
explaining phenomena in boomtown communities has also been called into question
(Smith, Krannich, and Hunter 2001; Wilkinson, Reynolds, Thompson, and Ostresh 1984).

Many authors, especially within the sociology of community” (particularly the
boomtown) literature, appear to have erroneously equated the entire social
disorganization theory with one of its hypothesized exogenous sources — urbanization.
Several classical theorists argued that changes in community structures lead to changes in

communal relations. For instance, Tonnies (1983) described a shift ﬁom Gemeinschaﬁ, a

47

community based on social bonds (which prevails in rural contexts), to Gesellschaft, a
more differentiated society characterized by exchange and means-end relationships.
Similarly, Durkheim referred to a shift from mechanical to organic solidarity. Wirth

(l 93 8) argued that increases in population result in the dismantling of communalism, and
its replacement with urbanism, whereby social disorganization is increased. The social
disruption hypothesis of the boomtown studies is grounded in Durkheim and Tonnies’s
writings on modernization and Wirth’s work on urban social disorganization (Broadway
and Stull 2005). This perspective has been proposed to understand the effects of the
siting of slaughterhouses in rural communities, which is part of a general trend of
industrial agriculture ‘urbanizing’ the rural (Goldschmidt 1998).

The boomtown studies have been grouped into three categories: the early
research from 1974 through 1980; a period of critique and reevaluation from 1980 to
1984; and after 1984 revised approaches resulting in contrasting results (Smith, Krannich,
and Hunter 2001). The boomtown studies commenced in the early to mid-19705 with the
increase in energy resource development. The early studies found signiﬁcant increases in
divorce, mental disorders, school dropouts, juvenile delinquency, suicides, and crime
rates. Then between 1980 and 1984 the literature fell under critical scrutiny, and the
merits and shortcomings of the research were debated in an issue of Paciﬁc Sociological
Review (Albrecht 1982; Finsterbusch 1982; Freudenberg 1982; Gale 1982; Gold 1982;
Murdock and Leistritz 1982; Wilkinson, Thompson, Reynolds, and Ostresh 1982;
Wilkinson, Thompson, Reynolds, and Ostresh 1982). The central critiques of the early
literature included the contention that many of the studies relied on unproven

assumptions, weak empirical evidence, and potentially unreliable data; they relied on

48

single community case studies instead of comparative studies across communities; and
there was a tendency to rely on cross-sectional rather than longitudinal data (Smith,
Krannich, and Hunter 2001).

Since it has been proposed that the boomtown model and theorizing could be
applied to understand increases in crime rates in slaughterhouse communities, it is ﬁrst
necessary to examine how it has fared in predicting crime increases in energy boomtowns
in general. Studies on the crime rates in boomtowns have been somewhat contradictory
(Smith, Krannich, and Hunter 2001). In a review of the literature, F ruedenberg and Jones
(1991) group the studies into three categories: studies that compare growing counties
with stable counties, using available county-level data; cross-community studies that
employ survey data on criminal victimization; and case studies of crime statistics in
boomtown communities. The results in the ﬁrst category are mixed, and the authors
point to their reliance on ofﬁcial crime statistics as a possible reason why signiﬁcant
relationships between population grth and crime were not found, since official crime
statistics would “be expected to increase measurement error, lowering the observed
correlations between community conditions and recorded crime rates” (F reudenberg and
Jones 1991, 624). The second group of studies, which employ victimization surveys, as
opposed to ofﬁcial report and arrests rates, also come to contradictory conclusions. The
ﬁnal group of studies, using case studies of aggregated data, are the most numerous. In
21 out of the 23 studies in this category, increases in crime outpace the increases in
population. In a more recent assessment of the literature, Hunter, Krannich, and Smith
(2002) state that research has generally demonstrated that rapid population grth is not

unifome associated with extreme disruption and disorganization. Despite little

49

differences in actual victimization, however, like earlier studies (see F reudenberg 1986;
Krannich, Berry, and Greider 1989), the authors ﬁnd that boomtown residents report
signiﬁcantly higher levels of fear of crime.

Approaching the subject from a different angle, F ruedenberg (1986) has found
that the decline in the density of acquaintanceship in boomtowns has signiﬁcant effects
on the control of deviance (long-time residents of boomtowns were more likely than
long-time residents of comparison communities to report locking their doors, fearing for
their safety, and to report having been victimized), socialization of the young (complaints
about deviant youths were more common in the boomtown), and caring for the weaker
members of the community (boomtowns were less likely to exhibit community-wide
caring for those in need to additional support or tolerance). Although these ﬁndings are
consistent with the notion that increased population leads to the weakening of social ties
and increased social problems, they do not support the part of the boomtown disruption
hypothesis that would expect increases in psychopathology. Fruendenberg did not ﬁnd a
signiﬁcant difference between the psychopathology of boomtown residents and the
residents in the control communities. In discussing the increase in psychopathology
hypothesis, Finsterbusch (1982) asserts that there is strong evidence demonstrating that
boomtowns cause social and psychological problems, but concedes that this is not the
case for every town and on every measure of social disruption. Regarding crime rates, he
states that often they do increase, but in a study of McLean County, Wyoming they
actually declined. He concludes, “the crime rates tend to increase. Usually crimes
against persons do not increase faster than the population, but crimes against property and

conduct offenses do. Many observers report the weakening of the mechanism of informal

50

social control in boomtowns and thus a degree of social disorganization” (Finsterbusch
1982, 318).

F ruedenberg also states that the data do not support what he refers to as the
‘compositionalist’ argument, that the social problems are the result of differences in
values and behaviors between the new residents and the long-time residents: there is no
basis for the assumption that the increases in social problems were created by the new
residents. In fact, he found that the individuals who have often been blamed for the
social problems — those who come to these towns looking for work but are unable to ﬁnd
it — accounted for only 0.5% of the mental health cases opened in an 8 year period.

F ruedenberg concedes that these data do not disprove or support the ‘simplistic’ version
of the boomtown disruption hypothesis:
Rather than a shredding or disintegration of the social fabric, the
boomtown appears to have experienced something more like a
popping apart at the seams — with the emphasis here on the fact that
numerous patches of the social fabric have remained very much

intact. What had once resembled a relatively even 'blanket' of
social ties might now be better represented as a patchwork quilt

(Freudenberg 1986, 56).
Some “patches” have remained intact and protected the psychosocial functioning, despite
the fact that disruptions are occurring in other areas of community functioning. He
contends that the boomtown social disruption thesis based on classical sociological
thinking positing an increase in individual-level psychological problems, cannot be
supported by the data. However, a newer hypothesis that psychological stresses can be
mediated by strong ties within a community (examined through density of

acquaintanceship) holds more promise. This perspective would expect that rapid growth

51

would be associated with signiﬁcant increases in crime rates; speciﬁcally crimes
generally not committed by those known by their victims.

Contrary to this hypothesis, however, many have found that rapid growth and
urbanism do not necessarily result in social disorganization (Bender 1978). Change in
community size does not automatically negatively affect communalism or neighboring.
Instead, some have found (see Berry, Krannich, and Greider (1990)) that other
characteristics of the communities, such as the socio-demographic composition, have
more important effects on neighboring. As a result of the ‘urbanism equals
disorganization’ hypothesis, most research on social disorganization has focused on
urban areas to the exclusion on nonurban areas (Kubrin and Weitzer 2003).

Some scholars have examined the possibility that boomtown features may have
varying effects on different members of the community. It has been hypothesized that
women are more negatively affected than men by boomtown development, especially
since they rarely beneﬁt from the incoming jobs as the men do. However, in examining
the personal assessments of residents Freudenberg (1981) found little support for that
hypothesis. Others have noted that children are often negatively affected by social
change due to their unique vulnerabilities (Camasso and Wilkinson 1990). In a pooled
time-series cross-sectional study, Carnasso and Wilkinson (1990) examined the effects of
energy development industries on severe child abuse in counties in the state of Utah.
They found that the energy-development indicators had strong positive effects on child
abuse (with mining employment having the strongest effect), but that the effect of
population change did not have as signiﬁcant an impact on child abuse. The authors also

examined the effect of employment in the manufacturing sector and found that it had a

52

signiﬁcant negative effect on child abuse. They explain, “Any disruptive potential in
manufacturing grth appears to be outweighed by employment that is less prone to the
vagaries of boom-and-bust change” (Camasso and Wilkinson 1990, 13). Thus, the
meatpacking industry (which is categorized as a form of manufacturing) might be unique
in its negative effects.

Wilkinson and his colleagues have been more critical of the boomtown literature
than most scholars. According to Wilkinson, Thompson, Reynolds, and Ostresh (1982)
“Research on western energy development is not based on a cogent theory, but many of
the postulates suggested in discussions of local social effects relate, at least implicitly, to
a theoretical perspective that can be traced to the typological approach used earlier by
Ferdinand Toennines, Emile Durkheim... A conceptual distinction is drawn between
rural and urban" (282). In a subsequent article, Wilkinson, Reynolds, Thompson, and
Ostresh (1984) point out the transition from rural to urban need not lead to a reduction in
the level of integration, unless preexisting structural cleavages (such as social inequality)
limit the emergence of organic solidarity and participation in social interaction. Further,
given how interconnected communities are with the larger society, much of the inﬂuence
on the behavior and mental health of people might be more independent of the changes in
the immediate community than previously thought (Wilkinson, Thompson, Reynolds, and
Ostresh 1982). Controlling for characteristics that could affect crime rates in subsequent
years (measures of urbanization, socio-economic status, and culture of violence) for 197
non-metropolitan counties, the authors ﬁnd that recent growth and energy development
have negligible effects on violent crime when the preexisting characteristics of the county

are controlled. They conclude, “The results of this analysis do not support the social

53

disorganization theory as this theory has been applied in previous discussions of western
energy development. While social integration is not measured directly, the ﬁndings show
clearly that conditions prior to the energy boom give a better explanation of variation in
the violent crime rate than do growth and energy development during the 19705”
(Wilkinson, Reynolds, Thompson, and Ostresh 1984, 252). Clearly there are some
inconsistencies and issues with applying the social disorganization theory to energy
boomtowns. Claims that this model can be applied to explain increased social disruption

in slaughterhouse communities are therefore likely premature.

Increased crime as a result of unemployment

Less frequently, it has been proposed that slaughterhouse communities might experience
increased crime rates because the high turnover rates in the industry (described earlier)
and the recruitment of workers into the community might result in increased
unemployment (Eisnitz 1997; Schlosser 2005[2001]). This proposed explanation of
increased crime rates is fairly straight-forward and does not require as much elaboration
as the demographic and social disorganization theories did. It is exempliﬁed by the
following quote: "Aﬁer quitting, thousands of legal and illegal workers recruited from
pockets of unemployment end up looking to slaughterhouse communities to pick up the
social costs associated with their joblessness and resultant crime" (Eisnitz 1997, 275,
emphasis mine). However, some general studies of the effects of unemployment have
not demonstrated the signiﬁcant positive relationship with crime that Eisnitz and others

hypothesize. For instance, Cohen and Land (1987) found that the unemployment rate had

54

a signiﬁcant negative relationship with motor vehicle thefts. They state that this ﬁnding
is consistent with prior theorizing that an increase in the unemployment rate results in a
reduction in the circulation of people and property, which actually reduces the
opportunities for offending and victimization. Additionally, a state-level of analysis
found that unemployment was not signiﬁcantly positively related to violent or property

crimes (Kawachi, Kennedy, and Wilkinson 1999).

Increased crime as a result of the characteristics of the work?

As explained thus far, the demographic characteristics of the workforce, the negative
effects of the population inﬂux on community social control, and increased levels of
unemployment have been theorized in the literature as causes of the social disruption
observed in communities where slaughterhouses are opened. Broadway has also
mentioned that “work related stress among males would be expected to contribute to
increases in the crime rate and the incidence of depression, divorce, alcoholism and child
abuse” (1990, 328, emphasis mine). The source of this ‘work related stress’, however,
has not been interrogated. In fact, the violent nature of the work inside the
slaughterhouses has not even been mentioned in the literature as a possible factor. Even
Gail Eisnitz, an animal cruelty investigator who details the violence within
Slaughterhouses in her expose', Slaughterhouse: The Shocking Story of Greed, Neglect,
and Inhumane Treatment Inside the US. Meat Industry (1997) — and pointed to
unemployment as causing crime in slaughterhouse communities — fails to draw a link

between the social disruption observed in slaughterhouse communities and the effects of

55

slaughtering and dismembering animals on the workers themselves and their

communities.

That Eisnitz does not at least mention the possibility of a connection between the
increased crime rates in slaughterhouse towns and the repetitive acts of killing and
dismembering animals is particularly interesting given the statements of some of her
informants that appear to corroborate a link between killing and dismembering animals
and emotional and behavioral problems. For instance, she quotes a man who was
employed as a ‘sticker’ (responsible for slitting the animals’ throats) as follows,

after a while you don 't give a shit. You’re just putting in your
time. And then it gets to a point where you're at the daydream
stage. Where you can think about everything else and still do
your job. You become emotionally dead. And you get just as
sadistic as the company itself When I was sticking down

there, I was a sadistic person (75).

Another sticker remarked,

[e]very sticker 1 know carries a gun, and every one of them
would shoot you. Most stickers I know have been arrested for
assault. A lot of them have problems with alcohol. They have
to drink, they have no other way of dealing with killing live,
kicking animals all daylong. If you stop and think about it,
you're killing several thousand beings a day (88).

An additional informant described the connection between working in a slaughterhouse,
substance abuse, and abuse of women:

Most men are not taught to do that [express emotions]. We

don 't share what’s going on inside us. So a lot of guys at

Morrell just drink and drug their problems away. Some of

them end up abusing their spouses because they can 't get rid

of the feelings (88).

And two other informants described the desensitization that occurs as a result of the job.

According to the ﬁrst,

56

 

The worst thing, worse than the physical danger, is the
emotional toll. If you work in that stick pit for any period of
time, you develop an attitude that lets you kill things but
doesn't let you care (87).
The second informant explicitly describes the blurring of the boundaries between killing
animals and killing humans:
My attitude was, it 's only an animal. Kill it. Sometimes I
looked at people that way, too. I've had ideas of hanging my
foreman upside down on the line and sticking him. I
remember going into the oﬂice and telling the personnel man
that I have no problem pulling the trigger on a person — if you
get in my face I'll blow you away (87).

In her book, Eisnitz does detail a problem in meatpacking facilities that has
increased with the line speeds: animals being improperly stunned, regaining
consciousness, and having their throats slit and being dismembered while conscious. The
workers she interviewed reported having to dismember conscious, struggling animals,
and how this deeply troubled them. Perhaps because Eisnitz is focused upon the negative
consequences of the modem slaughterhouse on the animal ‘products’ she overlooks an
important possible byproduct: a link between production in the plants and social
disruption in the communities.

Others have noted the troubling aspects of working in slaughterhouses.
Anthropologist Deborah Fink conducted participant observation for approximately ﬁve
months at an IBP plant in Iowa. In recounting her time there she states that in her life she
had never turned to alcohol for comfort before, but she had a strong desire to do so while

working in the slaughterhouse. She describes the serious emotional toll the work took on

her:

57

Before I left the plant I was carrying heavy depression and thoughts
of suicide that were more real than any I had known before... I
found that my mind would ﬁll with grotesque ﬂashbacks, and I was
unable to process events or emotions as I had before. I dreamed
about looking into a combination of meat [containers she threw the
meat she cut up in] and seeing detached arms and tormented faces
reaching up to me to be saved - or to pull me in (Fink 1998, 37).

Fink also reports that one woman worker committed suicide, another woman stabbed her

boyfriend in the heart, and a male co-worker ran his girlfriend over and almost killed

her.9 And in the epilogue to her book, Fink states,
The distress of the majority of packing workers emerges repeatedly
in newspaper items on crime or personal tragedies. Often, as when
Pete 's attack on his girlfriend got highlighted because of his prior
conviction of murdering his wife, the packinghouse connection was
not reported. So familiar was the state penitentiary at Anamosa that
plant workers spoke of their prison term as 'visiting Annie’. Those
packinghouse workers who have not experienced major problems
related to the stress of their jobs in the late 19905 are the exceptions
rather than the rule (1998, 192).

The possibility that the killing and dismembering of thousands of animals a day might

contribute to this stress and the larger social problems in these communities has

surprisingly not been mentioned in the literature.

The negative community effects of other forms of violent employment or service
have been documented. For instance, a relationship between military service (where
individuals are trained in the use of violence against others) and illegal forms of violence
has been documented. Military service has been found to be particularly strongly related
to forms of violence within the family. In a review of the literature investigating the link
between intimate partner violence and military service, Marshall, Panuzio, and Taft

(2005) state that reported rates of intimate partner violence are up to three times higher

than the rate found in samples of civilians; and a review of records from 1992 to 1996

58

found that the rate was ﬁve times that of the civilian population (Mercier 2000).
Evidence of increased general violence, as opposed to speciﬁc violence against intimates,
has also been found. For instance, a study of incarcerated men found that there was a
signiﬁcant relationship between imprisonment for violent crime and previous military
experience; this was the case even when age, race, and education were controlled for.
Military experience had the second strongest effect of the variables (age had the
strongest). Controlling for these variables, the length of time spent in the military was
positively and signiﬁcantly related to incarceration, but there was not a signiﬁcant
difference between those who had witnessed combat during their military careers and
those who had not (Allen 2000). However, combat experience has been found to have a
signiﬁcant impact on marital adversity, and the effect has been found to be mediated by
post-traumatic stress symptoms and antisocial behavior. The authors conclude that
combat directly increases violent and unlawful or antisocial behavior and stress, which
subsequently negatively affects marital quality and stability (Gimbel and Booth 1994).
Theorized reasons for this link include the following: congruent with social
learning theories the violence is learned in the service; the armed services recruit
individuals who are predisposed to violence (Marshall and McShane 2000); like police
and corrections ofﬁcers, a subculture of violence is created in the peer group whereby
violence is expected and tolerated and hypermasculinity often prevails (Marshall and
McShane 2000; Rosen, Kaminski, Parmley, Knudson, and F ancher 2003); for some
military service results in Post-Traumatic Stress Disorder, which makes them more prone
to act violently (Marshall, Panuzio, and Taft 2005); and stress is created by the low pay

and being isolated from friends and family for extended lengths of time (Mercier 2000).

59

It is not unreasonable to posit that there could be a connection between other jobs that
entail violence and violence outside of the workplace. In fact, "Steinmetz and Straus
(1974) argue that the more normalcy placed on aggressive behavior in the occupational
role, the greater the amount of violence” (Marshall and McShane 2000, 18).

In general, aggregate crime research has been characterized by a rudimentary
appreciation of the links between labor processes and crime. Most have only examined
the posited link between unemployment and crime. Few have examined how the
industrial structure of an area might affect crime rates, and those few have focused on
urban communities (Lee and Ousey 2001).

The social consequences of manufacturing have been discussed to a limited
degree in the literature. Attention has been paid to the effects of Taylorism, or scientiﬁc
management, which include an extreme division of labor and gross deskilling that
characterize manufacturing work today. The result is that the workers are separated from
the means of production and from the products of their labor, and they surrender their
interest in the labor process (Braverrnan 1974/1998). However, employment in
manufacturing does not appear to contribute to social problems as much as employment
in other industries. Recall that in examining the effects of employment in mining,
construction, and transportation on cases of child maltreatment, Carnasso and Wilkinson
(1990) found that manufacturing had a signiﬁcant negative effect on child maltreatment,
and that any disruptive potential of this type of work appears to be muted by the steady
and reliable nature of the work. In an examination of nonmetropolitan counties, another
study found that the presence of small manufacturing ﬁrms is associated with lower rates

of homicide, robbery, burglary, and motor vehicle theft. It is theorized that small

60

manufacturing jobs are particularly helpful because smaller companies often invest in the
communities (unlike multinational corporations), they may provide an entree for low and
semi-skilled workers into the labor force, and they may recruit high school children (Lee

and Ousey 2001).

Summary

Thus, the apparent negative consequences of slaughterhouses need to be explained. As
detailed herein, reasons for the negative social consequences of slaughterhouses have
been hypothesized, but not yet addressed empirically, which has lead Broadway to
conclude, “Clearly, further research is needed to determine the reasons behind the
increases in social disorder and whether Garden City’s experience is representative of
small towns which undergo rapid growth” (1990, 342). Further, the possible
explanations that have been put forth — that increased crime rates in slaughterhouse
communities are the result of the inﬂux of immigrants and young, adult males, assumed
to pose an increased criminogenic risk; that the inﬂux of people results in social
disorganization and increased crime; and that the increase in crime may be the result of
increased unemployment caused by the high turnover rate in the meatpacking industry -
have not included another possibility. The possibility that at least some of the social
disruption that has been observed in communities is due to the nature of the work that is
conducted in slaughterhouses has not been mentioned, much less examined (this has been

substantiated by Broadway (2005, personal communication)). The purpose of this study

61

is to begin to ﬁll this theoretical and empirical gap in the literature. The next chapter

explicates how this objective was actualized.

62

Chapter 2

RESEARCH METHODOLOGY

1 fully agree with you about the signiﬁcance and educational value of methodology as
well as history and philosophy of science. So many people today — and even professional
scientists — seem to me like someone who has seen thousands of trees but has never seen
a forest.

- Albert Einstein in a letter to Physics Professor Robert A. Thorton, 1944

This chapter provides a detailed examination of the design of the study, including the
hypotheses tested, the data used, and the operationalization of the variables included in
the study. Subsequently, the analysis techniques employed are discussed, with particular
attention paid to explaining why these techniques were selected and describing what they
entail. The chapter closes with a discussion of the strengths and limitations of the design

of this study and the analysis techniques used therein.

Study Design

As stated in the introductory chapter, three hypotheses are tested in this study. These

hypotheses are restated below.

Hypothesis 1: General and speciﬁc crime rates will increase as the number of

slaughterhouses and slaughterhouse employees increases.

63

Hypothesis 2: Controlling for the variables proposed in the literature (the unemployment
rate, the number of people in poverty, net immigration, net migration, the
number of non-white or Hispanic residents, the number of young males,
the total number of males, and the population density of the county),
slaughterhouse presence and employment will be associated with
increased crime rates in counties, more so than industries that utilize the
same type of labor force, have high injury and illness rates, and entail

routinized labor, but do not involve killing and dismembering animals.

Hypothesis 3: Controlling for the variables proposed in the literature (the unemployment
rate, the number of people in poverty, net immigration, net migration, the
number of non-white or Hispanic residents, the number of young males,
the total number of males, and the population density of the county), rape
and family violence rates in particular will increase in counties where

there is an increase in slaughterhouse employees.

A test of the ﬁrst hypothesis is needed because the studies conducted thus far have
only been able to examine a few communities where extremely large slaughterhouses
have opened. Therefore, it is necessary to test whether or not results from those studies
are generalizable. Testing the second hypothesis is necessary to ascertain whether the
increase in crime in slaughterhouse communities can be explained by the variables

proposed in the literature, and if the effects are unique to slaughterhouses, or if

employment rates in similar industries characterized by high levels of immigrant

64

employment, high injury and illness rates, and routinized work would result in similar
increases in crime. The ﬁnal hypothesis is grounded in the theorizing of scholars such as
Adams (1991; 1995), Patterson (2002), and Spiegel (1996) positing a link between the
victimization of animals and the victimization of less powerful human groups, such as
children and women. Testing this hypothesis will demonstrate whether or not there is a
speciﬁc link between the amount of slaughterhouse employment in counties and the
victimization of less powerful groups of people, such as women and children, in the form
of rape and family violence.

This study analyses quantitative, aggregate-level data for three central reasons.
First, the studies'of slaughterhouse communities conducted thus far (as described in the
previous chapter) have been limited to a few communities where especially large
slaughterhouse facilities have opened. A study was needed that included all communities
with slaughterhouses (not just those with newly opened ones) and thus did not restrict the
variance in the independent variable. In this study, counties with large slaughterhouses,
small slaughterhouses, and those without slaughterhouses are all are included in the
analysis, thus making the results more generalizable. Additionally, this type of analysis
is best suited for testing the theories of increased crime in slaughterhouse communities
proposed in the literature because it permits the statistical control of other variables
hypothesized to account for the increases in crime (e.g., the proportion of young males)
in assessing the impact of the level of slaughterhouse employment on crime. Third,
quantitative, aggregate-level analysis also enables a comparison of the effects of
slaughterhouses and slaughterhouse employment to the presence and levels of

employment in other types of industries that utilize the same type of workforce and entail

65

the same type of work, minus the killing and dismembering of animals. Such a
comparison was not feasible in the ethnographic community studies conducted thus far.

Secondary data are utilized in this study since systematic, centrally-collected data
are available for the necessary variables for multiple counties over a long time period,
making panel data analysis possible. The use of this secondary data has many .
advantages, but it also comes with limitations. For instance, there were limits on the time
period that could be included in the study. This study covers the time period from 1994
to 2002. The analysis could not extend back before the year 1994 because the Uniform
Crime Report data collected prior to 1994 should not be compared with data collected
later due to the implementation in that year of a new imputation algorithm to adjust for
incomplete reporting by law enforcement jurisdictions. This change is believed to make
estimates for longitudinal analysis more accurate because the data are less sensitive to
changes in the extent of reporting between years (ICPSR 2004). The year 2002 is the
most recent year included in this study because some of the demographic and County
Business Patterns data on the industries of interest for more recent years were not yet
available.

The unit of analysis for this study, the county, is large enough to permit access to
uniform data across cases and time, but small enough so that each case should be
sensitive to the effects of the slaughterhouses and presumably be less affected by
confounding variables than larger units, such as the state. Disadvantages to using the
county as the unit of analysis also exist and the following two are potentially the most
consequential: (1) social phenomena, of course, do not always follow geopolitical

divisions (Wilkinson, Reynolds, Thompson, and Ostresh 1984) and, (2) subcounty

66

differences cannot be measured (Camasso and Wilkinson 1990; Rice and Smith 2002;
Wilkinson, Reynolds, Thompson, and Ostresh 1984). (According to Camasso and
Wilkinson (1990), however, this latter disadvantage is ameliorated somewhat by the fact
that county-level data are able to capture phenomena at both the center and peripheral
areas of the community.) Overall, it was felt that the advantages of using the county as
the unit of analysis (especially having standard longitudinal data for all variables of
interest) vastly outweighed the disadvantages.

The sample in this study is delimited in two ways in order to mitigate some
potential confounding effects. First, only counties in right-to-work states are included in
this study. In these states employees cannot be required to join or pay dues to a union
and may resign from the union at any time, while still enjoying the beneﬁts of the
collective agreement. The following states currently have right-to-work laws (and did
during the entire study time period, with the exception of Oklahoma): Alabama, Arizona,
Arkansas, Florida, Georgia, Idaho, Iowa, Kansas, Louisiana, Mississippi, Nebraska,
Nevada, North Carolina, North Dakota, Oklahoma, South Carolina, South Dakota,
Tennessee, Texas, Utah, Virginia and Wyoming (National Right to Work Legal Defence
Foundation 2005). Counties in Oklahoma are included in the sample because right-to-
work laws were enacted in the state during the time period examined in this study.
Restricting the sample to counties in states with right-to-work laws aids in minimizing
the potential suppressant effects strong unions might have on the community impacts of
the slaughterhouses, as it is unlikely that slaughterhouses have strong unions in right to
work states. However, it also means that the results cannot be generalized to counties in

states without right-to-work laws.

67

Table 1. Rural-Urban Continuum Codes, 2003

 

[Code [Description

1 [Counties in metro areas of 1 million population or more

 

 

[Counties in metro areas of 250,000 to 1 million population

 

[Counties in metro areas of fewer than 250,000 population

 

 

2
3
4 [Urban population of 20,000 or more, adjacent to a metro area
'5

rban population of 20,000 or more, not adjacent to a
etro area

 

[Urban population of 2,500 to 19,999, adjacent to a metro area
[:eran population of 2,500 to 19,999, not adjacent to a metro

 

 

 

 

 

rea

8 ompletely rural or less than 2,500 urban population, adjacent
o a metro area

'9 ompletely rural or less than 2,500 urban population, not
djacent to a metro area

 

Source: United States Department of Agriculture, Economic Research Service (2005)

Additionally, the study sample excludes metropolitan counties and those adjacent
to metropolitan areas in order to remove the potentially confounding effects of
urbanization and spillover from metropolitan areas to rural counties, which is oﬁen found
in criminological research (see, for example, Lee and Ousey 2001). Non-metropolitan
counties and those not adjacent to metropolitan areas were operationalized as those
counties categorized by the Ofﬁce of Management and Budget in 2003 as
nonmetropolitan and not adjacent to metropolitan areas as reported by the United States
Department of Agriculture, Economic Research Service (the counties coded as 5, 7, and 9
in the continuum in Table 1). The result of these criteria is that 581 non-metropolitan
counties not adjacent to metropolitan areas in right-to-work states are included in this

study (see Appendix A for a complete list of the counties included in the study). Thus, a

68

sample of a larger population is not analyzed per se; rather, the entire population of non-
metropolitan counties not adjacent to metropolitan areas in right-to-work states are
analyzed. Again, it is worth noting that there are limits to the generalizability of this
study: the results cannot be generalized to counties in or adjacent to metropolitan areas.
Subsequent studies might extend the analyses to include metropolitan counties and those

adjacent to metropolitan areas in states without right to work laws.

Data Collection

Because secondary data analysis was used in this study the data did not have to be
collected for each case separately; instead, the data had to be compiled from several
sources. The steps for doing so were rather straight forward, although time consuming
(spanning over a period of approximately four months). Most of the data were
downloaded directly from governmental websites into Microsoft Excel and later imported
into SPSS 14.0. The Uniform Crime Report data, however, were able to be directly
exported from the Inter-University Consortium for Political and Social Research website
into SPSS. Some of the data were not in the necessary format or proper level of
aggregation (for instance, the demographic data were grouped by age, race/ethnicity, and
sex) and had to be manipulated in order to get it into the proper format. Additionally, the
County Business Patterns data for the years 1994 to 2002 were acquired on CD5 from the
State Library and while the data for the later years could be imported directly into Excel,

the data from earlier years had to be manually entered into the data set.

69

All of the data were compiled for each individual year into separate ﬁles in SPSS.
New variable names were given, the data were checked and cleaned, and new variables
were created (described later) in SPSS. Then all of the data were merged together into

one ﬁle in STATA 8. The analysis techniques used in STATA are described shortly.

Instrumentation and Measurement

As mentioned earlier, the data for this study were collected from six different sources. It
was necessary to access data from so many different sources because the independent
variables, dependent variables, and the variables that needed to be included in the model
as control variables to assess the applicability of the theorized causes of increased crime
in slaughterhouse communities were not available from a single source. Below each of
the variables, the source from which they were obtained, and any special measurement
issues related to them are described. A summary table describing the variables, including

descriptive statistics, can be found in Appendix B.

Independent Variables:

All of the independent variables (the number of slaughterhouse employees and

establishments, as well as the number of employees and establishments for the

comparison industries) were drawn from the County Business Patterns (US. Department

of Commerce and US. Census Bureau). This is the most comprehensive source of

70

county-level annual employment by detailed industry classiﬁcations and is used in
studies where it is necessary to delineate industries and employment at the county-level
(for instance, see Camasso and Wilkinson 1990; Lee and Ousey 2001); however, it only
includes employees covered by the federal social security program (Camasso and
Wilkinson 1990). Therefore, it excludes self-employed individuals, employees of private
households, railroad employees, agricultural production employees, and most
government employees. The self-employment and agricultural exclusions, however, may
be useful in excluding small ‘custom’ slaughter facilities (which are discussed in detail in
Chapter 4) to some extent. The program has used the North American Industry
Classiﬁcation System (N AICS) since 1998. Prior to 1998, the data are based on the
Standard Industrial Classiﬁcation (SIC) System (US. Census Bureau 2006).

Due to the time period under study, having data that were bridgeable between the
NAICS and SIC classiﬁcation systems was a concern. Fortunately, the slaughterhouse
variable (N AICS category 311611) — “animal (except poultry) slaughtering” - is
bridgeable and is comprised of SIC category 2011 (“Meat packing plants”) and 20% of
SIC category 2048 (“Prepared feeds and feed ingredients for animals and fowl, except
dogs and cats”).

It should be noted, however, that in the year 1998 custom slaughtering facilities
were added to the animal (except poultry) slaughtering category. Custom slaughter
includes (a) slaughter or processing of uninspected meat food animals for the sole
consumption of the owner; (b) slaughtering/processing animals as a custom service for an
individual who owns the animal, and uses the meat for his/her own consumption. These

tend to be very small establishments. As demonstrated in Table 2 this resulted in an

71

increase from 1997 to 1998 in the smaller slaughterhouse facilities. A potential
consequence of this classiﬁcation change is that the effects of slaughterhouses on crime
in these years could be diluted in the aggregate data by the increase in these small

facilities.

Table 2. Changes in the Number of Slaughterhouse Facilities from 1997 to 1998

 

 

 

 

 

 

 

 

 

 

 

 

 

1997 1998 Chgge
Total number of employees 140,883 146,305 5,422
Total number of 1,406 2,151 745
establishments
Establishments with 1-4 532 1,046 514
workers
Establishments with 5-9 273 421 148
workers
Establishments with 10-19 218 286 68
workers
Establishments with 20-49 142 155 13
workers
Establishments with 50-99 77 78 1
workers
Establishments with 100-249 62 64 2
workers
Establishments with 250-499 36 35 -1
workers
Establishments with 500-999 18 16 -2
workers
Establishments with 1000+ 48 50 2
workers

 

 

 

 

 

Source of data: County Business Patterns, 1994-2002

Having bridgeable data was also a signiﬁcant factor in selecting the comparison
industries to be used in this study: industries that are similar to the meat packing industry
in that they use a similar labor force, have high injury and illness rates, and entail

routinized work, but not the killing and dismembering of animals. The selected

72

industries include Iron and Steel Forging, Truck Trailer Manufacturing, Motor Vehicle
Stamping, Sign Manufacturing, and Industrial Launderers. Table 3 provides information
for each of these industries regarding the number of workers, immigrant concentration,

and illness and injury rates.

Table 3. Slaughterhouse and Comparison Industries

 

 

 

 

 

 

 

 

 

INAICS [Name 0. of [Immigrant Concentration [Injury/Illness
mployeesll
311611 'mal 142,374 art of Food Manufacturing, which [#15 for injury
except s#7 1n immigrant concentration and illness (as
oultry) of 2004)
Slaughtering
332111 ron and Steel 26,432 artof Fabricated Metal Products, #8 for injury /
orging Ehich 15 #18 1n immigrant 7 for injury
oncentration d illness
336212 ruck Trailer 30,678 artof Motor Vehicles and 12 in injury
anufacturing quipment manufacturing, which 15 d #12 in
35 in immigrant concentration njury and
'llness
336370 otor Vehicle 126,905 artof Motor Vehicles and 19 in injury
etal quipment manufacturing, which 15 d illness
Stamping 35 1n immigrant concentration
339950 Sign 82,956 artof Miscellaneous ot among the
anufacturing anufacturing, which 15 #4 1n 'ghest rates
mmigrant concentration
I812332 ndustrial 1,908 art of Personal and Laundry ot among the
aunderers [‘8 ervices, which is #5 in immigrant 'ghest rates
oncentration

 

 

 

Sources: Information on the industry classiﬁcation and number of employees
obtained from County Business Patterns website (US. Census Bureau
2006); Information on immigrant concentration obtained from Cortes
(2005); Information on illness and injury rates obtained from Bureau of
Labor Statistics, US. Department of Labor (2004a; 2004b).

73

Control Variables:

The control variables for this study can be grouped into three categories in line with the
causes theorized in the literature: demographic variables, social disorganization
variables, and unemployment. These categories are not intended to be mutually
exclusive: in fact, many studies of crime incorporate all three. Rather, these categories

reﬂect the theorizing in the slaughterhouse literature. ’0

Demographic Variables:

All three of the demographic variables used in this study — the number of young men, the
total number of men, and the total number of nonwhites or Hispanics (by deﬁnition this
category includes people who identify their race as white but their ethnicity as Hispanic)
— were obtained from the Population Estimates (US Census Bureau). These data are
estimates for July 1 of each year. The Census Bureau used what they term a distributive
cohort method in creating the estimates. They reportedly applied this method in several
steps: (1) they began with previously developed county population estimates by age, and
state population estimates by age, sex, race, and Hispanic origin; (2) they estimated the
variables for each county by estimating post-censal change in the populations with a
cohort component model; and, (3) they applied these distributions to the original county
estimates by age and state characteristics estimates (US Census Bureau).

A variable measuring the number of males in the counties was included because it

has been asserted, especially in the boomtown literature, that there is a large inﬂux of

74

men into these areas and that men are more likely to be offenders; therefore, the number
of men in an area is believed to be positively related to the crime rates. Therefore, it is
common in aggregate crime research for the sex ratio to be used as a control variable (for
example, see Lee and Ousey 2001).

A variable was also created to measure the number of young men. This variable
was created by summing the number of males in the following categories: males age 15-
19, males age 20-24, males age 25-29, and males age 30-34. Thus, the variable used in
this study represents the number of males age 15 to 34. This is a rather large age span;
however, it was necessary to capture all of years of increased crime risk. Studies have
found that there is a peak in criminality around age 30 and that arrests for violent offenses
peak a few years after the peak for property offenses (Cohen and Land 1987). There is,
however, no standard grouping of age categories when studying the relationship between
age and crime (Krivo and Peterson 2004), so this aggregation reasonably captures the
group usually expected to have higher criminality.

The number of nonwhites and/or Hispanics variable also requires brief
elaboration. The collection of the race/ethnicity data changed in the year 2000 with the
introduction of multiracial categories in the census, thus making comparisons for speciﬁc
categories across the time period under study (1994-2002) unreasonable. Consequently,
controlling for the number of individuals in a speciﬁc group, such as Hispanics, was not
feasible. Instead, a non-white and/or Hispanic variable was created by subtracting the
number of white non-Hispanic males and the number of white non-Hispanic females

from the total county population. Thus, while the effects of speciﬁc racial/ethnic groups

75

cannot be ascertained across the study time period, the effects of the total non-white

and/or Hispanic population can be.
Social Disorganization Variables:

Several variables believed to contribute to social disorganization (in addition to the
demographic variables listed above that are often implicated in social disorganization
theory) were included in the model. These variables include a measure of poverty,
population density, international immigration and domestic migration. The poverty in an
area is a critical social disorganization variable. Three variables measuring
poverty/income were available from the Small Area Income and Poverty Estimates (US.
Census Bureau): the number of people in the county in poverty, the percent of the
county’s population in poverty, and the median household income. It was decided that
the median household income variable would be inadequate because it would not give a
reasonable estimation of poverty across counties (especially given variations in the
standard of living across the country). Therefore, the number or percent of people in
poverty would more accurately reﬂect the theorized cause of social disorganization that
this variable was intended to tap. These variables have been used in studies as measures
of economic deprivation (see Lee, Martinez, and Rosenﬁeld 2001).

The income and poverty estimates are obtained by modeling the relationship
between income or poverty and tax and program data (such as the Food Stamp program)
using estimates of income/poverty from the Current Population Survey Annual Social

and Economic Supplement. The deﬁnition of poverty is based on the characteristics of

76

the family (number of people, ages, etc.) and the household income. Unfortunately, the
data were not available for the years 1994 and 1996 because prior to 1998, county-level
data were only produced every second year, for odd-numbered years. It was possible to
impute the 1996 data from the 1995 and 1997 data; however, it was not possible to do so
for the 1994 data because the 1993 data were not available.

Another variable of import to studies of social disorganization is population
density. As discussed earlier, studies of social disorganization (notably those of energy
boomtowns) hypothesize that high population density results in the loosening of social
bonds because people presumably do not know their neighbors and are unable to provide
surveillance (Lee and Ousey 2001), as well as with increased population density there is
an increase in potential targets for offending (Cohen and Land 1987). The population
density variable was created by dividing the total county population by the square mile
area of the county.

Social disorganization theorists have also proposed that social disorganization
occurs in areas with high levels of population turnover and consequently measures of
turnover are frequently included in analyses of crime (see, for example, Lee and Ousey
2001). In addition to absolute turnover, it is argued that international immigrants in an
area pose particular risks for social disorganization, at least partially because they may
hold values at odds with those of the host population. Therefore, it was necessary in this
study to include a measure of domestic migration and international immigration into the
counties. These variables were obtained from the Population Estimates (US Census
Bureau), which provides estimates for July 1 of each year. As with the demographic

variables obtained from the Population Estimates, the migration and immigration

77

variables were created using a distributive cohort method (described above). The
immigration and migration variables are net change variables, meaning that the variable
represents the difference between in—migration and out-migration. Internal in- and out-
migration includes moves within the United States (excluding Puerto Rico). International
in- and out-migration consists of movement across US. borders, which includes lawful
permanent residents, temporary migrants, humanitarian migrants (i.e. refugees), and
“people illegally present in the United States” (US. Census Bureau, State and County
Terms and Deﬁnitions). The data for the year 2000 only contained measures of
immigration and migration for a four month period. Therefore, it was necessary to

estimate the annual value by multiplying that ﬁgure by three.

Unemployment variable:

As mentioned earlier, unemployment has been theorized as a cause of increases in crime
in slaughterhouse communities. Therefore, it was necessary to include it in the model.
The unemployment rate for each county was accessed through the Local Area
Unemployment Statistics (Bureau of Labor Statistics). Unemployed persons are those
who were not employed during the reference week, but were available for work, and had
made efforts to ﬁnd employment during the previous four weeks. The data are collected
from a variety of sources, including the Current Population Survey, the Current

Employment Statistics program, and unemployment insurance program.

78

Exposure Variable:

One of the analysis techniques used in this study (negative binomial regression) requires
that an exposure variable be identiﬁed to differentiate across cases differences in the
possibility of being ‘exposed’ to the effect. Long and Freese (2006) use the example of
time as an exposure variable. In this study, however, it is not time that differentiates the
likelihood of crime in the counties, it is differences across counties in population (a larger
number of people makes the possibility of offending or being victimized greater).
Therefore, in this study the exposure variable is county population. Including the
exposure variable adds the natural log of the size of the population at risk to the model.
Thus, in essence, the model analyzes per capita rates of crime instead of merely counts of
crime even though the dependent variable is a count, not a rate. This is standard practice
in the quantitative criminology literature (Osgood 2000). Using the population as the
exposure variable also permits an acknowledgement in the model that rates based on
larger populations have greater precision, which addresses the issue of heterogeneity of
variance, which is problematic in the use of OLS regression on count variables (Krivo
and Peterson 2004; Osgood 2000). The population data for all of the counties were

obtained from the Population Estimates (US Census Bureau).

79

Dependent Variables:

All of the dependent variables used in this study (there are 22 in total) come from the
Uniform Crime Reports (UCR), compiled by the Federal Bureau of Investigation. Both
arrest and report variables (14 and 8 variables respectively) are included in the analysis.
It has been argued that report rates are more valid than arrest rates because they are less
susceptible to bias (Kawachi, Kennedy, and Wilkinson 1999; Krivo and Peterson 2004);
however, report rates are collected for signiﬁcantly fewer offenses.

Although many studies of crime rely on the UCR for their data (such as Kawachi,
Kennedy, and Wilkinson 1999; Krivo and Peterson 2004; Lee and Ousey 2001;
Wilkinson, Reynolds, Thompson, and Ostresh 1984), shortcomings of the data have been
identiﬁed. For instance, official statistics obviously exclude those crimes that law
enforcement ofﬁcials are not aware of. However, for some offenses, such as motor
vehicle theft and homicide (Kawachi, Kennedy, and Wilkinson 1999), and serious crimes
more generally (Sampson 1987), the undercount is trivial. There are also problems
related to the ability of victims and witnesses to recall and report accurate information,
limitations of police resources for making arrests, and inconsistencies in the deployment
of resources and enforcement of laws across geographic areas (Krivo and Peterson 2004;
Sampson and Groves 1989). The validity of the data has been questioned particularly in
areas undergoing rapid growth. It is possible that increases in ofﬁcial crime rates in
growing areas are the result of increases in police staff, additions which are common in
boomtowns. Relatedly, it is also possible that increases in crime rates in boomtowns

might be partly due to increased reports by law enforcement ofﬁcials in an attempt to

80

justify increasing their resources (Gold 1982). However, residents in stable areas have
been known to assert that the police record even minor incidents because their time is not
occupied with serious offenses (F reudenberg and Jones 1991), thus potentially increasing
crime rates at the less severe end of the spectrum.

Despite the critiques of ofﬁcial arrest and report data, these data are the best
sources of systematic offense information at the county-level: “Although they have their
limitations, police data represent the only regular source of information on incidents of
violence that is readily available and timely” (Miles-Doan 1998, 629). Some have
suggested that victimization data be used instead of arrest and report data; however,
victimization data are more limited and few differences have been found between the
arrest rates of the UCR and offending rates estimated from the national victimization

survey (Sampson 1987).

Data Analysis

In this section the type of empirical speciﬁcations used in this study are described. In
describing the speciﬁcations they used in their longitudinal study of the effects of
graduated driver licensing on teen trafﬁc fatalities, Dee, Grabowkski, and Morrisey
(2005) state that their choice was motivated by two factors: ﬁrst, they wanted to take
advantage of the longitudinal data that were available; second, they needed a form of
speciﬁcation that would account for the count nature of their dependent variable. These

same two concerns motivated the selection of the empirical speciﬁcations used in this

81

study. Before delving into the logistics of the speciﬁcations it will be helpful to brieﬂy

discuss these two factors in detail.

The Longitudinal Dimension

Although data collected over several time periods is often referred to as panel data

(N erlove 2002), there are some subtleties that must be addressed. Panel data consist of
repeated cross-section data on a large number of cases which are sampled and are
generally only observed a few times. For instance, conducting a survey of a sample of
teenagers every couple of years would create panel data. In such data the cases
(individual teenagers in this example) are not of interest; rather, the underlying
population (all teenagers) that was sampled is of interest. In contrast, pure time series
design entails several observations on only a single case.

With pooled time series cross section (TSCS) analyses there are generally a large
number of cases and a relatively large number of observations of these cases (Beck 2001;
Nielsen 1999). With TSCS data the units themselves are of interest: they are ﬁxed and
not sampled (Beck 2001). TSCS data are generally annual observations of ﬁxed political
units, such as countries, states, and counties. The combination of the time-series and
cross-section dimensions can enhance the quantity and quality of the data above and
beyond what is possible using only one of these dimensions (Yaffee 2003). This study
falls into the TSCS category because the units themselves (non-metropolitan counties in
right-to-work states) are of interest and are not being used to draw inferences to a larger

population. As well, there are a relatively large number of cases and observations.

82

Although some might think that the inclusion of a longitudinal dimension
introduces new problems into an analysis, others (such as Nielsen 1999) argue instead
that longitudinal data provide the leverage necessary to address some of the problems
confronted by cross-sectional research. These types of analyses have several advantages.
First, unlike static cross-sectional analyses, panel or TSCS designs incorporate change
(Finkel 1995; Kalton and Citro 2000; Yaffee 2003). In fact, “To study change... it is
necessary to use a longitudinal design” (Rose 2000, 8). Second, panel data techniques
make it possible to control for all time-invariant unit-speciﬁc variables that are not
included in the models (also referred to as heterogeneity bias), such as culture, history,
climate, etc., that could potentially render the relationship between the dependent variable
and the independent variable of interest partially or fully spurious (Allison and Waterman
2002; Finkel 1995; Halaby 2004; Nerlove 2002; Nielsen 1999; Rose 2000). More
speciﬁcally, in cross-sectional studies, the estimates of the regression coefﬁcients can be
biased if the independent variables are correlated with time-invariant variables that are
affecting the dependent variable, because these time-invariant variables are thrown into
the error term. With TSCS analyses, however, the time-invariant variables fall out of the
equation due to the differencing process between years used in estimations, resulting in
the independent variable no longer being correlated with the error term (F inkel 1995).

Many scholars argue that attending to these time-invariant variables is often
critical. For instance, in examining the consequences of economic growth in developing
countries, Firebaugh and Beck contend that because nations are unique, it is highly likely
that unmeasured characteristics unique to them have important effects on the variables

that many sociologists study. Such unique characteristics could include geographic

83

location, topography, climate, rainfall, resources, history, culture, economic system,
political system, legal system, religious composition, and relationships with neighboring
countries. They accordingly assert, “when enduring attributes are important — as we
suggest they often are — the results of quantitative cross-national research can be
seriously biased” (Firebaugh and Beck 1994, 637). The differences between counties on
these types of unique characteristics might not be as large as for countries, but they
nonetheless could be related to the dependent variables (crime in this case) in important
ways.

Another advantage of TSCS data is that more observations are usually available
(Nerlove 2002), since the data for each year and case can be combined, which is referred
to as ‘pooled’ data. Pooled design is beneﬁcial because it combines the inferential
strengths of time-series and cross-sectional methods. However, in addition to enjoying
the beneﬁts of both designs, pooled methods also inherit the estimation problems of both
(Fording 1997).

Because pooled TSCS data have two dimensions (the cross-sectional unit of
observation and the temporal dimension), the error term has two dimensions (one for the
cross-sectional unit and one for time) (Gordon, Lahey, Kawai, Loeber, Stouthhamer-
Loeber, and Farrington 2004; Yaffee 2003). There are therefore two ways of estimating
the models: ﬁxed effects and random effects. “Both separate the error term from a
standard regression model into person-speciﬁc [county-speciﬁc in this case] and time-
speciﬁc components. The ﬁrst differs across individuals [counties in this case] but
remains the same within each individual across time. The second differs across both

individuals and time” (Gordon et a1. 2004, 65). In the random effect model the error is

84

dealt with by assuming that the intercept for each unit is a random outcome variable.
With the ﬁxed effects model the intercept is case-speciﬁc, differing from county to
county. In the simplest version of this model there are no estimated temporal effects, but
there are estimated differences among the units. This model is sometimes referred to as
the Least Squares Dummy Variable Model because dummy variables are used to capture
these effects. A potential drawback of this model is that numerous cases will require
numerous dummy variables and could remove degrees of freedom from the model, thus
compromising statistical tests. Additionally, many dummy variables could result in
multicollinearity (Yaffee 2003). As I will discuss below, there is an alternative way to
estimate this model that avoids the problems involved in estimating a model with
hundreds of dummy variables, one for each cross-sectional unit.

The choice of which type of effects to use depends upon whether or not the
unobserved effects are correlated with the independent variables. If the correlation
between the time varying independent variables and the time-invariant unobserved
variables is 0, then the assumptions of the random effects model are satisﬁed and the
random-effects model is apparently better suited for the analysis. On the other hand,
when the correlation is not equal to 0, the random-effects estimators are likely to be
biased and ﬁxed-effects should be used as they will remove that bias (Allison and
Waterman 2002). Halaby (2004) argues that researchers too quickly reject ﬁxed effects
in favor of random effects. He maintains that without theoretical grounds or empirical
evidence to substantiate the random effects assumption (no correlation between the
unobserved variables and the independent variables), bias and consistency considerations

should lead one to select ﬁxed effects. The ﬁxed effects approach has become

85

increasingly popular in sociology recently (Conley and Springer 2001) and has been used
in studies of the effects of welfare spending on infant mortality rates (Conley and
Springer 2001), the efﬁcacy of the Head Start program (Currie and Thomas 1995), and in
studies of crime (Jacob and Lefgren 2003).

Choosing between ﬁxed and random effects is also related to the inferences that
are to be drawn from the research. If inferences are going to be limited to the effects in
the model, then the effects should be considered ﬁxed (Beck and Katz 1996). That is the
case in this study, wherein inferences are not drawn outside of the non-metropolitan
counties in right-to-work states under study. The ﬁxed effect model effectively factors
out the unobserved differences between cases (counties) in analyzing the effect of the
regressors on the dependent variab1e(s). For these reasons, this study estimates models
with ﬁxed effects for each county.

Sociologists are slowly catching on to the beneﬁts longitudinal methods could
have for their studies. In a 2004 review of the use of panel data in sociology, Halaby
commented, “sociologists have been slow to capitalize on the advantages of panel data
for controlling unobservables that threaten causal inference in observational studies”
(Halaby 2004, 507). He explains that panel data are particularly suited to the types of
non-experimental data commonly used in social research. The issue of causality is often
difﬁcult to deal with in such studies. But according to Halaby, the problem of drawing
causal inferences is generally related to the unobservables, which is exactly where panel
techniques make their contributions. As discussed in the previous chapter, the tendency
to rely on cross-sectional as opposed to longitudinal data was one of the critiques of the

energy boomtown research (Smith, Krannich, and Hunter 2001).

86

It is worth noting, however, that these models cannot control for variables that do
change over time, and are not explicitly included in the model (and are therefore
unmeasured), such as changes in local policing policies over time (Gordon et al. 2004).
Additionally, longitudinal data are vulnerable to two types of non-sampling error: non-
response and measurement error (Skinner 2000). Non-response is particularly
problematic for longitudinal studies because often the researcher is trying to obtain data
from the same cases across a period of time. This is a common problem when
individuals are the unit of analysis; however, even in studies such as this one with geo-
political areas as the unit of analysis, some data can be missing. As mentioned earlier, in
this study information on poverty and income was not available for the years 1994 and
1996. If there are no missing values then the data set is referred to as a balanced panel,
and where there are missing values (such as in this study) the data set is described as an
unbalanced panel (Yaffee 2003). Measurement error can also be problematic and can be
compounded by the fact that several years of data are included in the study. In this study
measurement error could have occurred in the ﬁrst place with the collection of the data,
subsequently with the entry of the data, and ﬁnally in the compilation of the data for all
of the years from all of the sources for this study. Efforts were made to ameliorate
potential errors, such as examining descriptive statistics for each variable at each year and

cross-referencing the data.

87

The Count Dimension of the Dependent Variables

Historically, much aggregate-level criminological research has been conducted on areas
with large populations, such as cities, where the number of crimes reported are generally
large enough to use OLS regression techniques. However, with units of analysis
containing small populations (such as the rural counties analyzed in this study), some of
the assumptions of OLS regression that justify small sample estimation cannot adequately
be met; most speciﬁcally the assumptions of homogeneity of error variance and normal
error distributions are frequently violated (Osgood 2000). In brief, the ﬁrst problem is
that predicting crime rates become less precise when the population of the geographic
unit (i.e., the county) decreases. Therefore, the error variance and population size are
inversely related, which violates the assumption of homoscedasticity. The second
problem is that when the population base and crime counts are small (likely in rural
counties), the assumption of normality is more likely violated (Lee and Ousey 2001).
The error term is the composite of the measurement errors in crime and the ability of the

model to predict crime, and both elements are likely to be larger in small communities.

Factor Analysis and Scale Creation

One way to mitigate the assumption violations of OLS regression encountered when a

crime count is the dependent variable is to convert the count to a rate and combine

numerous crime rate variables together to create a crime rate scale. Doing so addresses

the problem of analyzing small counts because the counts are converted to rates and then

88

combined together into one scale creating a more normal distribution of scores;
consequently, small counts no longer present a problem. An additional general advantage
of using scales when possible is that “analyses of factors are usually more powerful and
reliable than measures of individual items” (Leary 1991, 170).

Therefore, in this study one of the techniques used involves factor analyzing the
crime data in order to create scales. Principal Components Analysis (PCA) is used to this
end. The PCA method does not uncover latent factors or variables as traditional factor
analysis does. Instead, this technique is employed when the goal is data reduction and
does not entail testing a hypothesis (Fabrigar, Wegener, MacCallum, and Strahan 1999;
Henson and Roberts 2006). In a review of exploratory factor analysis, Fabrigar,
Wegener, MacCallum, and Strahan (1999) describe the purpose of PCA as follows: “the
goal of PCA is to account for variance in the measured variables. That is, the objective
of PCA is to determine the linear combinations of the measured variables that retain as
much information from the original measured variables as possible” (p. 275). The model
ﬁtting procedure (used to estimate the factor loadings) utilized here is iterative principal
factors. This method requires no distributional assumptions (Fabrigar, Wegener,
MacCallum, and Strahan 1999).

It should be acknowledged that there are a few issues that can complicate factor
analysis. These are identiﬁed as sampling variability, selection bias, and measurement
errors (Kim and Mueller 1978, 64-70). The ﬁrst issue, sampling variability, is not
problematic in this context because a sample of non-metropolitan counties in right-to-
work states is not used; rather all of the counties meeting these criteria are included in the

analysis. The second issue, selection bias, cannot be excluded in this case, since “no

89

researcher can avoid making a certain number of judgmental decisions” (Kim and
Mueller 1978, 68). This refers to the bias involved in selecting the variables to be
included in the factor analysis, Admittedly, the crime variables selected for inclusion in
the factor analysis described in the subsequent chapters are those that are of theoretical
interest, and the inclusion of other variables might have resulted in different factor
structures. Finally, the factor analysis could be affected by measurement errors. Factor
analysis can accommodate random measurement error; however, systematic errors are
more difﬁcult to deal with. It is not apparent that the data used in this study are

vulnerable to systematic errors.

Negative Binomial Regression

Recent criminological studies examining aggregate crime rates with expected small
counts have instead utilized regression models based on the Poisson distribution rather
than OLS regression for the reasons noted above (Krivo and Peterson 2004; Lee,
Martinez, and Rosenﬁeld 2001; Lee and Ousey 2001; Osgood 2000; Rosen et a1. 2003).
In these models, the population at risk is set as the exposure variable, whereby a variable
with the log of the population is created and the coefﬁcient is ﬁxed at 1. This then
transforms such models into analyses of rates rather than counts of crimes (Krivo and
Peterson 2004; Osgood 2000).

As Osgood (2000) demonstrated in his article in the Journal of Quantitative
Criminology, however, the basic Poisson regression model assumes that the variance

equals the mean (for a detailed description of the Poisson distribution, see Long 1997).

90

This assumption is often violated in analyses of crime data. Violating this assumption
produces underestimates of the standard errors and misleading signiﬁcance test results.
In instances of overdispersion (where variance exceeds the mean), negative binomial
regression (using Poisson distribution) is preferred, as it allows for overdispersion (Long
1997; Osgood 2000). There is a notable consequence to using negative binomial
regression: the negative binomial standard errors are larger than the standard errors in
Poission regression (Allison and Waterman 2002), making it a more conservative test
(resulting in more difficulty obtaining signiﬁcant results).

Due to the discrete count nature of the dependent crime variables used in this
study both factor analysis with OLS regression and negative binomial regression of
counts/rates are used to overcome the difﬁculties posed by the data. Both techniques
have their strengths and limitations, and as will be demonstrated in the subsequent
chapters, the use of both techniques helps to provide a more complete picture of crime
over time in counties with slaughterhouses and more conﬁdence that the results are

robust.

Methodological Strengths and Limitations

As with any study, this one has design-related limitations. The most potentially

consequential limitations are examined below.

(1) Limitations of Generalizabilty. As mentioned earlier, because only counties that

are not in or adjacent to metropolitan areas in right-to-work states are included in

91

(2)

this study, the results cannot be generalized to counties in states without right-to-
work laws and to counties in or adjacent to metropolitan areas. Subsequent
research expanding these delimitations might provide interesting information

about the effects of labor unions and urbanization.

Limitations of aggregate-level data. A quantitative, aggregate-level analysis was
chosen to examine whether the few slaughterhouse community studies conducted
thus far are generalizable, to assess the applicability of the causes of crime in
slaughterhouse communities theorized in the literature (which had never been
tested), and to compare the effects of slaughterhouses to other industries. The
aggregated-level of the data, however, poses two limitations: (a) due to the high
level of aggregation the effects of slaughterhouses might be muted. This could
make the analysis rather conservative; (b) these data provide a broad picture, but
do not enable gaining a clear understanding of the dynamics in these
communities, such as who is actually committing the crimes, or if some jobs in
slaughterhouses are more problematic than others. This difﬁculty is shared by
many ecological investigations of crime (see, for example, Camasso and
Wilkinson 1990; Kawachi, Kennedy, and Wilkinson 1999). As Camasso and
Wilkinson (1990) point out in their study of child maltreatment in boomtowns,
due to the aggregate level of the data it is not possible to determine if increases in
child maltreatment are the result of the isolation of newcomers, problems among
long term residents, or a form of social disorganization that cuts across all of the

social groups.

92

Therefore, there is a risk of making inaccurate ecological inferences, or
succumbing to ecological fallacy. King (1997) explains that there are two
statistical issues that can cause ecological inferences to be incorrect. The ﬁrst is
referred to as aggregation bias. It refers to the loss of individual-level information
that takes place when the individual-level data are aggregated. Inferences that fail
to take this loss of individual-level information into account are consequently
biased. The second set of issues King identiﬁes entails statistical problems
endemic to this type of research, such as heteroskedasticity, which are not taken
into consideration when inferences are made. The consequence of ecological
fallacy is that the observed relationship at the aggregate level may be a ‘statistical
artifact’ of dynamics at the individual level.

Part of this problem might be ameliorated by the TSCS design of this
study. For instance, Conley and Springer (2001) point out in describing their
ﬁndings on the welfare state and infant mortality rates that it is possible that the
reduction in infant mortality they observed resulting from state investment is not
occurring among those in the population who are actually receiving the state
beneﬁts. They argue, however, that their use of the ﬁxed-effects approach
mitigates this possibility because their analysis focuses on within-country changes
in state spending and the infant mortality rate. In contrast, a cross-sectional
approach would be more vulnerable to ecological fallacy.

In these types of studies, in order to understand the macro processes, the
individual-level processes must be bracketed out (at least for the time being).

Accordingly, in a study of the effects of immigration on crime, the authors

93

(3)

explain, “in order to focus on immigration as a macrological process this study
ignored individual-level issues regarding whether either the perpetrators or
victims of homicide were immigrants. We have directed our attention to the
conditions of communities that are associated with variation in group levels of
criminal poverty, immigrant-status, and other factors within communities” (Lee,
Martinez, and Rosenﬁeld 2001, 573). Without individual-level data it is also not
possible to tell if effects such as these are actually community effects or
aggregated individual effects (Miles-Doan 1998).

Although one should be wary of conclusions drawn about individuals
based upon aggregated data, “aggregate data are sometimes useful even without
inferences about individuals” (King 1997, 7). Thus, while this study cannot
permit one to draw conclusions about the individuals who work in
slaughterhouses, it nonetheless is a ﬁrst step in better understanding what is
occurring in slaughterhouse communities (and what is not occurring in
communities without slaughterhouses). A subsequent ethnographic study would

permit a more nuanced analysis of what is occurring at the individual level.

Are slaughterhouse workers predisposed to disruptive behavior? Relatedly, this
study cannot control for the possibility that work in slaughterhouses might attract
people who are already predisposed to or involved in disruptive behavior and
consequently that the work itself has not caused their anti-social behaviors (akin
to the problem of knowing if there are true ecological-level effects of community

on crime, or if those with prior antisocial behavior are selected or self-select into

94

(4)

(5)

certain communities (Kawachi, Kennedy, and Wilkinson 1999)). However, there
is nothing in the literature on slaughterhouse workers to indicate that this is the
case. On the contrary, the literature suggests that the one thing slaughterhouse
workers have in common is that they are hard working individuals who take these
jobs because they are available and they are desperate to support themselves and
their families. Again, subsequent ethnographic would be required to investigate

this possibility.

Spatial Autocorrelation. The possibility of spatial autocorrelation is also another
potential limitation of this research. One of the assumptions of OLS regression is
that errors are independent across the cross-sections (F ording 1997). However,
when dealing with geo-political areas they are rarely spatially independent. For
instance, crime in one neighborhood might inﬂuence crime in an adjacent
neighborhood; additionally variables often linked to crime (such as poverty)
generally span across geographic boundaries (Kubrin and Weitzer 2003).
Techniques have been developed to deal with the issue of spatial autocorrelation
in OLS regression; however, techniques for dealing with this issue in Poisson-
based models of discrete events are just now beginning to be developed (Lee and

Ousey 2001).

Variable Limitations. Due to the unit of analysis and the fact that the study period
covered the years 1994 to 2002, there are some variables that could not be

included in the study because they were not available either at the county-level

95

and/or for each year under study. Such variables include more intervening
dimensions of social disorganization, such as community organization
membership and family disruption, as well as other dependent variables, such as

suicide rates.

The design of this study also presents some signiﬁcant beneﬁts, which have been

implicitly acknowledged earlier in this chapter, but are brieﬂy highlighted below:

(1)

The Longitudinal Dimension. The use of longitudinal data and analysis
techniques is advantageous compared to cross-sectional observations. Data
obtained for one year only affords a snapshot into what is transpiring in a
community and the reliance on such data has been critiqued in the boomtown
studies. Communities are very dynamic and examining a cross-section obscures
social change from the researcher’s purview. As will be demonstrated in the
subsequent analysis, the effect of slaughterhouses on communities is dynamic and
has changed over time, as has the level of slaughterhouse employment in the
counties under study. Consequently, one could draw very different conclusions
about the relationships between slaughterhouses and crime depending upon the
year which s/he chose to examine. Examining a series of years, as is done in this
study, permits us to examine social change and draw conclusions that are not
tightly tied to one year.

Additionally, panel or TSCS analysis allows one to control for the time

invariant variables that otherwise would not be included in the model. Doing so

96

(2)

(3)

mitigates against drawing spurious conclusions. Thus, this study is able to control
for factors that might confound the relationship between slaughterhouses and
crime, such as a history of a culture of violence, such as that believed to exist in

the southern United States (Flynn 2002).

Inclusion of Key Control Variables. An additional strength of the design of this
study is the number of control variables that have been included. The inclusion of
these variables makes it possible to assess the applicability of the causes of
increases in crime in slaughterhouse communities hypothesized in the literature
and to better assess the unique effects of slaughterhouse employment. The
partialling out of various effects and the assessment of theories proposed have not
been possible in the community studies conducted thus far that examine the

impacts of slaughterhouses.

Use of Comparison Industries. An additional strength of this design is the
inclusion of several industries that were carefully matched with slaughterhouses
on characteristics such as the risk of injury/illness, the concentration of immigrant
workers, and the routinized nature of the work itself. The inclusion of these
industries allows us to assess whether or not the impacts of slaughterhouses are
unique or if other types of similar industries also result in similar increases in

crime.

97

(4) Examination of Many Counties with and without Slaughterhouses. Finally, due to
the time-intensive nature of ethnographic community studies, only a few
communities have thus far been examined. This study, however, entails an
examination of 581 counties. This sample includes counties that do and that do
not have slaughterhouse located within their boundaries. Thus, this design
enables us to see if the effects found in the ethnographic community studies
where large slaughterhouses have opened are generalizable to other counties with
slaughterhouses, including those with smaller slaughterhouses and those that have
been in operation for a number of years. Additionally, it permits a comparison

between crime trends in communities with and without slaughterhouses.

Summary

This chapter has provided an overview of the design of this study, a description of the
variables included in the models, the sources of the data, and how the data are analyzed.
It has been demonstrated that numerous issues have been taken into consideration in
designing this study; most speciﬁcally the longitudinal dimension and the discrete count
nature of the dependent variables. Like all studies, this one has limitations, but it also
possesses several signiﬁcant strengths and notably examines an area where there is a
clear dearth of information and understanding. The knowledge generated by this research

is examined in the subsequent chapters, which detail and discuss the analyses of the data.

98

Chapter 3

DESCRIBING THE EFFECTS OF SLAUGHTERHOUSES

Science may be described as the art of systematic oversimplification. — Karl Popper

This is the ﬁrst of two chapters which present the results of the data analysis. This
chapter begins with a description of the slaughterhouse variables, followed by an
examination of the bivariate relationships between the variables included in this study.
Next, attention is paid to testing the ﬁrst of the research hypotheses, which addresses the
following question: Do counties with slaughterhouses experience more crime? The next
chapter introduces models which contain control variables and comparison industries in

order to test the second and third hypotheses.

The Slaughterhouse Variables

There are two independent variables of interest in this study: the number of
slaughterhouse establishments in the county and the number of people employed in
slaughterhouses in the county. Table 4 provides the number of county/year combinations
where slaughterhouses are present. There are 5,227 cases in total: 581 counties
multiplied by 9 years minus 2 missing cases (data on slaughterhouse presence and
employment is missing for Kinney, Texas for the years 2001 and 2002. It is unclear why

they did not report data for these two years, but the small size of the county — the

99

population is slightly greater than 3000 persons — might have something to do with issues

related to reporting).

Table 4. Frequency Distribution of Slaughterhouses for County/Y ear

Frequency Percent

 

 

Do not have slaughterhouse 3,706 70.9
Have slaughterhouse(s) 1,521 29.10
Total 5,227 100

Table 5 presents data illustrating trends in the two slaughterhouse variables over
the time period under study. The table reveals a dramatic increase in the mean number of
establishments in 1998. This increase is due to the inclusion of ‘custom slaughter’
facilities into the slaughterhouse category (the consequences of which will be elaborated
upon in the next chapter). The slaughterhouse employment variable shows a general
increase over time, peaks in 1999 at a mean value of 73.94 and tapers off afterward. It is
also worth noting that in the counties under study the maximum number of
slaughterhouse employees increased in 1999 and 2000 to 7500 workers. This maximum
value is the result of an increase in the number of slaughterhouse employees in Finney
County, Kansas (which has been the focus of many of the slaughterhouse community
studies described earlier) from 3750 employees in 1998 to 7500 in 1999 and 2000.11

Later in this chapter we examine whether the crime rates in the 1,521 county/year
combinations that contain slaughterhouses differ signiﬁcantly from the 3,706 county/year
cases that do not and in the next chapter we examine how the above trends in

slaughterhouse employment are related to the crime rates in these counties.

100

Table 5. Trends in Slaughterhouse Establishment and Employment Variables,

 

 

 

1994-2002

Slaughterhouse Establishments Slaughterhouse Employment

Mean Minimum Maximum Mean Minimum Maximum
1994 0.28 0 6 57.14 0 3750
1995 0.28 O 4 60.08 0 3750
1996 0.29 0 4 67.02 0 3750
1997 0.28 0 4 63.33 0 3750
1998 0.47 0 5 64.86 0 3750
1999 0.44 O 5 73.94 0 7500
2000 0.44 0 5 71.89 0 7500
2001 0.44 0 5 62.55 0 3750
2002 0.38 0 4 57.49 0 3750

 

Bivariate Relationships

This section details the bivariate relationships between the variables of interest in this
study. First, the correlations between the independent variables (the number of
slaughterhouse employees and the number of slaughterhouse establishments), the control
variables, and the summary arrest rate variables are examined. Next, the correlations
between the independent variables, control variables, and speciﬁc arrest rate variables
are examined. Finally, the correlations between offense report rate variables (of which
there are fewer) and the independent and control variables are discussed.

In examining the below results it is important to remember that “[t]here are good
reasons to exercise caution in the interpretation of the zero-order correlations” (Baron and
Straus 1988, 133). It is possible that one or more of the control variables are suppressing
the bivariate relationship between the slaughterhouse variables and the crime variables

(arrests and reports). Tables 3 through 5 do demonstrate that some of the independent

101

variables are intercorrelated. Multivariate analyses often helps to clarify such issues. For
instance, in their examination of the causes of homicide, Baron and Straus (1988) found a
small correlation between their Legitimate Violence Index and homicide. However,
multivariate regression demonstrated that once the effects of the other independent
variables were held constant, their index became a signiﬁcant predictor of homicide.
Further, of particular relevance in the case of the data analyzed here is the fact that the
correlation coefﬁcient is really a transformation of the b coefﬁcient from OLS regression
and is therefore subject to all of the problems in using OLS regression with discrete count

data described earlier.

Summary Arrest Rate Variables

Table 6 presents measures of central tendency (mean and median) and dispersion
(standard deviation, minimum and maximum values), in addition to the bivariate
correlations for the independent variables, control variables, and the summary arrest
variables (total number of arrests, total number of index offenses, total number of violent
crimes, and total number of property crimes). Subsequent tables report the descriptive
statistics for the speciﬁc arrest variables and report variables.

Before discussing the bivariate correlations between the predictor variables and
the summary arrest rate variables it is worth noting that there is quite a bit of variation in
some of these variables. The mean number of slaughterhouse employees in these

counties is fairly small (64.26) and there is notable variation (standard deviation =

102

402.36). Understandably, there is not as much variation in the number of slaughterhouse

establishments themselves, and the mean number (0.37) is fairly small.

Table 6. Descriptive Statistics and Zero Order Correlations, Independent

Variables, Control Variables, and Summary Arrest Variables

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1. 2. 3. 4. 5. 6. 7.
1. Slaughterhouse 1
employment
2. Slaughterhouse 0.338 1
establishments
3. Total arrests 0.021 0.015 1
4. Index offenses 0.024 0.004 0.754 1
5. Violent crime -0.013 -0.05 0.545 0.803 1
6. Property crime 0.037 0.027 0.755 0.967 0.625 1
7. Number in 0.08 0.133 0.398 0.462 0.361 0.45 1
poverty
8. -0.062 -O.12 0.322 0.37 0.381 0.322 0.344
Unemployment
rate
9.1mmigration 0.513 0.188 0.095 0.095 0.052 0.102 0.366
10. Migration -0.176 -0.08 -0.035 -0.016 0.023 -0.031 -0.139
11.Non-white or 0.113 0.058 0.362 0.421 0.354 0.401 0.822
Hispanic
residents
12. Population 0.08 0.054 0.358 0.47 0.3122 0.482 0.548
density
13.Numberof 0.168 0.213 0.344 0.392 0.294 0.388 0.88
males
14. Males age15- 0.187 0.195 0.304 0.35 0.257 0.349 0.849
34
Mean 64.26 0.37 3348.05 364.69 99.07 265.48 2239.99
Median 0 0 3002.79 278.11 67.97 195.85 1281.50
Standard 402.36 0.65 2605.23 342.31 111.06 260.99 2548.04
Deviation
Minimum 0 0 0 0 0 0 8.25
Maximum 7500 6 27044 4077 1288 2903 21450

 

 

 

Note: Correlations and descriptive statistics for dependent variables are shown as crime rates per 100,000
population.

103

Table 6 (cont’d).

 

10.

ll.

12.

13.

14.

 

 

1 .
Slaughterhouse
employment

 

2.
Slaughterhouse
establishments

 

3. Total arrests

 

4. Index
offenses

 

5. Violent crime

 

6. Property
crime

 

7. Number in
POVCTIL

 

8.
Unemployment
rate

 

9. Immigration

0.163

 

10. Migration

-0.071

-0.198

 

1 I. Non-white
or Hispanic
residents

0.426

0.477

-0.231

 

12. Population
density

0.171

0.169

0.007

0.425

 

13. Number of
males

0.173

0.439

-0.119

0.7

0.597

 

14. Males age
15-34

0.187

0.51

-0.l76

0.727

0.552

0.959

 

 

Mean

5.02

15.95

-33.84

3077.2

22.2

6816.0

1988.33

 

Median

4.20

-25

764.5

12.54

4529

1184

 

Standard
Deviation

3.06

47.48

260.96

5449.5

32.09

6908.2

2394.04

 

Minimum

-6

-4083

01

34

3

 

 

Maximum

 

38.4

 

777

 

3281

 

47049

 

519

 

66194

 

22118

 

104

 

There is also signiﬁcant variation in the crime rates in these counties (especially
given that they are all rural). For instance, the standard deviation for the total number of
arrests variable (variable #3 in the table) is 2,605.23, the minimum value is 0 and the
maximum value is 27,044. The fact that the minimum value of all of the summary crime
variables is 0 is indicative of the small populations and little crime in some of the
counties and additionally the imperfect nature of collecting and compiling crime
statistics.

The control variables provide an economic and demographic snapshot of these
communities across the time period under study. The mean number of people living in
poverty in these counties (2239.99) works out to be an average of 16.16% of the county
populations, which is slightly higher than the 2000 national average of 12.4% (US
Census Bureau 2000 Fact Sheet). The mean unemployment rate of approximately 5% is
consistent with the national average. Overall these counties do not appear to have
unusual levels of economic hardship; however, as indicated by a maximum value of
38.4% unemployment, some counties (this speciﬁc one is a small county in Texas) during
the time period under study have experienced severe economic hardship.

The means of the net immigration and migration variables indicate that on
average there has been a small mean increase in the number of immigrants moving to
these non-metropolitan counties in right-to-work states, but the amount of migration out
of these counties is nearly double that value. Additionally, there is less variation in the
net number of immigrants into these counties than there is in the net number of migrants.

The relatively small mean population density of 22.2 people per square mile is consistent

105

with the fact that only non-metropolitan counties not adjacent to metropolitan counties
were included in this study.

The mean number of non-white or Hispanic residents (3077.25) represents 22% of
the county population on average. This is less than the national average of approximately
30%, which is not surprising given that the counties included in this study are from the
south and Midwest, which according to the 2000 Census have the highest concentrations
of white residents (Grieco 2001).

The mean number of males represents 49% of the average county population,
which is consistent with the national average of 49.1% (US Census Bureau 2000 Fact
Sheet). The number of males age 15-34 accounts for 14% of the average population of
these counties, or 29% of the male population in these counties.

The bivariate correlations between slaughterhouse employment and
establishments and the control and dependent variables are of particular interest and are
examined in some detail in what follows. The number of slaughterhouse employees and
the number of slaughterhouses in the counties have fairly small correlations with the
summary arrest variables; however the slaughterhouse employment variable generally
has higher correlations with the crime variables than the slaughterhouse establishment
variable. The largest of these correlations is between slaughterhouse employment and
arrests for property crime (r = 0.03 7). The correlations between slaughterhouse
employment and establishments and arrests for violent crimes are negative (but small).

The correlations between the slaughterhouse variables and the control variables
range from weak to strong and are both negative and positive. The correlations between

the economic variables and the slaughterhouse variables are fairly small. The correlation

106

between slaughterhouse employment and establishments and the number of people in
poverty are 0.08 and 0.133 respectively. The relationship between slaughterhouse
employment and establishments and unemployment is small but negative (r = -0.062 and
-0.12).

Notably, there is a strong positive correlation (r = 0.513) between the number of
slaughterhouse employees in the counties and the net amount of international
immigration, which is not surprising given the demographic shifts in the industry
discussed in the literature review chapter. Interestingly there is a negative relationship
between net migration into a county and slaughterhouse employment and establishments
(r = -0.176 and -0.08). This negative relationship runs counter to the theorizing in the
literature.

The positive correlations between slaughterhouse employment and the number of
non-white or Hispanic residents (r = 0.113), the number of males (r = 0.168) and the
number of males age 15 to 34 (r = 0.187) were anticipated. However, by themselves
these relatively small correlations do not provide much evidence in support of the
contention in the literature that there are profound demographic differences in counties
containing slaughterhouses. This could be related to the fact that this analysis includes
slaughterhouse employment and establishments on all scales, not just counties that have
large-scale facilities and employment therein.

Also of particular interest are the relationships between the control variables and
the summary arrest variables, since these control variables were included because they
were hypothesized in the literature as having signiﬁcant effects on crime increases in

slaughterhouse communities. We ﬁnd here that the number of people in poverty has

107

moderate-strong correlations with the four summary arrest variables (r = 0.398 with total
crime, r = 0.462 with index crime, r = 0.361 with violent crime, and r = 0.45 with
property crime). Unemployment has slightly smaller correlations with these variables —
with the exception of violent crime, which is slightly larger (r = 0.322 with total crime, r
= 0.37 with index crime, r = 0.381 with violent crime, and r = 0.322 with property crime).

The net international immigration variable has relatively small correlations with
the summary crime variables (r = 0.095 with total crime, r = 0.095 with index crime, r =
0.052 with violent crime, and r = 0.102 with property crime). The net migration variable
has small but negative correlations with three of the summary crime variables (r = -0.035
with total crime, r = -0.016 with index crime, and r = -0.139 with property crime), and a
small positive correlation with violent crime (r = 0.023).

The relationship between the rest of the control variables and the summary crime
variables are generally consistent with what is hypothesized in the literature. Population
density has variable but fairly strong correlations with the four summary crime variables
(r = 0.358 with total crime, r = 0.47 with index crime, r = 0.312 with violent crime, and r
= 0.48 with property crime). The number of non-white or Hispanic residents is similarly
correlated with the crime variables (r = 0.362 with total crime, r = 0.421 with index
crime, r = 0.3 54 with violent crime, and r = 0.401 with property crime). The number of
males is also positively correlated with the summary crime variables (r = 0.344 with total
crime, r = 0.392 with index crime, r = 0.294 with violent crime, and r = 0.3 88 with
property crime), and the strength of the correlations between the number of young males

and the crime variables are slightly less than those for the number of males in general (r =

108

0.304 with total crime, r = 0.35 with index crime, r = 0.257 with violent crime, and r =
0.349 with property crime).

In sum, the strongest correlations with the summary crime variables are observed
among the economic variables (number of people in poverty and unemployment rate) and

the population density variable.

Speciﬁc Arrest Rate Variables

Table 7 presents the descriptive statistics and correlations for the predictor variables and
the speciﬁc arrest variables of interest. The mean number of arrests for these speciﬁc
offenses, especially the index offenses (murder, rape, robbery, and aggravated assault),
are fairly low. The most common of the included offenses is other (meaning non-
aggravated) assaults. Like the summary arrest variables, the speciﬁc crime arrest
variables demonstrate quite a bit of variation across the counties.

The slaughterhouse employment and establishment variables have fairly small
correlations with the speciﬁc arrest variables and some of the correlations are negative.
The positive correlations are found between the slaughterhouse employment and
establishment variables and rape (r = 0.036 and 0.022 respectively), other assaults (r =
0.076 and 0.026), sex offenses (r = 0.029 and 0.016), and disorderly conduct (r = 0.029

with the slaughterhouse employment variable).

109

Table 7. Descriptive Statistics and Zero Order Correlations, Independent
Variables, Control Variables, and Individual Arrest Variables

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1. 2. 3. 4. 5. 6. 7. 8. 9.
1. Slaughterhouse 1
employment
2. Slaughterhouse 0.338 1
establishments
3. Murder -0.01 1 -0.026 1
4. Rape 0.036 0.022 0.1 16 1
5. Robbery -0.008 -0.025 0.229 0.249 1
6. Aggravated -0.016 -0.052 0.216 0.297 0.5 1
Assault
7. Other assaults 0.076 0.026 0.194 0.286 0.495 0.482 1
8. Sex offenses 0.029 0.016 0.092 0.147 0.182 0.218 0.325 1
9. Offenses -0.004 -0.017 0.129 0.168 0.33 0.269 0.352 0.193 1

_2_1gainst the family

10. Disorderly 0.029 -0.015 0.161 0.188 0.384 0.345 0.522 0.185 0.312
conduct
11. Number in 0.08 0.133 0.145 0.206 0.436 0.311 0.445 0.122 0.216
poverty
12. -0.062 -0.12 0.178 0.151 0.331 0.349 0.319 0.102 0.235
Unemployment
rate
13. Immigration 0.513 0.188 -0.009 0.038 0.068 0.048 0.144 0.05 0.011
14. Migration -0.l76 -0.08 -0.001 -0.006 -0.031 0.032 -0.089 0.019 -0.017
15. Non-white or 0.113 0.058 0.168 0.197 0.469 0.294 0.466 0.12 0.253
Hispanic residents
16. Population 0.08 0.054 0.095 0.145 0.31 0.287 0.36 0.098 0.166
density
17. Number of 0.168 0.213 0.086 0.16 0.335 0.261 0.398 0.126 0.147
males
18. Males age 0.187 0.195 0.075 0.145 0.307 0.226 0.355 0.108 0.127
15-34
Mean 64.26 0.37 3.56 6.44 8.3 80.21 289.45 14.74 48.87
Median 0 0 0 0 0 53.51 218.76 7.67 15.81
Standard 402.4 0.65 11.64 1 1.76 15.95 94.96 283.35 23.63 92.21
Deviation
Minimum 0 0 0 0 0 0 0 0 0
Maximum 7500 6 366.1 245.1 174.98 1143.4 2108.6 530.51 1680.2

 

Note: Correlations and descriptive statistics for dependent variables are shown as crime rates per 100,000

population.

110

 

Table 7 (cont’d).

 

10.

l3.

l4.

16.

18.

 

 

1. Slaughter-
house
employment

 

2. Slaughter-
house
establish-
ments

 

3. Murder

 

4. Rape

 

5. Robbery

 

6.
Aggravated
Assault

 

7. Other
assaults

 

8. Sex
offenses

 

9. Offenses
against the
family

 

10.
Disorderly
conduct

 

1 1. Number
in poverty

0.311

 

12. Unemp-
loyment rate

0.284

0.344

 

13.
Imrrﬂration

0.023

0.366

0.163

 

14.
Migration

-0.092

-0.l39

-0.07l

-0.l98

 

15. Non-
white or
Hispanic
residents

0.322

0.822

0.426

0.477

 

16.
Population
density

0.182

0.548

0.171

0.169

0.007

0.425

 

17. Number
of males

0.226

0.88

0.173

0.439

-0.119

0.7

0.597

 

18. Males

age
15-34

0.203

0.849

0.187

0.51

-0.l76

0.727

0.552

0.959

 

 

Mean

134.18

2239.9

5.02

15.95

-33.84

3077.3

22.2

6816.1

1988.3

 

Median

70.72

1281.5

4.20

-25

764.5

12.54

4529

1184

 

Standard
Deviation

193.69

2548

3.06

47.48

260.96

5449.5

32.09

6908.2

2394

 

Minimum

8.25

-4083

0.1

34

 

 

Maximum

 

2098.6

 

21450

 

38.4

 

777

 

3281

 

47049

 

519

 

66194

 

22118

 

lll

 

Quite a bit of variation is observed in the correlations between the control
variables and the speciﬁc crime variables. The murder and rape variables have fairly low
correlations with the control variables (the highest of which is r = 0.206 between rape and
the number of people in poverty). Robbery, aggravated assault, and other assaults are
moderately-strongly positively correlated with the economic variables and some of the
demographic variables; sex offenses and offenses against the family are less strongly
correlated with the control variables; and disorderly conduct is most strongly related to
the economic variables and the number of non-whites or Hispanics. It is worth re-
emphasizing that these variables represent rates of arrests, an activity which is imbued
with much police discretion, especially for certain types of offenses. Therefore, for
instance, the high correlation between the number of non-white and/or Hispanic
individuals in counties and the number of arrests for disorderly conduct might be at least
partly due to the willingness of police ofﬁcers to arrest minorities for certain types of
offenses, such as disorderly conduct, which are highly discretionary. The correlations
appear fairly consistent with the literature, with one notable exception: the immigration
and migration variables do not have strong correlations with the crime variables (the
largest is r = .114 between net immigration and arrests for “other assaults”), and in fact,

many of the correlations are negative.

Crime Report Variables

Data on the number of offenses reported provide a nice supplement to the arrest data.

Table 8 presents the crime report data and includes one summary report variable (Index

112

reports) and several speciﬁc crime report variables. All of the speciﬁc report variables
available (with the exception of larceny) are included here because they are all used in
analyses detailed in the subsequent chapter, whereas only some of the available arrest
variables are included in the analyses. Consequently, data are presented for three report
variables not included in the arrest analysis: burglary, motor vehicle theft, and arson.

The slaughterhouse employment and establishment variables have positive
correlations with all of the crime report variables, with the exception of murder. The fact
that the values of the correlations for the murder arrest rate and the murder report rate
with slaughterhouse employment and establishments are so similar is not surprising
because due to the nature of the crime the arrest and report rates are very similar (as
evidenced by only a 0.31 difference in the mean rates of murder arrests and murder
reports).

Again, however, the correlations between the slaughterhouse and crime variables
are modest, ranging from a low of r = -0.22 between slaughterhouse establishments and
murder to a high of r = 0.16 between slaughterhouse establishments and larceny. It is
interesting to note that the correlations between the slaughterhouse establishment variable
and the report rates tend to be slightly higher than that between the slaughterhouse
employment variable and the report rates, which is the reverse of the general trend
observed with the arrest rate data, where the correlations for the slaughterhouse
employment variable was generally slightly higher than that for the establishment
variable. Overall, there is a stronger positive relationship between the report rates
and the slaughterhouse variables than between the arrest rates and the slaughterhouse

variables. For instance, the correlation between the rape arrest rate and slaughterhouse

113

Table 8. Descriptive Statistics and Zero Order Correlations, Independent
Variables, Control Variables, and Crime Report Variables

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1. 2. 3. 4. 5. 6. 7. 8. 9.
1. Slaughter- 1
house
employment
2. Slaughter- 0.338 1
house
establish-
ments
3. Index 0.095 0.132 1
reports
4. Murder -.004 -0.022 0.197 1
5. Rape 0.086 0.082 0.544 0.135 1
6. Robbery 0.031 0.028 0.595 0.219 0.35 1
7. Assault 0.01 0.017 0.679 0.22 0.406 0.51 1 1
8. Burglary 0.032 0.059 0.847 0.226 0.467 0.59 0.601 1
9. Motor 0.076 0.1 17 0.78 0.18 0.438 0.494 0.532 0.672 1
vehicle theft
10. Arson 0.009 0.029 0.378 0.111 0.264 0.186 0.293 0.355 0.344
1 1. Number 0.08 0.133 0.526 0.144 0.326 0.58 0.385 0.498 0.455
in poverty
12. Unemp— -0.062 -0.12 0.133 0.136 0.067 0.24 0.248 0.21 0.103
loyment rate
13. 0.513 0.188 0.283 0.026 0.18 0.173 0.15 0.204 0.261
Immigration
14. -0.176 -0.08 -0.074 -0.006 -0.101 -0.116 -0.016 -0.039 -0.031
ﬂgration
15.Non- 0.113 0.058 0.443 0.162 0.2737 0.618 0.35 0.443 0.371
white or
Hispanic
residents
16. 0.08 0.054 0.416 0.082 0.258 0.401 0.294 0.318 0.355
Population
density ,
17. Number 0.168 0.213 0.562 0.081 0.346 0.492 0.341 0.455 0.475
of males
18. Males 0.187 0.195 0.517 0.073 0.32 0.457 0.316 0.4138 0.425
age
15-34
Mean 64.26 0.37 1827.9 3.25 14 15.743 142.1 429.53 87.83
Median 0 0 1523.5 0 0 0 83.20 367.60 70.19
Standard 402.36 0.65 1611.9 8.82 20.88 31.55 181.13 390.19 88.16
Deviation
Minimum 0 0 0 0 O 0 0 0 0
Maximum 7500 6 9630.4 173.71 245.1 405.34 2808.7 3076.9 697.67

 

Note: Correlations and descriptive statistics for dependent variables are shown as crime rates per 100,000

population.

114

 

Table 8 (cont’d).

 

10.

ll.

12.

13.

14.

15.

17.

 

 

1 .
Slaughter-
house
employme
nt

 

2.
Slaughter-
house
establish-
ments

 

3. Index
reports

 

4. Murder

 

5. Rape

 

6. Robbery

 

7. Assault

 

8. Buglary

 

9. Motor
vehicle
theft

 

10. Arson

 

1 1.
Number in

poverty

0.169

 

l 2.
Unemp-
loyment
rate

0.044

0.344

 

13.
Immigratio
11

0.071

0.366

0.163

 

14.
Migration

-0.038

-0.139

-0.071

-0.l98

 

15. Non-
white or
Hispanic
residents

0.114

0.822

0.426

0.477

-0.231

 

16.
Population
density

0.1443

0.548

0.171

0.169

0.007

0.425

 

17.
Number of
males

0.171

0.88

0.173

0.439

-0.119

0.7

0.597

 

18. Males
age
1 5-34

0.147

0.849

0.187

0.51

-0.l76

0.727

0.552

0.959

 

 

Mean

11.83

2239.99

5.02

15.95

-33.84

3077.25

22.2

6816.09

1988.33

 

Median

1281.50

4.20

-25

764.5

12.54

45 29

1184

 

Standard
Deviation

22.51

2548.04

3.06

47.48

260.96

5449.53

32.09

6908.21

2394.04

 

Minimum

8.25

-4083

1

0.1

34

3

 

 

Maximum

 

528.17

 

21450

 

38.4

 

777

 

3281

 

47049

 

519

 

66194

 

22118

 

115

 

employment is 0.036, whereas the correlation between the report rate and slaughterhouse
employment is 0.086.

The poverty control variable is moderately-strongly related to the crime variables
(with the exception of arson and murder, which are weaker), but the unemployment
variable performs less well. The immigration and migration variables continue to display
relatively low correlations with the crime variables (especially the migration variable,
which is negatively correlated with all of the crime report variables), compared to the
other control variables. Population density is particularly correlated with economically
motivated crimes, such as burglary, larceny, and motor vehicle theft, perhaps reﬂecting
the availability of targets and the effects of social distancing. The demographic variables
(number of minorities, males, and young males) are also most strongly related to the
economically-motivated offenses. They are less related to the violent offenses, especially
murder (correlation ranges between 0.073 and 0.162).

Overall, the slaughterhouse variables appear to be only modestly correlated with
the crime variables, although the relationship between the slaughterhouse variables and
the report rates is greater than that with the arrest rates. The control variables are more
strongly correlated with the crime rate variables, although the immigration and migration
variables do not appear to perform as well as has been hypothesized in the literature.

Before proceeding to the more telling multivariate analyses discussed in the next
chapter, it is necessary to examine whether crime is actually higher in counties with

slaughterhouses.

116

Increased Crime in Counties with Slaughterhouses?

As we saw in the previous section, there are positive correlations between the
slaughterhouse variables and many of the crime rate variables. However, the question
remains whether there is a signiﬁcant difference in the crime rates in slaughterhouse and
non-slaughterhouse counties. Thus, in this section we test research hypothesis number I

— general and speciﬁc crime rates will be higher in counties with slaughterhouses.

Examining the Diﬂerences between Counties with and without Slaughterhouses

An easy way to demonstrate that differences in crime rates exist between counties with
and without slaughterhouses is to graph the rates of both across time. Figures 1 through 4
below provide examples using total arrest rate, index offense report rate, rape arrest rate,
and rape report rate.

Figure 1 demonstrates that the total arrest rate is greater in county/year
combinations with slaughterhouses than in those without for the years 1995 through
1998. The reversal of the trend in 1999 likely related to the inclusion of custom
slaughterhouse facilities in the slaughterhouse category, which caused many more
counties to be listed as having slaughterhouses, even if the only ones they contain are
small custom slaughter facilities. Figure 2 shows a consistently higher number of index
reports in counties with slaughterhouses than those without, although the gap narrows

after the slaughterhouse classiﬁcation change.

117

Rate per 100,000

Figure 1. Total Amt Ram in Counties with and without Slaughterhouses

3600

3500

1’
8

+Total arrests - counties with slaughtemouses

3100 +Total enacts - counties Mthout
slaughterhouses

 

3000
1994 1995 1996 1997 1998 1999 2000 2001 2002
Your

The arrest rate for rape illustrated in Figure 3 is the least consistent of the four

ﬁgures. The rate is higher in slaughterhouse counties in the years 1994, 1996, 1998,

2000, and 2001. Again, the lines draw closer together after the change in slaughterhouse

classiﬁcation. The rape report rates in Figure 4 present more consistent results, with the

rate

being greater in counties with slaughterhouses than in counties without for all of the

years under study. While these graphs assist us in visually assessing the differences in

crime between slaughterhouse and non-slaughterhouse counties, they do not provide us

with a means to test whether or not the differences are signiﬁcant. To do so we turn to

another method.

118

Rate W1w,000

appearance

Figure 2. Index Report Rates for Counties with and without Slaughterhouses

2400

2000

1800

“00 +lndex report rates - counties with ‘
slaughterhouses l

+|ndexreponrates« oountieewithout
slaughterhouses

 

1200 .
1994 1995 1996 1997 1998 1999 2000 2001 2002
Year

Figure 3. Rape Arrest Rate in Counties with and without Slaughterhouses

8.5
+ Rape rate — counties with slaughterhouses

8 + Rape rate - counties without slaughterhouses

7.5

8.5

5.5 .

 

1994 1 995 1996 1 997 1998 1 999 2000 2001 2002

119

Figure 4. Rape Report Rates for Counties with and without Slaughterhouses

+ Rape report rates
slaughterhouses

-l— Rape report rates
slaughterhouses

 

1994 1995 1996 1997 1998 1999 2000 2001 2002
Year

A straight-forward way to assess whether or not there is a signiﬁcant difference in
crime rates in counties with and without slaughterhouses is to perform a t test of the
difference in means. Accordingly, t tests were performed on speciﬁc and general or
summary crime variables of interest, with unequal variances assumed and the non-
directional hypothesis that the difference between the two means does not equal 0. Since
the mean crime rate of counties with slaughterhouses is subtracted from the mean crime
rate of counties without slaughterhouses (counties without slaughterhouses — counties
with slaughterhouses = mean difference), negative values of the difference of means and
the t statistics indicate that there are more of those crimes in counties with

slaughterhouses.

120

The values in Table 9 demonstrate that although there is a mean difference in the
total arrest rate between counties with and without slaughterhouses (with slaughterhouse
counties having an increased total arrest rate), this difference is not statistically
signiﬁcant. We also observe an increase in the rate of arrest for property crimes, rape,
and other assaults; however, these differences are also not statistically signiﬁcant. The
only signiﬁcant mean difference in arrest rates in the expected direction is for sex
offenses, which is signiﬁcant at the .05 alpha level.

Table 9. Difference in Crime Rate Means between Counties with and without

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Slaughterhouses
Variable Mean Difference t value p value
Total arrests -58.064 -0.792 0.429
Index offense arrests 1.097 0.111 0.912
Violent crime arrests” 1 1.505 3.666 0.0003
Property crime arrests -10.444 -1 .368 0.1713
Murder arrests 0.656 1.888 0.059
Rape arrests -0.332 -1.032 0.302
Robbery arrests 0.859 1.903 0.057
Aggravated assault 10.228 3.82 0.0001
arrests”
Other assault arrests -5.108 -0.614 0.54
Sex offense arrests* -1.53 -2.192 0.029
Offenses against the 2.407 0.942 0.346
family arrests
Disorderly conduct 9.964 1.875 0.061
arrests
Index offense -435.838 -8.669 0.000
reports***
Murder reports 0.444 1.952 0.051
Rape reports*** -2.993 -4.807 0.000
Robbery reports“ -2.384 -2.417 0.016
Assault reports -8.433 -1.6 0.11
Burglary reports* ** -55.339 -4.647 0.000
Motor vehicle theft -21.676 -7.872 0.000
reports***
Arson reports“ -1.53 -2.433 0.015

 

 

 

 

 

* Signiﬁcant in the expected direction at p < .05; "”" at p < .01; "* at p < .001.
’ Signiﬁcant in the unexpected direction at p. 05; H at p < .01; m at p < .001.

121

Two of the arrest variables have signiﬁcant mean differences that are not in the
expected direction. These results indicate that arrests for violent crimes and for
aggravated assault are signiﬁcantly higher in counties without slaughterhouses than in
those with slaughterhouses. These counterintuitive results, however, are not found
among the report rates and are not found in the subsequent analysis comparing the means
of counties with high levels of slaughterhouse employment with those with low levels or
no slaughterhouse employment.

The mean difference in reported crime between counties with and those without
slaughterhouses is statistically signiﬁcant in the expected direction for all except reports
of murder and assault reports. The results indicate that counties with slaughterhouses
experience on average approximately 3 more reports of rape, 2 more reports of robbery,
55 more burglary reports, 345 more larceny reports, 22 more motor vehicle theft reports,
and 1.5 arson reports than counties without slaughterhouses.

Clearly there is a difference in crime report rates between counties with and
without slaughterhouses, with counties with slaughterhouses reporting more crimes.
There is not as much evidence, however, of a difference in arrest rates: only arrests for
sex offenses are signiﬁcantly higher in the expected direction. This does not necessarily
contradict the literature on slaughterhouse communities, since those studies have
exclusively examined communities with high levels of slaughterhouse employment.
Accordingly, a subsequent analysis was undertaken to determine if there is a difference
between counties with high levels of slaughterhouse employment and those with little or

no employment in slaughterhouses.12

122

Examining the Effects of High Levels of Slaughterhouse Employment

The above results indicate that the only signiﬁcantly higher arrest rate in slaughterhouse
counties is for sex offenses, and that all of the crime reports, with the exception of murder
and assault, are signiﬁcantly greater in counties with slaughterhouses. Of the summary
crime variables, only the number of index offense reports is signiﬁcantly greater in
counties with slaughterhouses than in those without.

The results are notably different when we examine the difference between
counties with high levels of slaughterhouse employment (deﬁned here as 1000 or more
people employed in slaughterhouse facilities) and counties with less or no employment in
the industry (see Table 10). Three of the four summary arrest variables are signiﬁcant in
the expected direction: the total number of arrests, arrests for index offenses, and arrests
for property crime are signiﬁcantly higher in counties with 1000 or more slaughterhouse
employees. The mean difference in the total number of arrests is quite large, with
counties with high levels of slaughterhouse employment experiencing 737.996 more
arrests on average. The only summary arrest variable that is not signiﬁcantly greater is
arrests for violent crimes.

Of the speciﬁc arrest variables, the arrests for sex offenses variable is again
signiﬁcant, as are the arrests for rape, for other assaults, and disorderly conduct variables.
The mean number of arrests for offenses against the family is higher in counties with
high slaughterhouse employment; however, it does not meet the threshold to be

considered statistically signiﬁcant.

123

Table 10. Difference in Crime Rate Means between Counties with Slaughterhouse
Employment of 1000+ and Counties with 0-999 Slaughterhouse Employees

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Mean Difference t value p value
Total arrests" -737.996 -3.025 0.003
Index offense -92.135 -2.483 0.014
arrests*

Violent crime 4.077 0.572 0.568
arrests

Property crime -95.972 —3.01 8 0.003
arrestsM

Murder arrests 0.965 1.847 0.067
Rape arrests*" -4.014 -3.921 0.0001
Robbery arrests 0.531 0.529 0.598
Aggravated assault 6.247 1.066 0.289
arrests

Other assault -162.595 -4.616 0.000
arrests* * *

Sex offense -9.121 -3.043 0.003
arrests”

Offenses against -3.476 -0.538 0.592
the family arrests

Disorderly conduct -47.078 -3.025 0.003
arrests"

Index offense -973.96 -4.49 0.000
reports***

Murder reports 0.55 1.19 0.236
Rape reports*** —11.255 -4.72 0.000
Robbery reports -2.531 -1 .226 0.223
Assault reports 7.321 0.563 0.574
Burglary reports* —80.011 -2.135 0.035
Motor vehicle theft -37.438 -3.497 0.0007
@0115“ * *

Arson reports -2.386 -1 .314 0.191

 

 

 

 

 

* Signiﬁcant in the expected direction at p < .05; *" at p < .01; *" at p < .001.

Slightly fewer of the crime report variables are signiﬁcantly higher in counties
with high slaughterhouse employment than in the previous analysis testing the difference
between counties with and without slaughterhouses. The summary report variable (index

offense reports) is again signiﬁcant (p < .001), and the mean difference (-973.96) is more

124

than double what it was in the previous analysis. Further, reports of rape, burglary, and
motor vehicle theft are signiﬁcantly higher in counties with high slaughterhouse
employment. Reports of robbery and arson were signiﬁcant in the previous analysis but
are not signiﬁcant here (however, they are in the expected direction).

Overall, this analysis demonstrates that notable differences in arrest and report
rates exist between counties with one thousand or more slaughterhouse employees and
those with fewer or no slaughterhouse employees. Whether the differences in crime rates
between counties with and without slaughterhouses are related to employment in the
industry or instead can be accounted for by mediating variables cannot be assessed by
these analyses. To do so, we must turn to the multivariate techniques used in the next

chapter.

Summary

This chapter has reported the following central ﬁndings: (1) Several crime variables are
positively correlated with slaughterhouse employment and establishments. These
variables include total arrests, arrests for index offenses, property offenses, rape, other
assaults, sex offenses, and disorderly conduct, reports of index offenses, rape, robbery,
assault, burglary, motor vehicle theft, and arson; (2) Compared with counties without
slaughterhouses, counties with slaughterhouses have signiﬁcantly higher levels of arrests
for sex offenses, reports of index offenses, rape, robbery, burglary, motor vehicle theft,
and arson; (3) Counties with high levels of slaughterhouse employment (1000 or more

employees) when compared with counties with less or no employment in slaughterhouses

125

have signiﬁcantly higher levels of total arrests, arrests for index offenses, property crime,
rape, other assaults, sex offenses, disorderly conduct, reports of index offenses, rape,
burglary, and motor vehicle theft. These results therefore provide partial support for
research hypothesis 1 (that general and speciﬁc crime rates are higher in slaughterhouse

counties).

While these analyses are informative, they nonetheless are limited because they
do not control for several variables that are believed to mediate the relationship between
slaughterhouse employment and crime rates, do not entail comparisons with other similar
industries, and do not take advantage of the time dimension in the data. Accordingly, the
next chapter builds upon this one, examining models that incorporate control variables,

comparison industries, and the longitudinal dimension of the data.

126

Chapter 4

INTRODUCING CONTROL VARIABLES AND COMPARISON INDUSTRIES

All models are wrong. Some models are useful. — George E. P. Box

In this chapter the second and third research hypotheses are tested. Recall that these
hypotheses state that controlling for the variables proposed in the literature,
slaughterhouse presence and employment will be associated with increased crime rates in
counties, more so than the comparison industries, and that rape and family violence rates
in particular will increase in counties where there is an increase in slaughterhouse
employees. As discussed in Chapter 2, the variables of interest in this study could not
simply be analyzed using OLS regression due to the count nature of the crime data.
Instead, the following two techniques are employed in this study: (1) using factor
analysis to determine which crime variables meaningfully group together, scales are
created and used as dependent variables in OLS regression; (2) negative binomial
regression, which takes the count nature of the dependent variable into account, is used to
examine the individual crime variables. Both techniques demonstrate that
slaughterhouse employment has eﬂects on certain crime rates over and above what can
be explained by the control variables and that similar eﬂects are not found among the
comparison industries. In this chapter we ﬁrst examine the OLS regression results using
the factor analysis-derived scales to gain a better understanding of the effect of
slaughterhouse employment rates on crime in general. Then we examine the negative

binomial regression results which enable us to focus on speciﬁc crime variables.

127

Analyzing Slaughterhouses and Crime Rates Utilizing OLS Regression

Factor Analysis

Two factor analyses were performed with the aim of creating two scales: one for the
arrest variables and one for the report variables. These scales are discussed in turn

below. '3

(1) Arrest Rates

Principal components analyses were run, which reveal that the following arrest rate
variables load onto a single factor: rape, robbery, burglary, other assaults, forgery,
possessing stolen property, vandalism, offenses against the family, and disorderly
conduct.

Two criteria were used to decide how many factors to extract. The ﬁrst criteria is
Kaiser’s eigenvalue rule, whereby factors with eigenvalues greater than 1 are retained
because they explain more of the variance than the average amount explained by one of
the variables alone (DeVellis 1991; Henson and Roberts 2006). (The eigenvalues for the
factors are listed in Table 11 below). This method has been critiqued as being rather
arbitrary, since a value of 1.1 might be retained while a value of 0.99 could be discarded.
Therefore, it is advisable to use more than one criterion in deciding how many factors to

retain. Accordingly, a second criteria was used: Cattell’s scree test (DeVellis 1991).

128

According to this criteria the second factor and below are in the scree and should
therefore be excluded. Thus, by both criteria we are left with a one factor structure (and

as a result, rotation was not required).

Table 11. Eigenvalues for the Arrest Rate Factor Analysis

Component Eigenvalue Cumulative Proportion Explained
3.86846 0.4298
0.92478 0.5326
0.82554 0.6243
0.71329 0.7036
0.70711 0.7821
0.61769 0.8508
0.53658 0.9104
0.47939 0.9637
0.32713 1.0000

\OOONQMADJNt—i

 

 

Using Iterative Principal Factors factor loadings were obtained which indicate that
all of the variables load reasonably well onto the factor. It is suggested that factor
loadings be have a minimum value between 0.30 (Leary 1991) and 0.40 (Wagenaar 1981)
to be considered adequately loaded on to a factor. The table below illustrates that all of

the factor loadings in this case exceed that threshold.

Table 12. Factor Loadings for Arrest Variables

 

Variable Factor Loadiggs
Rape 0.41538
Robbery 0.64845
Burglary 0.70605
Forgery 0.66792
Stolen Property 0.51142
Vandalism 0.51847
Other assaults 0.83718
Offenses against the family 0.49265
Disorderly conduct 0.61896

 

 

129

Next, a scale was created using these variables. The scale was created in Stata as
follows: a score was created for every observation and the summative score was divided
by the number of items (in this case 9) used to create the scale. The internal consistency
of the scale is measured by Cronbach’s alpha, which has a value of 0.6728. This is
slightly less than the generally aimed for value of 0.7; however, for use as a dependent
variable that guideline is less critical than for independent variables since measurement
error in dependent variables are less of a problem for inference than measurement error in

independent variables.

(2) Report Rates

This same process was followed in creating a scale of the crime report variables. First,
all of the report variables were included in the factor analysis. (Please see note number
13 for a discussion of the removal of outliers). Two variables were subsequently
dropped: The murder reports variable was removed because it did not load well onto the
factor and larceny reports was dropped because it lowered the reliability of the scale.
Principal Components analysis revealed that the remaining crime report variables (rape,
robbery, assault, burglary, motor vehicle theft, and arson) load well on to one factor.

Using the same criteria used in the factor analysis of the arrest variables
(eigenvalues greater than 1 and the scree test), it was determined that one factor should be
extracted. (The eigenvalues are listed in Table 13.)

Using Iterative Principal Factors the factor loadings were obtained. All of the

loadings are above 0.40 and therefore load well onto the factor (see Table 14 below).

130

The internal consistency of this scale is slightly lower than the arrest variable scale, with

a Cronbach’s alpha value of 0.6062.

Table 13. Eigenvalues for the Report Rate Factor Analysis

 

Component Eigenvalue Cumulative Proportion Explained
1 3.28176 0.547
2 0.83931 0.6868
3 0.67591 0.7995
4 0.46533 0.877
5 0.43879 0.9502
6 0.29891 1.0000

 

 

Table 14. Factor Loadings for Report Variables

Variable Factor Loadings
Rape 0.54284
Robbery 0.68135
Assault 0.75140
Burglary 0.86041
Motor vehicle theft 0.77803
Arson 0.44815

 

 

Multiple Regression Analyses

In all of the regression models tested in this study the slaughterhouse and comparison
industry variables were lagged one year because their impact on crime would likely not
be felt in the same year in these counties. More likely, the impact would be felt the
following year (especially in cases where the industry opened or expanded late in the
year). Lags are used in panel studies, such as in studies of how residential racial

composition affects political party identiﬁcation (Giles and Hertz 1994) and how the

131

opening of casinos in counties impacts crime rates (Grinols and Mustard 2006), to model
dynamic effects. The number of variables lagged in a study varies. In some studies,
more than one variable is lagged, while in others only one is lagged while the others are
held contemporaneous with the dependent variable (Halaby 2004). For instance, in the
racial composition and political party identiﬁcation study cited above, the percent of the
population that is black was lagged one year, while other independent variables were not
(Giles and Hertz 1994). Similarly, in the study of the effect of casinos on crime rates
(Grinols and Mustard 2006), lead and lag variables were included for casino openings
and laws permitting carrying concealed ﬁrearms but not for control variables such as
population density, economic variables, and demographic variables. Following those
studies, the slaughterhouse and comparison industry variables are lagged in this study,
while the control variables are dated contemporaneously with the dependent variables.
The logic followed here is that employment may take some time to have an effect on
communities, whereas the effects of the other variables can be expected to be more
immediate.

After creating the two scales (described above) they were entered into OLS
regression models as dependent variables. The following three time-series cross-section
(TSC S) models were run with each of the scales in turn as the dependent variable (each
with ﬁxed effects): (1) With the number of slaughterhouse workers as the sole
independent variable; (2) with the control variables added; (3) with the comparison
industries added.14 The results with the arrest scale as the dependent variable are
presented in Tables 15 and the results with the report scale as the dependent variable are

found in Table 16.

132

Table 15. Multiple Regression with Arrest Scale as the Dependent Variable

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Independent Model 1 Coeff. Model 2 Coeff. Model 3 Coeff.
Variables (Std. Error) Std. Error) (Std. Error)
Slaughterhouse .019 (.004)*** .013 (.004)** .013 (.004)**
employment

Unemployment 1.17 (.346)"‘* 1.164 (.346)**
Number in poverty .0003 (.0007) .0003 (.0007)
Immigration .072 (.028)* .069 (.028)*
Migration .004 (.003) .003 (.003)
Number of non- .008 (.001)*** .008 (.001)***
whites and/or

Hispanics

Young males -.003 (.002) -.003 (.002)
Total number of -.009 (.002)*** -.009 (.002)***
males

Population Density -.563 (.257)* -.556 (.257)*
Iron and steel -.204 (.126)
forgigg

Truck trailer -.016 (.02)
manufacturing

Motor vehicle metal -.035 (.061)
stamping

Sign manufacturing -.011 (.013)
Industrial .086 (.062)
launderers

Model — F value 21.36*** 19.83*** 1972*"
R-squared 0.004 0.04 0.03

 

 

* Signiﬁcant at p < .05; ** Signiﬁcant at p < .01; *** Signiﬁcant at p < .001.

The ﬁrst thing to note about the regressions in Tables 15 and 16 is that the number of
slaughterhouse employees is a signiﬁcant predictor in all six models. In Model 1 with the
Arrest scale as the dependent variable the coefﬁcient of the slaughterhouse variable is
.019 (signiﬁcant at the p < .001 level) and with the addition of the control and other
industry variables in Models 2 and 3 the coefﬁcient is .013 (signiﬁcant at the p < .01
level). This means that controlling for all of the variables in the model, when the number
of slaughterhouse workers increases by 1 the arrest rate scale increases by .013 arrests.

At ﬁrst this may not seem like much, but this value reﬂects the average risk posed by one

133

slaughterhouse worker. The addition of an average sized slaughterhouse of 175 workers
would result in an increase in the arrest rate of 2.28 arrests (and many counties have more
than one slaughterhouse of this size). A larger slaughterhouse of one thousand
slaughterhouse workers would accordingly result in an increase of 13 arrests per 100,000
people for the crimes included in the scale (rape, robbery, burglary, other assaults,
forgery, possessing stolen property, vandalism, offenses against the family, and

disorderly conduct). The results are even more substantial when we examine the crime

Table 16. Multiple Regression with Report Scale as the Dependent Variable

1
Variables

(

2.027
.006 .001
.263
.014
non- . .012 (
whites and/or

Y males -.003 .003 -.003
T number -.019 (.003)*** -.019 (.003
males

.308 . .312 .
Iron steel -.363 (.24)

T .06 (.038)

Motor vehicle metal -.1 13 (.l 17)

-.018
.016 (.118)
launderers

— F 10.39***
.068

 

134

report scale instead of the arrest scale. Controlling for all of the variables, the coefﬁcient
for slaughterhouse employment is .027 (signiﬁcant at the p < .01 level). Accordingly,
one average sized slaughterhouse would result in an increase in the report rate of 4.73
crime reports and an increase of 1000 slaughterhouse workers would be associated with
an increase of 27 reports of the crimes included in the scale (rape, robbery, assault,
burglary, motor vehicle theft, and arson).

Some of the control variables also have signiﬁcant effects. Looking ﬁrst at the
results with the arrest scale as the dependent variable we see that unemployment is
signiﬁcant (p < .01). The value of the coefﬁcient is 1.2 (rounded) for models 2 and 3,
which means that for every one percent increase in the unemployment rate in these
counties, we can expect a 1.2 increase in the number of arrests. The other economic
variable, the number of people in poverty, is not signiﬁcant.

The net number of immigrants variable is also signiﬁcant (p < .05), with a
coefﬁcient value of .07 (rounded) for models 2 and 3, indicating that with each additional
immigrant into a county a .07 increase in the arrest rate scale can be expected. The other
population turnover variable, net migration, is not signiﬁcant.

Of the demographic variables, the number of non-whites and/or Hispanics and the
total number of males are signiﬁcant at the p < .001 level, while the number of young
males is not signiﬁcant. However, contrary to what is theorized in the literature, the total
number of males is negatively related to the crime rates (but with a value of .009, it is
very close to zero). Also contrary to the expectations in the literature, population density
has a signiﬁcant (p < .05) negative effect on the arrest scale. This result could be related

to the fact that only rural counties are included in this study; perhaps if urban areas had

135

been included we would have observed the expected positive effect of population density
on crime.

In the third model, all of the comparison industries are introduced. An
examination of the results indicates that none of the ﬁve industries have a signiﬁcant
effect, and except for one of the coefﬁcients (for Industrial Launderers), all are negative.

The R—square measures of model ﬁt are small for these models (ranging from .004
for model 1 and .04 for model 2). However, the F test for the Coefﬁcient of Multiple
Determination is signiﬁcant (with p < .001 in all three models). This means that we can
reject the null hypothesis that R-squared is 0 in the population based on the F test, but
substantively it is very close to zero. The small value of R-squared is not entirely
surprising given the relatively small number of independent variables in the model for
explaining something as multifaceted as crime, and the fact that many of the variables
hypothesized in the literature to be related to crime increases in slaughterhouse
communities (and included in this study as control variables) are derived from studies of
urban crime and might be less useful in predicting crime in rural areas. These small R-
squared values are generally consistent with other studies modeling the effects of
numerous variables on crime in rural counties: The r-squared values of Lee and Ousey’s
(2001) models ranged from 0.260 to 0.436 and the values of Osgood’s (2000) models
ranged from 0.226 to 0.585. Regardless, the purpose of this study is not to design
exhaustive models; rather, it is to examine whether slaughterhouses are associated with
increases in crime once other variables theorized in the literature as mediating the effect

are controlled for.

136

Turning to the results with the crime report scale as the dependent variable we
ﬁnd fairly similar results. The effect of slaughterhouse employment across the models is
also signiﬁcant, but the effect is somewhat stronger than that observed with the arrest
variable scale. Controlling for the other variables, an increase of one slaughterhouse
employee is expected to be associated with a .027 increase in the crime rate report scale.

The unemployment variable is also signiﬁcant (p < .01), but the effects are larger
than those observed for the arrest scale (coefficients of 2.035 and 2.027 for model 2 and 3
with the report scale as the dependent variable compared to 1.17 and 1.64 with the arrest
scale as the dependent variable). The other economic variable, the number of people in
poverty, is signiﬁcant in these models whereas it was not in the models with the arrest
rate scale as the dependent variable. However, the value of the coefﬁcient (.006) is fairly
small.

Net immigration is again signiﬁcant (p < .001), and has a larger coefﬁcient value
than it did for the arrest rate scale (.26 versus .07). Unlike the results with the arrest scale
as the dependent variable, with a coefﬁcient of .014 net migration is signiﬁcant in these
models (p < .01).

The effects of the demographic variables are similar. As with the arrest rate scale
as the dependent variable, the number of young males is not a signiﬁcant predictor of the
report rate scale; the total number of males has a signiﬁcant negative effect (p < .001)
which is slightly stronger (b = -.019) than for the arrest rate scale (b = -.009); and the
number of non-whites and Hispanics is signiﬁcant, but the coefﬁcient remains fairly
small (.012). Population density, however, is not signiﬁcant, whereas it was with the

arrest scale as the dependent variable. (It should be noted again, as discussed in endnote

137

number 14, that the high colinearity among some of these control variables might be
impacting their documented effects here.)

Like the arrest scale models, none of the effects of the comparison industries are
signiﬁcant. However, whereas only the coefﬁcient for Industrial Laborers was positive
with the arrest scale as the dependent variable, in the analysis with the report scale as the
dependent variable the coefﬁcient for Truck Trailer Manufacturing is also positive (but
we cannot reject the null hypotheses that they have no effect on the dependent variable).

The R-squared values for the models with all of the control variables are slightly
higher with the report scale as the dependent variable (model 2 and 3 = .068) as
compared with the arrest scale as the dependent variable (model 2 = .04 and model 3 =
.03). The F tests for the Coefﬁcient of Multiple Determination are also all signiﬁcant

(with p < .001 in all three models).

Estimating the Effects of Levels of Slaughterhouse Employment

Using these scales it is possible to estimate the effects of speciﬁed numbers of

slaughterhouse employees in counties. This is accomplished by ﬁxing all of the control

variables at their means and manipulating only the number of slaughterhouse employees

in a county, using the following equation for TSCS OLS regression with ﬁxed effects:

(gm —y,~+§) :a+(x,.—i.+§)ﬁ+(c.,—?.+Tz)+%f

(StataCorp 2005).

138

 

The results of the equation at different levels of slaughterhouse employment are
listed in Table 17. As the number of slaughterhouse employees increases so does the
number of expected arrests and reports. The mean number of employees per
slaughterhouse approximates 175. Thus, looking at the results for 175 slaughterhouse
employees we ﬁnd that in a typical county the addition of one slaughterhouse would be
expected to increase the arrest scale by 2.24 arrests and the report scale by 4.69 reports.
Particularly telling is the fact that the expected arrest and report values in counties with
7500 slaughterhouse employees are more than double the values when there are no
slaughterhouse employees. The increase in the expected arrests and reports associated

with the increase in slaughterhouse employment is clear in Figure 5.

Table 17. Results of TSCS OLS Equation at Varying Levels of Slaughterhouse
Employment, Keeping Control Variables Stable

 

 

 

 

 

 

 

 

 

 

 

Slaughterhouse Arrest Scale Report Scale
employment

0 employees 69.32 115.40
10 emploges 69.44 115.67
60 employees 70.09 117.01
175 emploges 71.56 120.09
375 employees 74.13 125.45
750 employees 78.94 135.50
1750 emplonges 91.78 162.30
3750 employees 117.45 215.90
7500 employees 165.59 316.39

 

 

139

 

Figure 5. Linear Prediction Equation Values for Arrest and Report Scales

 

   

 

 

 

 

250
200
Large slaughterhouse
Average size
‘2 150 slaughterhouse
r:
1: __,
g +Arrest Scale
g + Report Scale
’ ,
2
< 100
50
o

O 10 60 175 375 750 1750 3750
Number oi Slaughterhouse Employees

The above analyses using linear OLS longitudinal regression demonstrate the
effect of slaughterhouse employment, employment in other industries, and the control
variables on scales comprised of various arrest and report rates which factor analysis
demonstrated load meaningfully onto one factor. They establish that the effect of
slaughterhouse employment on these scales cannot be explained away by the control
variables, and further that the comparison industries do not have similar signiﬁcant
effects. Because the analyses reported herein employ ﬁxed effects, they also therefore
control for time-invariant variables in these counties that might impact the crime rates,
such as climate and history. Despite these ﬁndings, however, these analyses do not

afford insight into the effects of slaughterhouses, the comparison industries, and the

140

control variables on speciﬁc crime variables that are of particular interest in this study,
such as rape. As previously discussed, due to the discrete count nature of the crime data,
analyses of individual crime variables require negative binomial regression. This method
is generally more conservative than OLS or Poisson regression (Allison and Waterman
2002; Osgood 2000).15 The following section presents the results of the TSCS negative

binomial regression analyses of the data.

Analyzing Slaughterhouses and Crime Rates Utilizing Negative Binomial Regression

Time-series cross-section (TSCS) negative binomial regression was performed on
the data for two time periods: the period before ‘custom slaughter’ facilities were added
to the slaughterhouse (except poultry) category (1994-1997) and then for the entire time
period under study (1994 to 2002). The models included the following seven arrest and
four report variables as the dependent variables: (1) total number of arrests; (2) arrests
for violent crimes; (3) arrests for murder; (4) arrests for rape; (5) arrests for offenses
against the family; (6) arrests for sex offenses (excluding rape); (7) arrests for aggravated
assault; (8) total number of reports for index offenses; (9) reports of murder; (10) reports
of rape; and (11) reports of assault.

These analyses were modeled with population set as the exposure variable and
county fixed effects, as discussed in Chapter 2. Three models were run for each of the
dependent variables, as was done with the OLS regression. In the ﬁrst model the only
independent variable included was the number of slaughterhouse employees (lagged one

year in all of the analyses). In the second model the control variables hypothesized in the

141

literature were added. The comparison industries (also lagged one year) were added to
the ﬁnal model.

As discussed earlier, in 1998 the classiﬁcation of slaughterhouse facilities was
changed to include “custom slaughter” facilities. And as was illustrated in Table 2 in
Chapter 2 comparing the slaughterhouse establishment ﬁgures from 1997 to 1998, this
change in categorization resulted in a large increase in the number of small facilities.
This likely has a twofold effect: (1) At a conceptual level, it results in the combining of
two very distinct types of slaughterhouses: industrial slaughterhouses that employ many
people and are motivated by proﬁts, and small facilities where much fewer animals are
being slaughtered, and those slaughtered are being used for personal consumption. It is
possible that the custom slaughter facilities are located in more viable communities, with
higher levels of social capital. The effects of large industrial slaughterhouses and smaller
custom slaughtering facilities on the community are likely very different. (2) Due to the
way the County Business Patterns groups the number of employees (Please see Appendix
B for full details) the inclusion of the small custom slaughterhouse facilities likely
artiﬁcially increases the number of slaughterhouse workers. For instance, a custom
slaughterhouse facility operated by two people would be classiﬁed as employing 1-19
people, and this category would then be assigned a mid-point value of 10. Therefore, in
this instance, there is an over-estimate of 8 slaughterhouse workers. This may not seem
like much, but given that there was an increase of 1,475 slaughterhouse facilities
employing 1-19 people from 1997 to 1998 it could result in substantial blurring of the
relationship between slaughterhouses and crime by artiﬁcially reducing and inverting the

relationship between the two. Therefore, the results from the period before ‘custom

142

slaughter’ facilities were added (1994-1997) are examined in detail below, with the
results of the three models summarized in separate tables for each of the dependent
variables. Then these results are compared to results derived from an analysis of the
entire time period (1994-2002).

In the tables included below an incidence-rate ratio (IR) is reported for each
predictor variable. The IR values are included because they are more easily interpreted
than the slope coefﬁcients, as the IRR value represents em rather than b. The IR values
can be interpreted as the multiplicative factor by which a one unit change in the
independent variable affects the dependent variable, controlling for the other variables.
Therefore, an IR value below 1 indicates that the predictor variable (controlling for the
other variables) decreases the incidence-rate, which demonstrates a negative effect, and
an IR value greater than 1 indicates an increase in the incidence-rate, or a positive
effect. They can also be expressed as percentages using the following formula:
100[exp(b)-l] (Long and Freese 2006; Long 1997). The standard errors of the estimates
are included in parentheses after the IRR values. Note that R-squared “is not part of the
maximum-likelihood estimation of the basic Poisson and negative binomial models”
(Osgood 2000, 38) and therefore is not reported here. However, chi-square tests of the
model ﬁt are reported.

The results reported below indicate that analyzing data from the restricted period
(1994-1997) and controlling for the variables proposed in the literature, slaughterhouse
employment has a signiﬁcant positive eﬂect on the total number of arrests and arrests for

violent crimes. A subsequent analysis of the entire time period reveals that

143

slaughterhouse employment also has a significant positive effect on arrests for rape and

arrests for sex offences (an effect not paralleled by the comparison industries).

(1) Total number of arrests. The table below lists the results of the TSCS negative
binomial regression for the restricted time period (1994-1997) with the total number of
arrests in the counties as the dependent variable. The results indicate that in the ﬁnal
model with all of the control variables and industry variables included, slaughterhouse
employment has a signiﬁcant positive effect on the total arrest rate (the fact that the IRR
value is above 1 indicates that the predictor variable increases the incidence-rate, which
demonstrates a positive effect). We can interpret the coefﬁcient to mean that each
additional slaughterhouse employee increases the expected total arrest rate by a factor of
1.000454 or by .0454%. This may not seem like much, but given that some of the large
slaughterhouse facilities employ thousands of people, the effect could be more substantial
than it appears. For instance, 4,000 slaughterhouse workers would increase the expected
total number of arrests by 1.82% [.000454 x 4000 = 1.82].

Of the control variables in the full model, the number of people in poverty,
migration, and the number of non-whites or Hispanics have signiﬁcant positive effects on
the arrest rate. Unemployment, the number of young males, and the total number of
males have signiﬁcant negative effects on the arrest rate. Of the comparison industries,
the only signiﬁcant effect is a negative one by truck trailer manufacturing. Models 2 and

3 signiﬁcantly predict the dependent variable.

144

Table 18. Negative Binomial Regression with Total Arrests as the Dependent
Variable, 1994-1997

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

value

 

 

 

 

Independent Model 1 [RR7 (Std. Model 2 IRRI (Std. Model 3 IRRI (Std.

Variables Error) Error) Error)

Slaughterhouse 1.000042 (.00008) 1.000071 (.00007) 1.000454

employment (00015)"

Unemployment .9967286 (.00839) .9267538
(.00822)***

Number iyoverty .9999924 (.00001) 1.000018 (.00001 *

Immigration 1.001018 (@047? .9990475 (.00082)

Migration .9999158 (.00004)* 1.000243
(.00003)***

Number of non- .999941 1.000464

whites and/or (.00001)*** (.00003)***

Hispanics

Young males 1.000192 (.00008)* .9994193
(.00009)***

Total number of .9999338 (.00003)* .9993236

males (.00004)***

Population Density .9956289 (.00224) 1.073137

Iron and steel .9968899 (.00368)

forging

Truck trailer .9979979

manufacturing (.00034)***

Motor vehicle .9992525 (.00117)

metal stamping

Sign manufacturing 1.003135 (.00186)

Industrial 1.006312 (.00497)

launderers

Model — Chi square 0.25 97.09*** 93.63***

 

* Signiﬁcant at p < .05; ** Signiﬁcant at p < .01; *** Signiﬁcant at p < .001. ‘ Incidence-Rate Ratio (IR)

145

 

(2) Arrests for violent crimes. The results with arrests for violent crimes as the dependent
variable are similar to those with total arrests: Once the other variables and industries are
controlled for, slaughterhouse employment has a signiﬁcant positive effect on the violent
arrest rate. We can interpret the coefﬁcient to mean that each additional slaughterhouse

employee increases the expected violent arrest rate by a factor of 1.000221 or by .0221%.

Accordingly, 4,000 slaughterhouse workers would be expected to increase the number of

arrests for violent offenses by 0.88% [00022] x 4000 = 0.884].

Table 19. Negative Binomial Regression with Arrests for Violent Crimes as the
Dependent Variable, 1994-1997

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

value

 

 

 

Independent Model 1 IRRI (Std. Model 2 IRRj (Std. Model 3 IRRT (Std.
Variables Error) Error) Error)
Slaughterhouse .9999964 (.00015) 1.000222 (.00011)* 1.000221 (.00011)*
employment

Unemployment 1.000146 (.01328L .9981239 (.01315)
Number in poverty .9999899 (.00002) .9999896 (.00001)
Immigration .9969369 (.0014)* .9968228 (.00141)*
Migration .9999816 (.00007) .9999918 (.00007)
Number of non- 1.000026 (.00002) 1.000038 (.00002)*
whites and/or

Hispanics

Young males 1.000162 (.00012) 1.000081 (.00012)
Total number of .9998548 .9998762

males (.00004)*** (.00004)"
Population Density .9987652 (.00241) .9988433 (.0024)
Iron and steel .989946 (.01972)
forging

Truck trailer .9980742 (.0006)**
manufacturing

Motor vehicle .9946924 (.00244)*
metal stamping

Sign manufacturing 1.000377 (.00142)
Industrial 1.00489 (.00423)
launderers

Model — Chi square <0.001 89.61*** 106.04***

 

* Signiﬁcant at p < .05; " Signiﬁcant at p < .01; *** Signiﬁcant at p < .001.
' Incidence-Rate Ratio (IR)

146

 

Compared with the analyses with the total arrest rate as the dependent variable,
however, fewer of the control variables have signiﬁcant effects here. Only the number of
non-whites/Hispanics has a signiﬁcant positive (but fairly small) effect. Immigration and
the total number of males have signiﬁcant negative effects. Of the comparison industries,
only two have signiﬁcant effects (Truck Trailer Manufacturing and Motor Vehicle
Manufacturing), but these effects are negative. Again, only Model 2 and 3 have

signiﬁcant Chi-square values.

(3) Arrests for murder. With arrests for murder as the dependent variable, slaughterhouse
employment is not signiﬁcant in any of the three models, although the relationship is in
the expected (positive) direction in all three models. Only one of the control variables
(number of people in poverty) has a signiﬁcant effect, but it is a negative effect.
Additionally, the chi-square values for all of the models are not signiﬁcant. None of the
comparison industries have signiﬁcant effects either. Clearly the explanatory power of
these models with murder as the dependent variable are less than for the total arrests and
arrests for violent offenses variables. This is not surprising since in these rural counties

with small crime counts, instances of murder would be more random than other crimes.

147

Table 20. Negative Binomial Regression with Arrests for Murder as the Dependent
Variable, 1994-1997

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

value

 

 

 

Independent Model 1 IRR‘ (Std. Mode12 IRR‘ (Std. Mode13 IRR‘ (Std.
Variables Error) Error) Error)
Slaughterhouse 1.000394 (.00082) 1.000765 (.00088) 1.000869 (.00094)
employment

Unemployment 1.033133 (.03982) 1.040852 (.04064)
Number in poverty .9999223 (.00004)* .9999166 (.00003)*
Immigration .9948729 (.00282) .9953541 (.00302)
Migration 1.000184 (.00018) 1.000304 (.00019)
Number of non- 1.000023 (.00004) 1.000049 (.00005)
whites and/or

Hispanics

Young males .9998493 (.00024) .9996847 (.00025)
Total number of 1.000061 (.0001) 1.000111 (.0001)
males

Population Density .9954816 (.01172) .9972847 (.01236)
Iron and steel 1.031706 (.05151)
forging

Truck trailer .9979514 (.00119)
manufacturing

Motor vehicle .9985286 (.00655)
metal stamping

Sign manufacturing 1.004712 (.00899)
Industrial .9862215 (.00765)
launderers

Model — Chi square 0.23 15.03 23.27

 

* Signiﬁcant at p < .05; ** Signiﬁcant at p < .01; *** Signiﬁcant at p < .001.
' Incidence-Rate Ratio (IR)

148

 

(4) Arrests for rape. With rape as the dependent variable, the slaughterhouse

employment variable does not have a signiﬁcant effect (although the effect is in the

expected direction). Of the control variables, only population density has a signiﬁcant

(positive) effect. The effect is signiﬁcant in Model 2, but loses signiﬁcance in Model 3.

Additionally, none of the models have signiﬁcant chi-square values. Of the comparison

industries, only Iron and Steel Forging has a signiﬁcant effect, but it is a negative one.

Table 21. Negative Binomial Regression with Arrests for Rape as the Dependent
Variable, 1994-1997

 

 

 

 

 

 

 

Independent Model 1 IRRI (Std. Model 2 IRR‘ (Std. Mode13 IRR‘ (Std.
Variables Error) Error) Error)
Slaughterhouse 1.000066 (.00022) 1.000051 (.00024) 1.000059 (.00024)
employment

Unemployment 1.026603 (.03168) 1.024633 (.03184)
Number in poverty 1 (.00002) .9999978 (.00002)
Immigration .9982699 (.002) .9979653 (.00203)
Migration .9999886 (.0001) .9999951 (.00013)
Number of non- 1.000023 (.00007) 1.00001 (.0000?)
whites and/or

Hispanics

 

Young males

1.00002 (.0003)

1.00012 (.00033)

 

Total number of
males

.9999141 (.00011)

.9998887 (.00012)

 

Population Density

1.035381 way

1.036529 (.02165)

 

 

 

 

 

 

 

 

value

 

 

 

Iron and steel .9604903 (.01928)*
fogging

Truck trailer .9997399 (.00083)
manufacturing

Motor vehicle .9992979 (.00886)
metal stamping

Sign manufacturing 1.002019 (.00315)
Industrial .9974364 (.013)
launderers

Model — Chi square 0.09 9.33 14.39

 

* Signiﬁcant at p < .05; " Signiﬁcant at p < .01; *** Signiﬁcant at p < .001.
' Incidence-Rate Ratio (IR)

149

 

(5) Arrests for offenses against the family. With offenses against the family as the

dependent variable the effect of slaughterhouse employment is positive in all models but

is not signiﬁcant. Of the control variables, in both Models 2 and 3, immigration has a

signiﬁcant positive effect and the total number of males has a signiﬁcant negative effect.

Models 2 and 3 signiﬁcantly predict the dependent variable. Finally, none of the

comparison industries have signiﬁcant effects.

Table 22. Negative Binomial Regression with Arrests for Offenses against the
Family as the Dependent Variable, 1994-1997

 

 

 

 

 

 

 

 

 

Independent Model 1 IRRI (Std. Model 2 IRRI (Std. Model 3 IRRI (Std.
Variables Error) Error) Error)
Slaughterhouse 1.000203 (.00017) 1.000201 (.00016) 1.000222 (.00016)
employment
Unemployment .9703695 (.01974) .97006 (.0196)
Number in poverty .9999702 (.00003) .9999625 (.00003)
Immigration 1 .004046 1 .003 668
£00139)" (.00139 **
Migration .9998966 (.00013) .9998443 (.00014)
Number of non- 1.000001 (.00002) 1.000015 (.00007)
whites and/or
Hispanics
Young males 1.000055 (.00012) 1.000025 (.00012)
Total number of .9998909 (.00005)* .9998898 (.00005)*
males

 

Population Density

1.002977 (.00361)

1.002707 (.00347)

 

Iron and steel

1.024801 (.02695)

 

 

 

 

forging

Truck trailer .9983923 (.00128)
manufacturing

Motor vehicle .9985946 (.00278)
metal stamping

Sign manufacturing .9996684 (.00293)
Industrial 1.011544 (.00598)
launderers

 

 

Model — Chi square
value

 

 

1 .44

81.08***

 

 

8746*"

 

* Signiﬁcant at p < .05; ** Signiﬁcant at p < .01; *** Signiﬁcant at p < .001.
' Incidence-Rate Ratio (RR)

150

 

(6) Arrests for sex offenses (excluding rape). The effect of slaughterhouse employment
on arrests for sex offenses becomes positive in models 2 and 3, however, it does not reach
the threshold of signiﬁcance. None of the comparison industries have a signiﬁcant effect

on the dependent variable. Of the control variables, only the number of people in poverty

has a signiﬁcant effect, which is negative. Models 2 and 3 signiﬁcantly predict the

dependent variable.

Table 23. Negative Binomial Regression with Arrests for Sex Offenses as the
Dependent Variable, 1994-1997

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

value

 

 

 

 

Independent Model 1 IRR‘ (Std. Model 2 IRR‘ (Std. Model 3 IRR‘ (Std.
Variables Error) Error) Error)
Slaughterhouse .9999815 (.00025) 1.000055 (.00024) 1.000083 (.00023)
employment
Unemployment 1.026606 (.02897) 1.03124 (.0291 5)
Number in poverty .9999057 .9998982
(00003)“ 900003?"

Immigration 1.000403 (.00176) 1.000434 (.00178)
Migration .9999971 (.00013) .9999715 (.00013)
Number of non- .9999967 (.00003) 1.000006 (.00003)
whites and/or
Hispanics
Young males .9998339 (.00014) .9998274 (.00014)
Total number of .9999934 (.00006) .9999842 (.00006)
males
Population Density 1.008042 (.00677) 1.00699 (.0065)
Iron and steel 1.001263 (.02127)
forging
Truck trailer 1.000406 (.00094)
manufacturing
Motor vehicle .9997529 (.00565)
metal stamping

iigp manufacturing 1.003013 (.00206)
Industrial 1.009035 (.00618)
launderers
Model — Chi square 0.01 50.36*** 5842*M

 

" Signiﬁcant at p < .05; ** Signiﬁcant at p < .01; *** Signiﬁcant at p < .001.
' Incidence-Rate Ratio (IRR)

151

 

( 7) Arrests for aggravated assault. Slaughterhouse employment does not have a

signiﬁcant effect in any of the models, however, its effect is in the expected direction in

Models 2 and 3. Two of the control variables, immigration and the total number of

males, have signiﬁcant negative effects. And of the comparison industries, Truck Trailer

Manufacturing has a signiﬁcant negative effect. Models 2 and 3 signiﬁcantly predict the

dependent variable.

Table 24. Negative Binomial Regression with Arrests for Aggravated Assault as the
Dependent Variable, 1994-1997

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

value

 

Independent Model 1 IR] (Std. Model 2 IRRI (Std. Model 3 IRR‘ (Std.
Variables Error) Error) Error)
Slaughterhouse .9999752 (.00015) 1.000203 (.00012) 1.000196 (.00012)
employment

Unemployment 1.006588 (.01449) 1.004418 (.01437)
Number in poverty .9999917 (.00002) .9999922 (.00006)
Immigration .9965188 (.00143)* .9965581 (.00143)*
Migration .9999593 (.00008) .9999702 (.00008)
Number of non- 1.000026 (.00002) 1.000036 (.00002)
whites and/or

Hispanics

Young males 1.00022 (.00012) 1.000141 (.00012)
Total number of .9998284 .9998497

males (.00004)*** (.00004)***
Population Density 1.000063 (.00263) 1.000122 (.00263)
Iron and steel .9894973 (.01884)
forging

Truck trailer .997983 (.00065)**
manufacturing

Motor vehicle .9953782 (.00254)
metal stamping

Sign manufacturing .999614 (.00142)
Industrial 1 .00595 (.00421)
launderers

Model — Chi square 0.01 90.68*** 104.99***

 

 

 

 

* Signiﬁcant at p < .05; ** Signiﬁcant at p < .01; *** Signiﬁcant at p < .001.
lIncidence-Rate Ratio (IRR)

152

(8) Total number of reports for index offenses. Slaughterhouse employment has a
signiﬁcant negative effect in the ﬁrst model, but once the other variables are controlled
for in the subsequent models the effect becomes insigniﬁcant. More of the control
variables are signiﬁcant in these models with the number of reports for index offenses as
the dependent variable than in the earlier analyses. Unemployment and population
density have signiﬁcant negative effects in both Models 2 and 3. The migration, number
of non-white and/or Hispanics, and total number of males variables have signiﬁcant
negative effects in Model 3. The immigration variable has a signiﬁcant positive effect in
both Models 2 and 3 and the number of young males has a signiﬁcant positive effect in
Model 3. All three of the models signiﬁcantly predict the dependent variable. Of the
comparison industries, two (Motor Vehicle Metal Stamping and Sign Manufacturing)

have signiﬁcant negative effects.

153

Table 25. Negative Binomial Regression with Reports of Index Offenses as the
Dependent Variable, 1994-1997

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

value

 

 

 

Independent Model 1 IRRI (Std. Model 2 IRR‘ (Std. Model 3 IRR‘ (Std.

Variables Error) Error) Error)

Slaughterhouse .9994702 (.0002)** .9997627 (.00017) .9997331 (.00017)

employment

Unemployment .9391391 .9448218

(.01484)*** (013443"
Number in poverty 1.000016 (.00002) 1.000016 (.00001)
Immigration 1 .003632 1 .002738
(.00065)*** (.00044)***

Migration 1.000044 (.00007) .9999195
(.00003)**

Number of non- .9999917 (.00001) .9999082

whites and/or (.00002)***

Hispanics

Young males .9999872 (.00011) 1.000807
(.0001)***

Total number of .9999873 (.00005) .9997796

males (.00004)***

Population Density .9865924 (.0042)“ .9891606
(00342)“

Iron and steel 1.005325 (.004)

forging

Truck trailer 1.000048 (.0005)

manufacturing

Motor vehicle .9941526

metal stamping (.00191)"

Sign manufacturing .9896203
(.00153)***

Industrial 1.005605 (.00315)

launderers

Model — Chi square 6.83M 105.25*** 286.71 ***

 

‘1' Signiﬁcant at p < .05; ** Signiﬁcant at p < .01; *** Signiﬁcant at p < .001.
Incidence-Rate Ratio (IR)

(9) Reports of murder. The effect of slaughterhouse employment on reports of murder is

positive but not signiﬁcant in all three of the models. Additionally, none of the control

variables or the comparison industry variables exhibit signiﬁcant effects. Accordingly, as

154

 

evidenced by the chi-square values, none of the three models signiﬁcantly explain the
dependent variable.

Table 26. Negative Binomial Regression with Reports of Murder as the Dependent
Variable, 1994-2002

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Independent Model 1 IRR‘ (Std. Model 2 IR? (Std. Model 3 IRR‘ (Std.
Variables Error) Error) Error)
Slaughterhouse 1.001454 (.00078) 1.001363 (.00077) 1.001345 (.00078)
employment

Unemployment .9864899 (.04924) .9880815 (.05006)
Number in poverty 1.000025 (.00003) 1.000019 (.00003)
Immigration 1.000817 (.0029) 1.000711 (.00293)
Migration .9999473 (.00017) 1.000048 (.00018)
Number of non- 1.000062 (.00006) 1.00006 (.00006)
whites and/or

Hispanics

Young males .9997503 (.00031) .9997969 (.0004)
Total number of 1.000047 (.00013) 1.000049 (.00014)
males

Population Density .9929716 (.01196) .9937794 (.01259)
Iron and steel 1.037114 (.10903)
forging

Truck trailer .9995342 (.00085)
manufacturing

Motor vehicle 1.000061 (.00753)
metal stamping

Sign manufacturing .9935767 (.00671)
Industrial .993868 (.01104)
launderers

Model — Chi square 3.49 7.06 8.1 1

value

 

 

 

 

 

 

"‘ Signiﬁcant at p < .05; ** Signiﬁcant at p < .01; *** Signiﬁcant at p < .001.
' Incidence-Rate Ratio (IR)

(10) Reports of rape. The slaughterhouse employment variable has a positive effect on
the reports of rape variable in all three models, but it does not meet the threshold of
signiﬁcance. Of the control variables, the poverty variable has a signiﬁcant positive
effect in Model 2 and the migration variable has a signiﬁcant negative effect in Model 3.

The chi-square statistic demonstrates that Model 2 and Model 3 signiﬁcantly predict the

155

dependent variable. Further, two of the comparison industries have signiﬁcant effects.
The Motor Vehicle Metal Stamping Variable has a signiﬁcant positive effect on reports
of rape, controlling for the other variables, whereas the Sign Manufacturing variable has

a signiﬁcant negative effect.

Table 27. Negative Binomial Regression with Reports of Rape as the Dependent
Variable, 1994-1997

 

 

 

 

 

 

Independent Model 1 IRR‘ (Std. Model 2 IRR‘ (Std. Model 3 IRR‘ (Std.

Variables Error) Error) Error)

Slaughterhouse 1.000047 (.00021) 1.000044 (.00021) 1.000008 (.00021)

employment

Unemployment 1.00616 (.02882) .9971552 (.02901)

Number in poverty 1.000042 1.000032 (.00002)
(.00002)**

Immigration 1.001101 (.00125) 1.000913 (.00123)

Migration .9998491 (.00008) .9997468

(00008)“

 

Number of non-
whites and/or

1.000069 (.00004)

1.000065 (.00004)

 

 

 

 

 

 

 

 

 

 

 

Hispanics

Youngmales 1.000086 (.0002) 1.000182 (.00022)

Total number of .9999367 (.00007) .9999179 (.00007)

males

Population Density .9963716 (.00346) .9966124 (.00347)

Iron and steel .9883354 (.02219)

forging

Truck trailer .9996249 (.00042)

manufacturing

Motor vehicle metal 1.072275

stamping (.03061)*

Sign manufacturing .9949149
(00162)"

Industrial 1.005044 (.00708)

launderers

Model - Chi square 0.05 18.69* 4232*”

value

 

 

 

 

* Signiﬁcant at p < .05; ** Signiﬁcant at p < .01; *** Signiﬁcant at p < .001.

156

 

(11) Reports of assault. Slaughterhouse employment has a positive effect on reports of
assault in Models 1 and 2 and a negative effect in Model 3, but none of these effects are
signiﬁcant. However, several of the control variables are signiﬁcant. In Models 2 and 3,
unemployment, migration, the number of non-whites and/or Hispanics, and the total
number of males have signiﬁcant negative effects. The number of people in poverty and

Table 28. Negative Binomial Regression with Reports of Assault as the Dependent
Variable, 1994-1997

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Independent Model 1 IRR' (Std. Model 2 IRR‘ (Std. Model 3 IRRI (Std.

Variables Error) Error) Error)

Slaughterhouse 1.000134 (.00013) 1.000031 (.00015) .9999967 (.00015)

employment

Unemployment .9495 l 12 .94903 92
(01518)” (.0152)**

Number in poverty 1.000034 (.00002)* 1.000039

(.00002)*

Immigration 1.000824 (.00077) 1.000715 (.0008)

Migration .9998972 (.00005)* .999887 (.00005)*

Number of non- .9999508 .9999471

whites and/or (.00002)** (.00002)"

Hispanics

Young males 1.000557 1.00056
(.00012)*** (.00012)***

Total number of .9998001 .9997982

males L00004)*** (.00004)***

Population Density 1.000559 (.00232) 1.000516 (.00231)

Iron and steel .9877602 (.0168)

forging

Truck trailer .9987897 (.00062)

manufacturing

Motor vehicle 1.00843 (.00597)

metal stamping

Sign manufacturing .9920695

(.00175)***

Industrial 1 .004502

launderers (.003912)

Model — Chi square 1.01 54.5*** 78.91 ***

value

 

* Signiﬁcant at p < .05; ** Signiﬁcant at p < .01; *** Signiﬁcant at p < .001.
' Incidence-Rate Ratio (IRR)

157

 

the number of young males have signiﬁcant positive effects on reports of assaults. Both
Models 2 and 3 signiﬁcantly predict the dependent variable. Finally, only one of the
comparison industries has a signiﬁcant effect: Sign Manufacturing has a signiﬁcant

negative effect on reports of assaults.

Overall Effects of the Predictor Variables on the Crime Variables

The series of tables above provide an understanding of the effects of the predictor
variables on each dependent crime variable across the three models. However, what is
also needed is a sense of the performance of each variable across all of the 11 crime
variables analyzed. This overview is provided in Table 29. All of the predictor variables
are listed in the rows and the dependent variables are listed across the columns. Each cell
contains the IRR value for Model 3 (the most complete model).

We see that controlling for all of the variables in the model, the slaughterhouse
employment variable has a signiﬁcant positive effect on two variables: total arrests and
total arrests for violent offenses. The effects of Slaughterhouse employment on the rest of
the dependent variables are not signiﬁcant. However, it is worth noting that the effect of
slaughterhouse employment is in the expected direction (positive) for seven of the nine
individual variables. Additionally, these individual crime variables are very stochastic in
small counties. Therefore, it is not necessarily surprising that the effects of
slaughterhouse employment on these individual variables is not signiﬁcant, whereas we

ﬁnd signiﬁcant effects on the summary variables here (total arrests and violent arrests)

158

and the report and arrest scales in the previous analyses. This is because the summary
variables and the scales created in this study capture more instances of arrests and reports
and therefore reduce the randomness of those measures.

The unemployment variable did not have a signiﬁcant positive effect on any of
the crime variables, contrary to what has been theorized in the slaughterhouse literature.
In fact, it had a signiﬁcant negative effect on three of the crime variables (total arrests,
reports of index offenses, and reports of assault). The other economic variable, the
number of people in poverty, fared somewhat better. It had a positive effect on the
reports of assaults and on the total arrest rate (although the effect was less than the effect
of slaughterhouse employment). The variable also had a signiﬁcant negative effect on
two variables (murder arrests and arrests for sex offenses).

The results of the demographic variables are mixed. The number of non-white
and/or Hispanic residents had a signiﬁcant positive effect on the total number of arrests
and arrests for violent offenses. It also had a signiﬁcant negative effect on reports of
index offenses and assault reports. The number of young males had a signiﬁcant positive
effect on the two variables that the number of non-white and/or Hispanic residents
variable had a signiﬁcant negative effect on: reports of index offenses and reports of
assault. The number of young males also had a signiﬁcant negative effect on the total
arrest rate. The total number of males variable did not perform in a way consistent with
the theorizing in the literature. Not only did it not have a signiﬁcant positive effect on
any of the crime variables, it had a signiﬁcant negative effect on six of the variables.

The results of the other social disorganization variables were also mixed.

Population density only had one signiﬁcant effect (on reports of index offenses) and it

159

was a negative effect, contrary to the theorizing in the literature. The immigration
variable had a signiﬁcant positive effect on only two variables: arrests for offenses
against the family and reports of index offenses. It also had signiﬁcant negative effects
on two variables (violent arrests and arrests for assault). The migration variable had a
Signiﬁcant positive effect on the total number of arrests, but had a signiﬁcant negative
effect on three other variables (reports of index offenses, reports of rape, and reports of
assault).

There was only one case where one of the comparison industries had a signiﬁcant
effect on a crime variable: Motor Vehicle Metal Stamping had a signiﬁcant positive
effect on reports of rape. However, this same variable also had signiﬁcant negative
effects on two variables (violent arrests and index reports). Iron and Steel Forging had a
signiﬁcant negative effect on rape arrests; Truck Trailer manufacturing had. a signiﬁcant
negative effect on total arrests, violent arrests, and assault arrests; and Sign
Manufacturing had signiﬁcant negative effects on index reports, rape reports, and assault
reports. The Industrial Launderer variable did not have a signiﬁcant effect on any of the
crime variables. Therefore, although slaughterhouse employment did not perform as
expected with all of the crime variables (akin to the control variables), the comparison
industry variables demonstrated several negative relationships and only one significant

positive one.
Comparing the Results ﬁom the Restricted Time Period with the Entire Time Period

The previous analyses demonstrate that controlling for the variables proposed in the

literature and using the data from 1994-1997 (before ‘custom slaughter’ facilities were

160

 

 

 

 
 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

._8. v a a 28:55 5 :o. v a a 28%? t ”8. v a a 28595 .
eeougﬂt—
Nomvoo; vvomoo; wowm00. moonoo; m0noo._ mmo0oo._ 31m— _o._ 33‘500. n_NNow0. 0wvoo._ N- meoo; .3596:—
:. .. 5 a
m0ooN00. 0v _ 0v00. 000200. momo0w0. v _ 0000. m _ omoo._ vwoo000. 0 _ oNoo._ N _ h¢oo._ bhmooo; mm _ moo; =w_m
e t. a E ESw 33$—
mvwoo._ thNho; _ooooo._ ONW _ v00. wamm00. 0Nw0000. ov0mw00. 050N000. owmnw00. VNOOV00. mmmm000. o_o_:o> .5qu
a... .1. a: miss—£3.82
003000. 0v~o000. NVMm000. wvoooo; mw0h0 . oovooo; mm0mw00. 00mn000. :n0h00. Nitow00. 0000000. 3:30. goat.
. MEN—on—
Noohbw0. memww0. 3 Kmo; mmmmoo; mn0v0w0. mom—oo._ Swvmo; mo0voo0. 0o: no; ev00w0. 00wwo00. 30% can :9:
e» .9950
S mooo; 120000. v05. m00. coo _ 000. mm _ ooo._ 00ooo._ hohNoo; 0Nn©mo._ hvah00. mmvww00. 0m _ who; coon—=00.”—
ees a: a: s .5 ..ie mug—5E
moi-000. 0N. _0000. 0voooo._ 002.000. h0vw000. N0w0000. w0ww000. bwww000. _ _ _ooo._ Nohw00 . OMNM000. mo Lon—=3: _So.._.
8.... s: :3. 3:2: mason
omooo; Nw _ooo._ 0000000. bowooo; :looo; vnmw000. mNoooo._ N _ ooo._ hvwo000. .woooo; m0 _ V000. mo ..onEsz
as :3 e e: 0:8 £13253
:v0000. nooooo. _ ooooo. _ Nwo0000. omoooo; oooooo. _ n _ oooo. — _oooo. _ 0voooo. _ omoooo. _ vovooo. _ 1:02
a .8 3.. .3:
53000. M3300 . wvoooo; m0_0000. Nob0000. w _ 00000. Mvvw000. E00000. vomooo; E00000. MVNooo; 5533—2
.9: _.. on _._
m _ booo; m_0ooo._ _ :ooo; «MhNoo; .wnno00. vaooo._ «cemoo; mmo0000. _mem00. wmmwo00. 3.3000. EVEN—HEE—
s 8.3.. e _.. Nto>om
0moooo._ Nmoooo; 0_oooo._ o_oooo._ NN00000. Nw0w000. 360000. 0000000. 020000. o0w0000. w_oooo._ E eons—=2
1. :3 .33..
N0mo0v0. Nmm _ 000. m _ woww0. w _ vavq w _ goo; vvg mo._ Smoob0. mmovmo; mmwowo; 0mm _ M00. wmnbom0. EthoEEmED
6 if ~=DENO_NEO
0000000. wooooo. _ me _ oo._ _ M31000. o0 _ ooo._ mwoooo; NNNooo. _ 0moooo._ 0owooo._ _NNooo._ envooo; omzofuﬁwnﬂm
nut—ONO“ 3h°m8¢ EO 0M— mtnvmmuﬂ— 88Lh< 3m0hh< 38hh< 38hh< mushh< aghh< 38hh< 8—Dlmha>
a—ﬂﬁnn< on; Luau-u: IOU—u— u—Sﬂmm< XOW k—msﬂn— a; Luge—=2 “EU—OT, —3°p—t «CG—vﬂaueﬂ—
82-33

.328:— ue 83353 25.—U 2: no ﬁeEatEr .55.. o... no .36 no_na_._a> 3.5.80 one Economy—.5 2: we 33cm— .0N 93:.

161

added to the slaughterhouse categorization), slaughterhouse employment has a signiﬁcant
positive effect on total arrests and arrests for violent offenses. It is also instructive to
examine the analysis of the data for the entire time period from 1994 through to 2002.

Table 30 displays the IRR coefficients for slaughterhouse employment for the
pre-1998 period and the entire study period (1994-2002). These analyses were conducted
with the most complete model, referred to as model 3 in the analyses above, with all of
the control variables and industry variables included.

Table 30. Comparison of [RR1 values for the Effects of Slaughterhouse
Employment Prior to 1998 and for the Entire Study Time Period (1994-2002) on the

 

 

 

 

 

 

 

 

 

 

 

 

Dependent Variables
Dependent Variables Effect of Effect of
Slaughterhouse Slaughterhouse
Employment, Pre- Employment, 1994-
1998 2002
Total Arrests 1.000454M .9993113***
Violent Arrests 1 .000221* .999381 1*“
Murder Arrests 1.000869 1.000044
Rape Arrests 1.000059 1.000327“
Offenses against the 1.000222 .9996448***
Family Arrests
Sex Offense Arrests 1.000083 1.000202*
Assault Arrests 1.000196 .9994051***
Index Reports .9997331 .999141***
Murder Reports 1.001345 1.000081
Rape Reports 1.000008 .9993045***
Assault Reports .9999967 .9992509***

 

 

 

 

 

* Signiﬁcant at p < .05; *" Signiﬁcant at p < .01; *** Signiﬁcant at p < .001.
‘ Incidence-Rate Ratio (IRR)

There are notable differences in the results of these two time periods. Whereas
there were not any instances where slaughterhouse employment had a signiﬁcant
negative effect on the crime variables in the pre-l998 analyses, there are several of these

instances in the analysis of the 1994-2002 data. However, it is necessary to use caution

162

in interpreting many of the IRR coefﬁcients less than 1 for slaughterhouse employment in
the analyses across the entire study period due to the inclusion of “custom slaughtering”
facilities in the animal slaughtering category. Unfortunately, due to changes in the
Uniform Crime Reports data from 1993 to 1994, and the change in the categorization of
slaughtering in 1998, the 1994 to 1997 period is the only one in which we have crime
data that can be compared across time and slaughterhouse data that excludes small
“custom slaughter” facilities.

Additionally, in the 1994-2002 analysis of the data, slaughterhouse employment
has signiﬁcant positive effects on two of the crime variables: arrests for rape and arrests
for sex offenses (excluding rape). In the analysis of the restricted time period (1994-
1997), slaughterhouse employment had a positive effect on rape arrests and arrests for
sex offenses, but they were not signiﬁcant. It is likely that the increase in the number of
observations in the 1994-2002 analysis is related to the signiﬁcance of these ﬁndings.

Table 31 presents the effects of all of the control variables and comparison
industries for the entire time period under study (1994-2002). Particularly noteworthy is
the fact that many more of the control and comparison variables have signiﬁcant effects
for the entire time period than for the restricted time period (1994-1997) due to the
increased number of cases.

The unemployment variable in the 1994-2002 analysis continues to have some
signiﬁcant negative effects on the crime variables, but it also has some signiﬁcant
positive effects not found in the analysis of the restricted time period. The number of
people in poverty variable also has a signiﬁcant effect on two of the variables (index

reports and murder reports) not found in the restricted analysis.

163

 

   

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.8. v o a again ...: :e. v o a anemia .. .8. v a a Hagan .
o .. .. £23.53
:3. _ oo._ mvoooo. — 0meoo. _ M0 _ woo; vmww000. vooooo; 0oo_oo._ $5000. 0cmooo._ v _ oooo._ 0oNNoo._ Egg—...:—
.. ...... mg
NNwow00. w0oN000. 0mmw000. mmwN000. VN _ ooo._ £00000. 00 _ 0000. vmw0000. mwoooo; 0 _ oooo._ 00Nooo._ :m_w
a
N0vw00. 0vo0w00. mNoooo._ wwom000. 0_om000. cmgoo; 0000000. 00Nooo._ wNNooo._ nNom000. MNOooo._ o_oEo> .552
t. . a
vmvooo; m _ mooo._ ov0w000. oo0ooo._ mwoooo._ 0mmooo._ Nooo._ m0Nooo._ VONV000. wooooo; VN©N000. 3:20. x030.
...... .... .3... ...... n ...... ...... mEmhom
_NMON00. 0wN0o00. 0_0moo._ Nvo0m00. mo_VN00. OVOVN00. moovv00. mowON00. 030m00. V00w_00. 00M_N00. 30% was :0..—
s ... ...... .9250
new _ @000. Nm _ooo._ 0v. moo; ommww00. omomoo; Eomoo; _mmmoo._ NVnooo; m00moo._ owwmoo; wwwoo.— .5325?
«.... .. .3... ...... ...... ...... ...... ...... ...... mpg—ME
00—0000. V00000. va0000. mmo0000. m _ ww000. $00000. _0ww000. 350000. wm_0000. 0o0w000. 00ow000. mo nun—=3: .39..
... .. a... .. .... ... moi:— mczoun
V0m0000. Noo0000. 500000. VO00000. mow0000. m0w000. NNw0000. mw0000. ovoooo; vw0000. 0w00000. mo hon—5.2.
...i .... ...: ...: ...... ...... ...... .3... ...... a...» EcumEEoEEB
0Noooo. _ omoooo. _ vwoooo. _ mooooo. — 0voooo. _ 0N00000. voooo. _ wooooo. o mooooo. _ mmoooo._ mo _ ooo. _ 1:02
8.... .... .... ...... ...... ...... ...... ...... ......
unmooo. _ 3. _ ooo._ 00Nooo.— mmmoooé Nmmooo._ wooooo._ wvoooo; womooo; 0N3oo. _ wmmooo; _0mooo.— corny—92
...: .... .3... .. s
o0VNoo._ E m _ oo._ v0 _ 0000. 0 _ 0Noo._ momooo.— mow—oo; VVN _ oo._ m0mooo._ 0moooo.~ ~0vooo._ 0mmw000. cage—HEE—
: .. .... .... e b.5200
vmoooo; wooooo._ 0moooo._ mooooé _oooo._ 00v000 . vwm0000. 0 _ oooo._ wmw0000. wooooo._ vooooo; E 338:2
e ...... .. .... ...... n ......
mewvw0. woo _ oo._ mvgoo; m0omow0. ova _ o. _ vvowoo; mwmwv00. N0m0vo._ w _ 00 _ o. — 0ov _ o. _ N0vvmo._ EoEAEQEonD
...: ...... ...... ...... ... ...... a.» ...i ...»... “cognoEEo
0omN000. mvom000. _ woooo. _ :e _ 000. _ mov000. NoNooo. _ wvvo000. 0Nmooo. _ vvoooo. _ _ 33000. 2 _ M000. omnofoﬁwnﬂm
Own-Om“: ”Ht—Ga Owe—Cg mun-Ga mwmﬂhh< WHOOLL< 3w0hh< m~m0hh< SMOLL< 8m0hh< WuMOLL< 00_Qﬂmh“>
...—name. on; .5952 .30:— ==amm< new Egan an: .5952 .535 .58. «no—Enact:—
NooNtv00~

.3285 no 833...; «EEO on. no Ame—nets.» .35.. o... no 87o 833.:3 ......neU one «no—32.2.5 on. an 88.5— .nm 033—.

Many of the positive effects of the non-Whites and/or Hispanic variable found in
the restricted analysis become signiﬁcant in the analysis of the entire time period.
However, the effects are substantively very small. The number of young males variable
performs less well in the full analysis, having signiﬁcant negative effects on ﬁve
variables and no signiﬁcant positive effects (there were two signiﬁcant positive effects in
the restricted analysis). Additionally, the total number of males variable only has
signiﬁcant negative effects (as it did in the restricted analysis).

There are some interesting differences in the social disorganization variables as
well. Population density, which only had a signiﬁcant negative effect on index reports in
the restricted analysis, has a signiﬁcant positive effect on three variables in the full
analysis (total arrests, arrests for offenses against the family, and arrests for assault). In
the full analysis, immigration has a signiﬁcant positive effect on three additional
variables: arrests for sex offenses, reports of rape, and reports of assaults. The effects of
the migration variable are much more in line with what is theorized in the literature in the
full analysis: whereas it had one signiﬁcant positive effect and three signiﬁcant negative
effects on the crime variables in the restricted analysis, in the full analysis it has a
signiﬁcant positive effect on nine of the crime variables (the only two variables it does
not have a signiﬁcant effect on are arrests for family offenses and for sex offenses,
although the effect is in the expected direction).

Of the comparison industries, Iron and Steel Forging has many more signiﬁcant
negative effects than it did in the restricted analysis (seven versus one) and still no
signiﬁcant positive effects. In the full analysis Truck Trailer Manufacturing has a

signiﬁcant positive effect (on index reports) which it did not have in the restricted

165

analysis. In the full analysis, Motor Vehicle Metal Stamping does not have any
signiﬁcant effects (it had two negative effects and one positive effect in the restricted
analysis). Sign Manufacturing has two signiﬁcant negative effects in the full model
instead of three in the restricted model. Finally, the Industrial Launderers variable did
not have any signiﬁcant effects in the restricted analyses, but has two signiﬁcant positive
effects (on total arrests and index reports) in the full time period analysis. The
conclusions than can be drawn from the restricted and full time period analyses will be
discussed in the next chapter. Before moving to that chapter, the effects of various levels
of slaughterhouse employment when the control variables are restricted to their means are

explored below.

Estimating the Effects of Levels of Slaughterhouse Employment

In understanding the effects of slaughterhouse employment on crime it is useful to

estimate the effects of slaughterhouse employment keeping the control variables stable at

their means. For the negative binomial regression the following equation is used:

 

 

I)r(y'il : yil ----- yin, :31”..- l Xi: 21;] K}! : 2;];13/1!)
1(2)! 1’\ NZ: 1 3/ f +1) )H POW +11”)
[(20.11va + 2;:191! 1(Art)(yit +1)

(StataCorp 2005).
Table 32 presents the results of estimating the effects of differing levels of
slaughterhouse employment on four variables that the above analyses have indicated that

slaughterhouse employment signiﬁcantly predicts, net of the control variables: The total

I66

number of arrests and the total number of arrests for violent offenses (both restricted to
the pre-1998 data, before “custom slaughter” facilities were added) and arrests for rape
and sexual offenses (for the entire study time period). The values are reported as
incidence rates.

The results indicate that compared to a county with no slaughterhouse employees
an average slaughterhouse employing approximately 175 people in a typical county (with
all of the control variables held at their means) would result in a .0000175 incidence-rate
increase in the number of total arrests, .0000307 incidence-rate increase in arrests for
violent offenses, .000017 incidence-rate increase in arrests for rape, and .0000092
incidence rate increase in arrests for sexual offenses. The incidence rates with 7500
slaughterhouse employees are more than double the value with no slaughterhouse
employees for arrests for violent offenses, arrests for rape, and arrests for sexual offenses
(the value for the total number of arrests is nearly double). The increases in the predicted
incidence rates for these crime variables across the levels of slaughterhouse employment
are presented graphically in Figure 6. These may appear to be small effects but they are
signiﬁcant and in the expected direction; these effects are not paralleled by the
comparison industries; there are numerous factors that likely explain crime in rural
counties; and for counties with high levels of slaughterhouse employment these are not

trivial effects.

I67

Table 32. Incidence Rates Obtained via the TSCS Negative Binomial Regression
Equation at Varying Levels of Slaughterhouse Employment, Keeping Control

 

 

 

 

 

 

 

 

 

 

 

Variables

Slaughterhouse Total Arrests Arrests for Arrests for Arrests for
employment Violent Rape Sexual

Offenses Offenses
0 employees .0013987 .0007741 .0002903 .0002586
10 employees .0013997 .0007758 .0002912 .0002591
60 employees .0014047 .0007845 .000296 .0002617
175 employees .0014162 .0008048 .0003073 .0002678
375 employees .0014364 .0008413 .000328 .0002788
750 employees .0014751 .0009144 .0003 705 .0003007
1750 employees .0015835 .0011417 .0005129 .0003676
3750 employees .0018248 .00178 .000983 .0005497
7500 employees .0023 806 .0040928 .0033287 .0011687

 

 

 

 

 

 

Figure 6. Log Scale Prediction Equation Values for Total Arrests, Arrests for Violent

0.0045

Offenses, Rape, and Sexual Assaults

 

0.004

0.0035

.8.

0.0025

0.002

Incidence-Rate Ratios

0.0015

Average size

Large
slaughterhouse

slaughterhouse \

 

0.001

0.0005

 

 

/

 

 

#

375 750 1 750

Number of Slaughterhouse Employees

168

3750 7500

+Total Arrests

‘+Arrests for Violent Offenses

Arrests for Rape
7;: Arrests for Sexual Assaults

Summary

The results in this chapter provide partial support for research hypotheses 2 and 3. The
TSCS OLS regression with the arrest and report rate scales indicate that slaughterhouse
employment has an effect on the crime scales net of the control variables — an effect not
paralleled by the comparison industries. The results of the more conservative TSCS
negative binomial regression analyses demonstrate that controlling for the other variables
slaughterhouse employment has a significant positive ejfect on rape arrests and arrests
for sex oﬂ'enses across the entire study period (1994-2002) and that slaughterhouse
employment has a significant positive effect on the total number of arrests and the total
number of arrests for violent crimes for the time period before the addition of “custom
slaughter ” facilities to the animal slaughtering category. The next chapter provides a

detailed discussion and summary of the results presented in this and the previous chapter.

169

Chapter 5

DISCUSSION AND CONCLUSIONS

While stories of work-related tragedies at slaughterhouses are commonplace, the impact
that these facilities have on communities is every bit as disturbing. — Erik Markus, 2005

You have just dined, and however scrupulously the slaughterhouse is concealed in the
graceful distance of miles, there is complicity. - Ralph Waldo Emerson, 1860

This chapter reviews the results presented in the previous two chapters and outlines the
conclusions that can be drawn from this study. Then the implications for the literature on
slaughterhouse communities, as well as the broader literatures of community sociology,
criminology, and animal studies, are discussed. Finally, the recommendations that can be
made in light of the results of this study are explored, with particular attention paid to the

areas which this study indicates further research is needed.

Slaughterhouses and Crime: Making Sense of the Study Results

Bivariate correlation results

As mentioned in Chapter 3, bivariate correlations are not necessarily the best indication
of the relationship between variables because the effects of other variables are not
controlled for. Further, the correlation coefﬁcient is really a transformation of the b
coefﬁcient from OLS regression and is therefore in this case vulnerable to all of the

problems discussed in chapter 2 related to using OLS regression to analyze count data.

I70

Therefore, I only brieﬂy summarize the bivariate results below and subsequently focus on
the multiple regression results.

Of particular note from the bivariate correlation results is the variation in the
correlations between the slaughterhouse variables and the control variables. The
correlations between the slaughterhouse variables and the economic variables (poverty
and unemployment) were fairly weak, and the correlations between the slaughterhouse
variables and the unemployment variable were negative, meaning that an increase in
slaughterhouse employment is associated with a decrease in the unemployment rate. Of
the other social disorganization variables, population density had small positive
correlations with the slaughterhouse variables (increases in the number of slaughterhouse
employees and establishments are associated with small increases in the population
density). Contrary to the theorizing in the literature, the migration variable was
negatively correlated with the slaughterhouse employment and establishment variables (-
0.176 and -0.08), such that an increase in the slaughterhouse employment and
establishment variables is associated with a decrease in the net number of migrants into
the county. Consistent with the literature, however, there was a strong positive
correlation between the immigration variable and the slaughterhouse employment
variable (r = 0.513), demonstrating that an increase in slaughterhouse employment in a
county is associated with an increase in the net number of immigrants into the county.
(The correlation with the slaughterhouse establishment variable had a smaller value of
0.188.) The correlations between the slaughterhouse variables and the demographic
variables were also consistent with the theorizing in the literature. The number of non-

Whites and/or Hispanics (r = 0.1 13), the number of males (r = 0.168) and the number of

I71

young males (r = 0.187) all had moderate positive correlations with the slaughterhouse
employment variable. Thus, increases in slaughterhouse employment in the counties
studied are associated with increases in the number of non-White and/or Hispanic
residents, the number of males, and the number of young males.

The correlations between the slaughterhouse and crime variables also varied,
although there was a fairly consistent stronger relationship between the slaughterhouse
employment and crime variables than between the slaughterhouse establishment and
crime variables. The correlations between the slaughterhouse employment variable and
the summary arrest variables were modest, and some were close to zero. The values
range from -0.013 (with violent crime) to 0.037 (for property crime). The economic
variables (poverty and unemployment) and the population density variable demonstrated
the strongest correlations with the summary crime variables.

The slaughterhouse variables also had fairly small correlations with the specific
arrest variables, and some of the correlations were negative (meaning that an increase in
slaughterhouse employment is associated with decreases in those crime variables). The
values ranged from -0.016 for aggravated assault to 0.076 for other assaults. The results
of the correlations between the slaughterhouse variables and the crime report variables
were somewhat different: with the exception of one variable (murder) all of the report
variables were positively correlated with the slaughterhouse variables, demonstrating that
an increase in slaughterhouse employment in the counties studied is associated with
increases in crime reports (with the exception of murder). These correlations, however,
were also modest, ranging from a low of -0.22 with reports of murder to a high of 0.16

with larceny.

172

Overall, there was a stronger positive relationship between the crime report
variables and the slaughterhouse variables than between the arrest and slaughterhouse
variables. The control variables were more strongly correlated with the arrest and report
variables than the slaughterhouse variables were. Based upon only the bivariate
correlation results the immigration and migration variables do not appear to perform as
well as anticipated in the literature (the multiple regression results point to somewhat
different ﬁndings). Again it is important to note that conclusions about the control
variables and the research hypotheses cannot be drawn from the bivariate results. The
results of other techniques and the testing of the research hypotheses must instead be

consulted.

Testing Hypothesis 1

Hypothesis 1 addressed the ﬁrst important question: Do general and speciﬁc crime rates
increase as the number of slaughterhouse employees increases? Recall that this question
was addressed through the use of non-directional t tests of the difference in means (1)
between counties with slaughterhouses and counties without slaughterhouses, and (2)
between counties with high levels of slaughterhouse employment (1,000 or more
slaughterhouse employees) and counties with less or no slaughterhouse employees (0-
999)

In testing the difference between counties with high levels of slaughterhouse
employment and those with less or no slaughterhouse employment signiﬁcant differences

in the expected direction (higher crime rates in counties with high levels of

173

slaughterhouse employment compared to counties with less or no slaughterhouse
employment) were observed in two of the summary arrest variables (total arrests and
index offense arrests) and among four of the speciﬁc arrest variables (arrests for rape,
other assaults, sex offenses, and disorderly conduct). Of the report variables, signiﬁcant
differences were observed in the one summary variable (index offense reports) and
among three of the speciﬁc report variables, including reports of rape, burglary, and
motor vehicle theft.

The results of the tests of the differences in means between counties with and
without slaughterhouses demonstrated that although counties with slaughterhouses had
higher mean levels of total arrests, arrests for property crimes, rape, other assaults, and
reports of assault, these differences were not signiﬁcant. The only statistically signiﬁcant
mean difference in the expected direction for the arrest variables was for sex offenses.
For the report variables the following mean differences were signiﬁcant: reports of index
offenses, rape, robbery, burglary, motor vehicle theft, and arson. However, two of the
variables (violent crimes and aggravated assault) had signiﬁcant mean differences that
were not in the expected direction. These counterintuitive results were not found in the
analyses testing the mean difference in the crime variables between counties with high
levels of slaughterhouse employment (1,000 or more employees) and counties with less
or no slaughterhouse employment.

Thus, the results of this study partially support hypothesis 1. When we examine
the difference between counties with and without slaughterhouses, the results show that
counties with slaughterhouses tend to report signiﬁcantly more offenses (most

speciﬁcally, reports of index offenses, rape, robbery, burglary, motor vehicle theft, and

174

arson) than counties without slaughterhouses (the exceptions are reports of murder and
assaults). However, the differences in arrest rates between counties with and without
slaughterhouses are not as evident. Compared to counties without slaughterhouses,
counties with slaughterhouses only had a signiﬁcantly greater mean arrest rate for one
variable: arrests for sex offenses. Once the differences between counties with high
levels of slaughterhouse employment and counties with less or no slaughterhouse
employment are examined the distinctions become more clear. Here there are signiﬁcant
differences in many of the arrest variables, including total arrests, index offense arrests,
and arrests for rape, other assaults, disorderly conduct, and sex offenses (the only arrest
variable signiﬁcant in the slaughterhouse/no slaughterhouse comparison). Additionally,
the mean differences in the variables are dramatically different when we compare the
slaughterhouse/no slaughterhouse results with the high level of slaughterhouse
employment/little or no slaughterhouse employment. For instance, the mean difference
in total arrests between counties with and without slaughterhouses is -58.064, whereas the
mean difference between counties with high levels of slaughterhouse employment and
little or no slaughterhouse employment is -737.996. Thus, the mean difference in total
arrests between counties with high levels of slaughterhouse employment compared to
counties with little or no slaughterhouse employment is nearly thirteen times greater than
the difference in total arrests between counties with and without slaughterhouses.

It must be noted that the mean difference is not in the direction that the hypothesis
predicts for some variables in both of these analyses. These variables include arrests for
violent crimes, arrests and reports of murder, arrests for robbery, arrests for aggravated

assaults, and reports of assault. An examination of what these offense categories entail at

175

least partially explains these ﬁndings. The arrests for the violent offenses category is
comprised of arrests for murder, rape, robbery, and aggravated assault. Therefore, it is
not surprising that we do not see a signiﬁcant difference in the expected direction with
this variable, since such a difference is not observed in three of the variables that
comprise it (murder, robbery, and aggravated assault). It is also not surprising that we do
not witness a signiﬁcant difference in the expected direction for arrests for and reports of
murder. This is the most extreme type of crime and occurs relatively infrequently,
especially in non-metropolitan counties. It is also not surprising that there are not
signiﬁcantly more arrests for robbery in counties with slaughterhouses, since based on
the thesis of a spill-over of violence from the slaughterhouse into the community we
would not necessarily expect to see an increase in robbery, which is essentially a property
crime.

In contrast, the fact that there is not a signiﬁcant difference in the expected
direction for aggravated assaults at ﬁrst appears surprising. However, the exact deﬁnition
of this type of offense helps to clarify the findings. The deﬁnition of aggravated assault
is “an unlawful attack by one person upon another for the purpose of inﬂicting severe or
aggravated bodily injury. This type of assault usually is accompanied by the use of a
weapon or by means likely to produce death or great bodily harm” (Department of Justice
and Investigation 2004, 23). This includes attempted murder. Thus, these are very
serious violent offenses where weapons are used. Further, there are signiﬁcantly (p <
.001) more arrests for ‘Other Assaults’ in counties with more than 1,000 slaughterhouse
employees than in counties that do not have slaughterhouse employees or have less than

1,000. This category of offense “includes all assaults which do not involve the use of a

176

ﬁrearm, knife, cutting instrument, or other dangerous weapon and in which the victim did
not sustain serious or aggravated injuries” (Department of Justice and Investigation 2004,
24). Consequently, it appears that the while there might not be differences in the
extremely serious violent offenses (murder and aggravated assaults) there are signiﬁcant
differences in less serious violent offenses, such as Other Assaults.

Whether or not employment in slaughterhouses is responsible for the elevated
levels of some types of crime, such as other assaults, in counties with slaughterhouses
cannot be ascertained based on these analyses. It is possible that the relationships
documented here are spurious, and instead are the result of the variables proposed in the
slaughterhouse literature. In order to determine whether or not this is the case, the results

of testing Hypothesis 2 must be assessed.

Testing Hypothesis 2

Hypothesis 2 anticipated that controlling for the variables proposed in the literature,
slaughterhouse presence and employment would be associated with increased crime rates
in counties, more so than the comparison industries. Two techniques were employed to
test this (and the third) hypothesis. First, arrest and report rate scales were created using
factor analysis and then employed as dependent variables in ﬁxed effects time series
cross section (TSCS) OLS regression models containing the control variables and the
comparison industries. Second, utilizing the same model structures, individual arrest and

report variables were regressed using ﬁxed effects TSCS negative binomial regression.

177

The results demonstrate that slaughterhouse employment is a signiﬁcant predictor
of both the arrest and report rate scales. In the analysis with the arrest rate scale as the
dependent variable (comprised of the following nine variables: arrests for rape, robbery,
burglary, forgery, stolen property, vandalism, other assaults, offenses against the family,
and disorderly conduct), the regression coefﬁcient value is .019 (p < .001). Accordingly,
an increase of one slaughterhouse worker would be expected to increase the arrest rate
scale by .019 arrests. With the addition of the control variables the coefﬁcient decreases
to a value of .013, but this value is still signiﬁcant (p < .01).

A regression coefﬁcient value of .013 means that controlling for the other
variables, the addition of an average sized slaughterhouse of 175 workers in a community
would be expected to result in an increase in the arrest rate scale of 2.28 arrests. A larger
slaughterhouse of 1,000 employees would result in an increase of 13 arrests. The
comparison industries do not have parallel effects, and for all except one of the industries
(Industrial Laundering) the regression coefﬁcients are negative (but not statistically
signiﬁcant).

The effect of slaughterhouse employment is even more impressive with the report
rate scale (comprised of the following six variables: reports of rape, robbery, assault,
burglary, motor vehicle theft, and arson) as the dependent variable. In the bivariate
regression model, the coefﬁcient for slaughterhouse employment is .039 (p < .001). The
size of the coefﬁcient decreases to .027 with the addition of the control variables, but it
remains signiﬁcant (p < .01).

Controlling for all of the variables in the model we would therefore expect an

average slaughterhouse of 175 people to result in an increase in the report rate scale of

I78

4.73 reports and a large slaughterhouse of 1000 employees to result in an increase of 27
reports. Again, these results are not paralleled by the comparison industries, all of which
do not have signiﬁcant effects.

Positive effects of slaughterhouses on crime rates are also found using TSCS
negative binomial regression to regress individual arrest and report variables. Recall,
however, that the addition of ‘custom slaughter’ facilities in 1998 to the slaughterhouse
category affected the results of the analyses. Therefore, analyses of two time periods
from the data were undertaken: for the entire time period (1994-2002) and before
‘custom slaughter’ facilities were added (1994-1997).

In the results derived from the entire time period, and controlling for the
extraneous variables, slaughterhouse employment has signiﬁcant effects on arrests for
rape and arrests for sex offenses (increases in slaughterhouse employment can be
expected to result in increases in arrests for rape and sex offenses). The Incidence-Rate
Ratio (IRR) representing the effect of the slaughterhouse employment variable on arrests
for rape, controlling for all of the variables, is 1.0003 (rounded) and signiﬁcant (p < .01).

This means that for every additional slaughterhouse worker the expected number
of arrests for rape is increased by a factor of 1.0003 and that for every additional
slaughterhouse worker there is an expected increase of 0.03% in arrests for rape. Given
that some of the large slaughterhouse facilities employ thousands of people, the effect
could perhaps be more substantial than it appears. For instance, 4,000 slaughterhouse
workers would increase the expected number of arrests for rape by 1.2% [.0003 x 4000 =
1.2]. Of the comparison industries, only Iron and Steel Forging demonstrates a

signiﬁcant effect on arrests for rape, but it is a negative one, meaning that controlling for

179

the other variables an increase in employment in Iron and Steel Forging is associated with
a decrease in arrests for rape.

The IR value representing the effect of slaughterhouse employment on arrests
for sex offenses, controlling for the other variables, is 1.0002 (rounded) and is signiﬁcant
(p < .05). This value can be interpreted to mean that each additional slaughterhouse
worker increases the expected number of arrests for sex offenses by a factor of 1.0002 or
.02%. Therefore, an increase of 5,000 slaughterhouse employees in a county would be
expected to result in a 1% increase in arrests for sex offenses. None of the other
comparison industries demonstrate a signiﬁcant effect on the variable.

The effects of slaughterhouse employment on the arrests for rape and sex offense
variables are not signiﬁcant in the analysis of the data prior to the inclusion of ‘custom
slaughter’ facilities (1994-1997). (These facilities tend to be very small and are engaged
in slaughtering their own animals for their own consumption or the owned animals of
other individuals for their consumption.) This is not necessarily surprising given that the
analysis of the entire time period has more than double the number of observations than
the period before the inclusion of ‘custom slaughter’ facilities. For the analyses of the
entire time period (1994-2002), 4,646 observations are analyzed [581 counties x 8 years
(8 years of observations instead of 9 are included in the analyses as the result of the one
year lag) — 2 missing cases = 4,646]. For the analyses of the time period before ‘custom
slaughter’ facilities were added to the slaughterhouse categorization (1994-1997), 1,743
observations are analyzed [581 counties x 3 years = 1,743].

Slaughterhouse employment is a signiﬁcant predictor of two variables for the

period before ‘custom slaughter’ facilities were added to the slaughterhouse

180

categorization: total arrests and violent arrests. The IR (Incidence-Rate Ratio) value
for the slaughterhouse employment variable, controlling for the other variables, in
predicting total arrests is 1.0005 (rounded). This means that each additional
slaughterhouse employee is expected to increase the total arrests by .05% and by
extension be expected to result in an increase of 1.82% for a county with 4,000
slaughterhouse workers. Only one of the comparison industries (Truck Trailer
Manufacturing) has a signiﬁcant effect on the total arrests variable, but it is a negative
effect and therefore an increase in the number of Truck Trailer employees in these
counties would be expected to decrease the number of total arrests.

The IR value for the slaughterhouse employment variable in predicting violent
arrests is 1.0002 (rounded), controlling for the other variables. This translates into a
.02% increase in violent arrests for each additional slaughterhouse worker. Further,
4,000 slaughterhouse workers would be expected to increase violent arrests by 0.88%.
Two of the comparison industries (Truck Trailer Manufacturing and Motor Vehicle Metal
Stamping) have signiﬁcant effects on the violent arrests variable, but both are negative.
Again, we would therefore expect that an increase in the number of employees in these
industries would be associated with a decrease in the number of arrests for violent
offenses.

Thus, the results of the TSCS OLS regression and TSCS negative binomial
regression both demonstrate that slaughterhouse employment does have signiﬁcant
positive and unique effects, controlling for the variables proposed in the literature, on
some types of crimes. Most speciﬁcally, this study ﬁnds that controlling for the variables

proposed in the literature, slaughterhouse employment (1) has a signiﬁcant positive effect

181

on total arrests and arrests for violent offenses for the period before ‘custom slaughter’
facilities were added to the slaughterhouse category (1994-1997), which can be
interpreted to mean that as slaughterhouse employment increases in the counties under
study, we can expect an increase in the number of total arrests and arrests for violent
offenses; (2) has a signiﬁcant positive effect on arrests for rape and sex offenses for the
entire time period under study (1994-2002), which can be interpreted to mean that as
slaughterhouse employment increases in the counties under study, we can expect an
increase in the number of arrests for rape and sex offenses; (3) has a signiﬁcant positive
effect on the arrest rate scale (comprised of the following variables: rape, robbery,
burglary, forgery, stolen property, vandalism, other assaults, offenses against the family,
and disorderly conduct) across the entire time period, which can be interpreted to mean
that as slaughterhouse employment increases in the counties under study, we can expect
an increase in arrest rate scale; (4) has a signiﬁcant positive effect on the report rate scale
(comprised of the following variables: rape, robbery, assault, burglary, motor vehicle
theft, and arson) across the entire time period, which can be interpreted to mean that as
slaughterhouse employment increases in the counties under study, we can expect an
increase in the report rate scale. All of these effects are unique when compared to the
effects of the comparison industries (some of which actually have signiﬁcant negative
effects on the crime variables). Accordingly, we can conclude that Hypothesis 2 is

partially supported by the data.

182

Testing Hypothesis 3

Hypothesis 3 was more speciﬁc than Hypothesis 2, asserting that controlling for the
variables proposed in the literature, increases in slaughterhouse employment would be
associated with increases in rape and offenses against the family, and that increases in
these offenses would not be associated with increases in employment in the comparison
industries. This hypothesis is also partially supported here.

The effect of slaughterhouse employment on the offenses against the family
variable was signiﬁcant and negative for the analysis of the entire time period, and
positive but not signiﬁcant for the analysis of the 1994-1997 data. The negative effect
found in the 1994-2002 analysis could have been impacted by the custom slaughter
facilities. It is also worth noting that the offenses against the family variable consists of
unlawful nonviolent acts by family members against each other (Justice and Investigation
2004). Therefore, there is not a clear measure of family violence in the Uniform Crime
Reports that includes violence against family members. Perhaps the inclusion of violent
forms of offenses against the family in this variable would have made the effects of
slaughterhouse employment on it a little clearer. Additionally, we cannot assess the
effect of slaughterhouse employment on reports of offenses against the family, because as
previously mentioned only data on reports for Part I or Index offenses (these offenses
include murder, rape, robbery, aggravated assault, burglary, larceny, motor vehicle theft,
and arson) are collected.

As described above in the discussion of Hypothesis 2, the slaughterhouse variable

did have a signiﬁcant positive effect on the rape arrest variable across the entire time

183

period under study. However, this effect was not signiﬁcant when fewer observations
were analyzed for the period before ‘custom slaughter’ facilities were added (1994-1997).
Similarly, slaughterhouse employment did not have a signiﬁcant effect on the rape report
variable for the analysis of the years 1994-1997. Slaughterhouse employment did have a
signiﬁcant negative effect on the rape reports variable for the analysis of the entire time
period. It is possible that this result was impacted by the inclusion of the custom
slaughter facilities. As discussed in the previous chapter, due to the way the County
Business Patterns categorizes the employee data (eg. 1-19, 20-99) instead of reporting the
exact number of employees, the inclusion of small custom slaughter facilities likely
artiﬁcially increases the number of slaughterhouse workers in counties since the midpoint
of the ranges are used in the analyses, therefore diluting the possible effect of
slaughterhouse employment. Unfortunately, there is no way to disaggregate the
slaughterhouse data and exclude these facilities from the analysis.

The signiﬁcant positive effect of slaughterhouse employment on the sex offense
variable is also noteworthy. This variable excludes forcible rape and prostitution.
However, it does include sexual attacks on males, incest, indecent exposure, statutory
rape, and ‘crimes against nature’ (such as bestiality) (Justice and Investigation 2004).
Many of these offenses are perpetrated against those with less power and therefore this
evidence assists in supporting the intent of hypothesis 3, as discussed next.

Hypothesis 3 was tested because it was thought that the work done within
slaughterhouses might spill-over to violence against other notably less powerful groups,
such as women and children. This study provides partial support for this hypothesis.

Controlling for the variables proposed in the literature, slaughterhouse employment (1)

184

has a signiﬁcant positive effect on arrests for rape across the entire study period, which
can be interpreted to mean that as slaughterhouse employment increases in the counties
under study, we might expect an increase in the number of arrests for rape; (2) has a
signiﬁcant positive effect on sex offenses (excluding rape and prostitution) for the entire
study period, which can be interpreted to mean that as slaughterhouse employment
increases in the counties under study, we can expect an increase in the number of arrests
for sex offenses. Further, these effects are not paralleled by the comparison industries,
and it is possible that the effect of slaughterhouse employment on offenses against the
family would have been positive and signiﬁcant (instead of positive but not significant
for the period prior to the inclusion of ‘custom slaughter’ facilities) if violent offenses
committed by family members were included in this category.

Like much social scientiﬁc research, the results of this study are not clear-cut and
the research hypotheses cannot be fully supported or rejected here. Instead, only partial
support is found for all of the hypotheses. It is evident that there are some signiﬁcant and
unique effects of the slaughterhouse employment variable (on total arrests, arrests for
violent offenses, arrests for rape, arrests for sex offenses, the arrest rate scale, and the
report rate scale) that are not paralleled by the comparison industries — effects which the
variables proposed in the literature (unemployment, social disorganization variables, and
demographic variables) cannot explain. Again, it is worth noting, however, that given the
highly stochastic nature of the dependent variables in rural counties the signiﬁcant
ﬁndings presented here are quite suggestive. Additionally, the differences in the results
before and after custom slaughter facilities were added to the slaughterhouse category

also suggests that the industrialization of slaughter has the strongest adverse effects,

I85

whereas the addition of the smaller, custom slaughter facilities likely adds “noise” to the
analyses and may even be adding the effects of social capital (related to small businesses

and small-scale agriculture) to the analyses.

Theoretical Implications

This study has several theoretical implications, for the substantive literature on
slaughterhouse communities and also for the larger areas of community sociology,
criminology, and the study of non-human animals and society. The most critical
implications are discussed below.

This study substantiates the earlier slaughterhouse community studies that
documented increased crime in slaughterhouse communities and provides evidence for
the generalization of these case studies of communities where large slaughterhouses have
opened to communities with smaller and/or older slaughterhouses. When compared with
counties without slaughterhouses, those with slaughterhouses have signiﬁcantly higher
levels of arrest for sex offenses, and reports of index offenses, rape, robbery, burglary,
motor vehicle theft, and arson. And counties with 1,000 or more slaughterhouse
employees have signiﬁcantly higher levels of total arrests, index offense arrests, arrests
for rape, other assaults, sex offenses, and disorderly conduct, index offense reports,
reports of rape, burglary, and motor vehicle theft. The inclusion of comparison industries
in the analyses further substantiates the claim that there is something unique about the
effects of slaughterhouses and that they should be analyzed as a special case of

manufacturing.

186

As found in this study, through the inclusion of the comparison industries in the
models, employment in the manufacturing sector (excluding the slaughterhouse industry)
can actually have negative effects on crime. One study of non-metropolitan counties
found that small manufacturing establishments were associated with lower rates of
homicide, robbery, burglary, and motor vehicle theft (Lee and Ousey 2001). Camasso
and Wilkinson (1990) have speculated that manufacturing employment has a mitigating
effect on crime therein because the disruptive potential of growth in the manufacturing
sector is outweighed by the fact that employment tends to be more consistent and is not
dramatically affected by market variations.

The current study demonstrates, however, that not all manufacturing industries
have these calming effects on crime rates. Despite the consistent levels of employment in
slaughterhouse facilities, not only do they not have clear signiﬁcant negative effects on
crime rates like other manufacturing industries, slaughterhouse employment has
signiﬁcant positive effects on some types of crimes. Thus, the effects of slaughterhouses
on communities appear quite unique and warrant special examination. These nuances
should be attended to by community and criminology scholars.

This study also tests the theories proposed to explain the increase in crime rates in
slaughterhouse communities. These theories have been grouped herein into three
categories: unemployment, social disorganization, and demographic factors. By
introducing variables from these theories as control variables this study is the ﬁrst to test
whether or not these theorized causes have independent effects on crime rates in counties

with slaughterhouses. Some of the control variables included in the models performed

187

better than others. The relative performance of the variables, and by extension the

theories, are assessed below.16

Unemployment: As discussed in Chapter 1, it has been proposed in the literature that the
increases in crime in slaughterhouse communities might be related to increases in the
unemployment rate as a result of the combination of increases in the number of people
moving to the county for work and the high turnover rate in the industry. The county
unemployment variable demonstrated a signiﬁcant positive effect on the arrest and report
scales in the Time Series Cross Section (TSCS) OLS regression models. The regression
coefficient was 1.17 with the arrest rate scale as the dependent variable and 2.03 with the
report rate scale as the dependent variable. This can be interpreted to mean that a one
percent increase in the unemployment rate would increase the arrest rate scale by 1.17
arrests and the report rate scale by 2.03 reports, controlling for the other variables. In the
TSCS negative binomial regression models with individual crime variables as the
dependent variables, unemployment also has a signiﬁcant positive effect on the following
variables: total arrests, violent arrests, rape arrests, and assault arrests. The Incidence-
Rate Ratio (IRR) values corresponding to these effects range from 1.01 to 1.05 (rounded).
However, it should also be noted that the unemployment variable has signiﬁcant negative
effects on arrests for offenses against the family, reports of index offenses, and reports of
assaults. Therefore, in this data unemployment displays mixed effects on crime rates.
These mixed effects have been documented in other studies (Cohen and Land 1987;

Kawachi, Kennedy, and Wilkinson 1999).

I88

Social Disorganization Variables: The other economic variable included in this study,
which falls under the scope of the social disorganization theory, is the number of people
in poverty variable. Brieﬂy stated, the social disorganization theory proposes that
increases in population, ethnic heterogeneity and low socio-economic status, results in
the weakening of social control functions and increases in crime. The number of people
in poverty variable was not a signiﬁcant predictor of the arrest rate scale, but it did have a
signiﬁcant positive effect on the report rate scale (b = .006, meaning that an increase of
one person in poverty results in an increase of .006 in the report rate scale, controlling for
the other variables). In the TSCS negative binomial regression the poverty variable only
had signiﬁcant positive effects on some of the report variables (increases in the number
of people in poverty were associated with increases in the reports of index offenses,
reports of murder, and assault reports, controlling for the other variables). This variable
had signiﬁcant negative effects on two of the arrest variables (controlling for the other
variables, increases in the number of people in poverty were associated with decreases in
arrests for offenses against the family and sex offenses). Therefore, as the number of
people in poverty increases in non-metropolitan counties in right-to-work states, crime
reports increase but arrest rates do not (and in two cases — arrests for offenses against the
family and sex offenses — they actually decrease, controlling for the other variables).
Perhaps this discrepancy is related to poorer counties having fewer law enforcement
resources to follow-up on crime reports and make arrests.

Population density Was also included as a measure of social disorganization, since
it is hypothesized that higher density living results in weakening of social bonds and

social control functions. The explanatory power of this variable, however, is not

189

impressive. It was not a significant variable in explaining the report rate scale and had a
signiﬁcant negative effect on explaining the arrest rate scale (b = -.56, which means that
an increase of one person per square mile is associated with a decrease of .56 in the report
rate scale, controlling for the other variables). Further, in the TSCS negative binomial
regression models it was only a signiﬁcant positive predictor of three of the eleven crime
variables (total arrests, arrests for offenses against the family, and arrests for assault).

The IR values ranged from 1.003 to 1.006, which actually are fairly large effects once
we consider that these are the factor increases for a one person increase per square mile.
In considering its effect on the scales and the individual arrest and report variables,
however, the variable did not perform as well as anticipated in the slaughterhouse
literature, wherein it is anticipated that part of the increase of crime in slaughterhouse
communities might be attributed to the increased population concentration as the result of
the new employment opportunities. Other researchers (such as Camasso and Wilkinson
(1990) in examining energy boomtowns) have also found that increases in population
have not had the expected effects on many types of crime.

According to the social disorganization theory, the movement of people into a
community, especially those who possess values different from the host population and
increase ethnic heterogeneity, can result in social disorganization and impact crime rates.
In this study, this possibility was tested through the use of two variables: net immigration
and migration into the county. Both of these variables were signiﬁcant positive
predictors of the report rate scale. The regression coefﬁcient value for the immigration
variable is .264 (which can be interpreted as meaning that an increase in one net

immigrant into a county is expected to result in an increase of .264 in the report rate

190

scale) and for the migration variable is .014 (meaning that an increase in one net migrant
into a county is expected to result in an increase of .014 in the report rate scale).
However, only the immigration variable was a signiﬁcant positive predictor of the arrest
rate scale (b = .07; therefore, an increase of one net immigrant into the county is expected
to result in an increase of .07 in the arrest rate scale). The migration variable, however,
had a signiﬁcant positive effect on more of the crime variables in the TSCS negative
binomial regression than the immigration variable: the migration variable demonstrated a
signiﬁcant positive effect on nine of the crime variables (the only two it did not have a
signiﬁcant effect on were arrests for offenses against the family and sex offenses). The
IR values range from 1.0001 to 1.0004. The immigration variable had signiﬁcant
positive effects on ﬁve variables: arrests for offenses against the family, arrests for sex
offenses, reports of index offenses, reports of rape, and reports of assault. Increases in
the number of net immigrants can therefore be expected to result in increases in these ﬁve
offenses. The IR values for the effects of immigration on the crime variables are larger
and range from 1.001 to 1.003. The signiﬁcant positive effect of immigration on fewer
crime variables than the migration variable is not necessarily surprising, since some
studies have found immigration not to be a signiﬁcant predictor of various crime (such as
Lee, Martinez and Rosenﬁeld (2001)).

Of the social disorganization variables employed in these models, the migration
and immigration variables make larger contributions compared to the poverty and
population density variables. However, the theory as a whole is not overwhelmingly
supported here. This is consistent with Nielson, Lee, and Martinez’s (2005) statement

that social disorganization theory has been useful in explaining the distribution of violent

191

crime rates nationally, but that there are more local concerns that need to be addressed to
explain variations and deviations from what the theory would predict. The type of

industries located in communities is one such local concern.

Demographic variables: As discussed in Chapter 1, it is theorized in the literature (and in
slaughterhouse communities themselves) that the increases in crime observed in
slaughterhouse communities might be related to changes in the demographic
characteristics of the community populations. Three demographic variables were
included in the models here: the number of non-whites and/or Hispanics, the number of
young males, and the total number of males. Of these three, the former has the most
explanatory power: however, none of the variables perform as expected. The number of
non-whites and/or Hispanics is a significant positive predictor of the arrest rate scale (b =
.008) and the report rate scale (b = .012). We would therefore expect an increase of one
non-White and/or Hispanic resident in a county to result in a .008 increase in the arrest
rate scale and a .012 increase in the report rate scale, controlling for the other variables.
However, these effects are less than those of the slaughterhouse employment variable.
The variable is also a significant positive predictor of ten of the crime variables in the
TSCS negative binomial regression models (the only exception is the arrests for sex
offenses variable). It should be noted, however, that the effects are fairly close to zero,
with IR values ranging from 1.00003 to 1.0001. Other studies have also found racial
heterogeneity not to be an important predictor of many types of crime (such as Rice and

Smith (2002)).

192

The number of young males and the total number of males variables generally
produced results here contrary to what has been hypothesized in the literature. The total
number of males variable demonstrates a signiﬁcant negative effect on the arrest rate
scale (b = -.009) and the report rate scale (b = -.019), such that controlling for the other
variables, an increase of one male in a county would be expected to result in a .009
decrease in the arrest rate scale and a .019 decrease in the report rate scale. The effect of
the young males variable is not signiﬁcant in these models, but it does appear to be in the
negative direction. Further, in the analysis of the TSCS negative binomial regression
results, neither of the variables have a signiﬁcant positive effect on any of the crime
variables. In contrast, the young males variable has signiﬁcant negative effects on ﬁve of
the variables (total arrests, arrests for sex offenses, index reports, rape reports, and assault
reports) and the total number of males variable has signiﬁcant negative effects on all but
two of the crime variables (the exceptions are arrests for sex offenses and reports of
rape). It should be noted, however, that these negative effects are fairly close to zero.
The IR values for the young males variable ranges from .99998 to .9999 and for the
total number of males variable the values range from .99999 to .99992. So, although the
number of young males and total number of males variables have statistically signiﬁcant
negative effects on several crime variables, substantively these negative effects are very
small and might be reasonably considered zero. Therefore, we can conclude that the
number of males and young males in non-metropolitan counties in right-to-work states
have no meaningful effects on the arrest and report rates examined here. Similarly, other

studies have found the age and gender distribution not to be useful predictors of crimes

193

such as homicide and burglary (Lee and Ousey 2001; Lee, Martinez, and Rosenﬁeld
2001).

The above theories of crime causation have largely been developed and tested in
studies of metropolitan areas. Therefore, some theorized causes of crime might not be as
applicable in non-metropolitan areas, which is where most slaughterhouses have moved
to in recent years. For instance, another study of non-metropolitan counties found that
the proportion of young people is not a signiﬁcant predictor of some types of crime that
one would otherwise presume to be connected with young perpetrators (Lee and Ousey
2001). Those undertaking studies of rural communities, therefore, must be careful not to
erroneously accept assumptions based upon experiences in metropolitan areas.

According to the results, the control variables with the most explanatory power in
predicting the arrest and report rates are the unemployment variable and some of the
social disorganization variables (speciﬁcally migration and immigration). The effects of
the demographic variables were largely contradictory and close to zero. This study is not
meant to be considered the deﬁnitive testing of these theories. It is possible that some of
these variables might have displayed stronger effects if only counties with new
slaughterhouses were analyzed. As previously mentioned, there were too few of such
counties in this sample to meaningfully analyze them separately. Additionally, it is
possible that different operationalizations of the theories might have resulted in slightly
different ﬁndings. However, given the number of key control variables employed and the
two analytical techniques used to test the models, the central ﬁndings of this study can be

considered robust.

194

There are two central ﬁndings of this study. First, increases in slaughterhouse
employment can be expected to result in increases in the arrest and report rate scales and
some of the individual crime variables (total arrests, arrests for violent offenses, arrests
for rape, and arrests for sex offenses) in spite of controlling for the effects of the variables
proposed in the literature and the time invariant elements in the counties that can affect
crime rates, such as history and climate. Second, these positive effects are not paralleled
among the comparison industries, which is substantiated by previous studies which found
that levels of employment in the manufacturing sector has negative effects on some types
of crime in non-metropolitan areas (Camasso and Wilkinson 1990; Lee and Ousey 2001).
It is particularly noteworthy that the ﬁndings of this study are documented at a fairly high
level of aggregation — the county. These effects of slaughterhouse employment might be
even more substantial at a lower level of aggregation, such as the township.

The results of this study point us toward an explanatory factor not considered in
the literature dealing with increased crime rates in slaughterhouse communities: the
unique work of killing and dismembering animals undertaken in slaughterhouses might
result in negative spill-over effects into the larger community. To an individual who
accepts at face value the categorization of the slaughterhouse industry as a form of
manufacturing, the results of this study might appear surprising. In contrast, if one
considers the work undertaken in slaughterhouses to be different and more complexly
layered than the manufacturing of inanimate objects, these results might not be surprising
at all.

This study demonstrates that there are meaningful theoretical and empirical

distinctions that can be drawn between slaughterhouse employment and other types of

195

employment categorized as manufacturing (and perhaps that there are meaningful
differences between industrialized slaughter and small-scale ‘custom’ slaughter). Studies
of communities (and community crime in particular) need to attend to these nuanced
differences (additional recommendations for further research are detailed in the next
section). Although our society, and especially our economic institutions, treats food
animals as goods to be produced, research in this area can no longer overlook the
possibility that workers who engage with them on a daily basis may at least occasionally
have a difﬁcult time dealing with these animals merely as commodities imbued with
economic value. The social constructions of food animals as commodities may at times
slip, exposing workers to some unsettling experiences, as illustrated by a slaughterhouse
worker in the following quote: “You may look a hog in the eye that's walking around
down in the blood pit with you and think, God, that really isn't a bad looking animal.
You may want to pet it. Pigs down on the kill floor have come up and nuzzled me like a
puppy. Two minutes later I had to kill them — beat them to death with a pipe” (cited in
Eisnitz 1997). Social scientists in particular ought to be wary of accepting deﬁnitions
(such as that of food animals purely as economic goods) which obscure underlying power
relations. There are clearly reasons to delve deeper into the possibility of a connection
between the type of work conducted in slaughterhouses and the increased crime rates in
slaughterhouse communities.

The positing of a general link between the work conducted in slaughterhouses and
larger social problems is not without precedent. The Greek philosopher Plutarch (AD 46
— 120) suspected a connection between the slaughter of animals and the slaughter of

people. He states,

196

in the beginning, some wild and mischievous beast was killed and eaten,
and then some little bird or ﬁsh was entrapped. And the desire of
slaughter, being ﬁrst experimented and exercised in these, at last passed
even to the laboring ox, and the sheep that clothes us, and to the poor cock
that keeps the house; until by little and little, unsatiableness, being
strengthened by use, men came to the slaughter of men, to bloodshed and
wars (http://etext.libraQ/adelaide.edu.au/p/plutarch/essays/complete.html).
Additionally, during a tour of the infamous stockyards Sinclair wrote about, author and
poet Rudyard Kipling was hon'iﬁed by both the surroundings and the indifference
displayed by those witnessing the events. According to Cronon, this “indifference
seemed to Kipling the most frightening thing he saw at the stockyards, and made him
worry about the effect of so mechanical a killing house on the human soul” (1991, 208).
Further, as discussed in Chapter 1, connections between other types of
occupations that involve violence (such as military service) and crime (particularly
interpersonal violence) have been documented through research. It is not a large leap to
hypothesize that there might be a connection between the increased crime rates in
slaughterhouse communities and the type of work conducted within the facilities. After
all, the few interviews conducted with slaughterhouse workers point in that direction.
Recall the statements from Gail Eisnitz’s (1997) informants cited in Chapter 1, where
workers describe becoming ‘emotionally dead’ and ‘sadistic’. Even anthropologist
Deborah Fink (1998), who worked in a slaughterhouse for ﬁve months, reported feeling
depressed and suicidal.
Although this study cannot directly assess whether or not there is a connection
between the elevated crime rates in slaughterhouse communities and the work conducted

within the walls of slaughterhouses, there are obvious reasons to wonder and to brieﬂy

speculate about some possible reasons that these facilities might have negative spill-over

197

effects into the community. In a Meadian sense, perhaps certain interactions that
transpire in these facilities are especially problematic and have particular symbolic
values. For instance, perhaps the ‘stickers’ who are responsible for slitting the throats of
the animals at the beginning of the (dis)assembly line have particularly traumatic
experiences. If Eisnitz’s (1997) assessment of the increasing number of animals that are
improperly stunned and travel down the (dis)assembly line while still conscious is
accurate, a growing number of workers might be involved in very troubling interactions
with dying and partially dismembered animals.

From more of a Marxist slant, perhaps this form of production results in a special
form of alienation of the workers. Cronon’s comments about the historic development of
a separation between people and the animals they consume is telling in this regard: "The
growing distance between the meat market and the animals in whose ﬂesh it dealt may
have seemed civilizing to those who visited the Exchange Building in the 18605, but it
also betokened a much deeper and subtler separation —- the word ‘alienation’ is not too
strong — from the act of killing and from nature itself" (Cronon 1991, 212-213). The
alienation of the public from the act of slaughter may also result in the alienation of
slaughterhouse workers from the larger society and from the product of their labor. As
society has distanced itself from the production of meat and slaughterhouses have
consolidated, the work of killing and dismembering animals has been relegated to fewer
and fewer people. Largely gone are the days of keeping a cow on a family farm and
killing her for meat once milk production slows. Slaughterhouse workers have become
responsible for killing and dismembering hundreds to thousands of animals a day. The

consequences of this burden could be profound.

198

It is also possible that there is a mediating factor between slaughterhouse
employment and crime. Broadway (1990; 2000; 2001) has speculated that increased
alcohol consumption in slaughterhouse communities could be related to the increased
crime rates. It is conceivable that as a result of the interactions and alienation faced by
slaughterhouse workers, some turn to alcohol. All of these possibilities, however, await
further investigation. Indeed, much additional research is needed, and recormnended

avenues are discussed below.

Recommendations

The results of this study point to some general recommendations that can be made
(further research is needed before more speciﬁc recommendations can be articulated).

Three key recommendations are discussed below

Holding Corporations Responsible

This study provides evidence contrary to the claim that the ‘type of people’ employed in
slaughterhouses are the problem. As mentioned earlier, in some slaughterhouse
communities the immigrant workers have been scapegoated in explaining the increased
crime rates. Instead, the results of this study demonstrate that there is something unique
about the effects of slaughterhouse employment on crime rates and that the characteristics

of the workers cannot explain these effects.

199

The companies proﬁting from this industry need to be held responsible for the
extemalities they create in these communities. Further, communities and their citizens,
often desperate to attract businesses and jobs, need to be made aware of the potential
negative impacts of this particular industry; and the practice of offering tax incentives to
these companies, especially when the social costs are likely to be high, should be
reconsidered. Increasing attention is being paid to holding companies responsible for the
health and safety of workers in these plants. For instance, in 2005 for the ﬁrst time ever
Human Rights Watch issued a report criticizing a speciﬁc industry — the meatpacking
industry — for violating basic human rights. However, more attention needs to be paid to

the social consequences of this industry as well.

Educating Consumers

A less popular recommendation would be to hold the consumers of meat responsible for
creating such a high demand for a product that has been widely recognized as having
problematic environmental and health consequences, and as this study shows, has
problematic social consequences as well. David Nibert (2002) has clearly articulated the
connections between the consumption of meat in developed nations (especially the US)
and oppression and devastation in developing countries, where food animals are
increasingly being raised. Bringing such far-removed consequences of consumption into
the consciousness of consumers in the developed world can be very difﬁcult. This study,
however, points to social consequences literally in the backyards of many consumers.

These consequences need to be brought to the attention of the consumers, who should be

200

encouraged to make more sustainable and socially responsible dietary choices. The
director of the US programs for Human Rights Watch has stated, “Every country has its
horrors, and this industry is one of the horrors in the United States” (cited in Greenhouse

2005). Consumers need to be made more aware of these horrors.

Further Research

As mentioned earlier, this study is but one step in ﬁlling the gap in the substantive
literature on slaughterhouse communities. As such a ﬁrst step, it points us in the
direction where future research is warranted. First of all, this study was delimited in a
few important ways: it focused on non-metropolitan counties in states with right-to-work
laws. Future research might include counties categorized as metropolitan or adjacent to
metropolitan counties in order to examine if the degree of urbanization has an impact on
the social effects of slaughterhouses. More interesting research, however, might examine
if the effects of slaughterhouses are different in states without right-to-work laws than in
the states examined in this study that do have right-to-work laws, thus facilitating an
examination of whether or not labor union organization can mitigate some of the social
consequences of slaughterhouses.

This study was also limited to slaughterhouses that kill and process quadruped
animals (cows, pigs, horses). Subsequent research might compare the effects of these
slaughterhouse facilities with those that kill poultry (and perhaps even ﬁsh). It is possible

that facilities that slaughter animals lower on the sociozoologic scale (a term coined by

20l

Arluke and Sanders (1996) to describe the socially constructed hierarchy of animals that
exists in human cultures) than quadruped animals would not have the same degree of
community consequences, despite similarities in the bodies being processed, the type of
work conducted, and the demographic characteristics of the workers, because killing and
dismembering their bodies might not have the same signiﬁcance for the workers as those
engaged in slaughtering and dismembering large quadruped animals.

Although some work has been conducted examining the effects of increasing
consolidation in the meatpacking industry in Canada (see Broadway 1998), it might also
be interesting to compare the explanatory power of the variables proposed in the
literature (those used as control variables in this study) in the Canadian context with those
in the US. Controlling for the other variables, would we see the same increase in the
arrest and report scales, in total arrests, arrests for violent offences, rape arrests, and
arrests for sexual offences in Canadian slaughterhouse communities? Or perhaps due to
differences in the categorization of offences in Canada there would be increases in other
official crime rates, such as family violence? These results might also be compared with
those in other countries which have not experienced the geographic Shift in meat
production closer to livestock supplies, such as in Britain (Broadway 2002). Such
research would enable us to answer the following question: Do communities with
longstanding slaughter facilities experience the same social consequences of slaughtering
as communities with relatively new slaughterhouse facilities in Canada and the US? This
could prove important because afﬁrmative evidence would lend support to the thesis that
the type of work conducted in slaughterhouses is at least partially responsible for the

increased crime rates observed in the US and Canada.

202

This study has demonstrated that controlling for several key variables, levels of
slaughterhouse employment in the counties studied are signiﬁcantly positively related to
the arrest and report scales created, arrests for rape and sex offences over the entire study
period, and total arrests and arrests for violent offences before the slaughterhouse
categorization changed to include ‘custom slaughter’ facilities. These ﬁndings, however,
do not prove the thesis that the type of work conducted in slaughterhouses is at least
partially responsible for the increased crime rates observed in the US and Canada.
Rather, it demonstrates that this is a possibility that can no longer be ignored since the
theories that have been proposed in the literature do not explain the crime increases. In
order to explicitly test this theory more individual-level research is required. This could
involve interviewing and/or surveying slaughterhouse workers (and comparison groups,
such as those working in similar types of manufacturing industries and others employed
in occupations that involve violence, such as the military, police, and corrections) about
their own experiences. It would be interesting to examine if the type of job conducted in
the slaughterhouse is particularly important.

It would also be interesting to examine what role gender plays, especially since
this research has demonstrated that rape and other sex crimes are particularly related to
slaughterhouse employment. (As of 2003, slightly more than a quarter (26.6%) of meat,
poultry, and ﬁsh possessing workers in the US were women (US Census Bureau 2003)).
Are women and men affected differently by employment in slaughterhouses?
Additionally, at least in Canada and the US, some incarcerated populations are employed
in slaughterhouses. What might the effects be of taking already at-risk populations and

exposing them to a type of violent work that might have spill-over effects even among

203

the general public? Clearly, there are many interesting avenues that remain to be

explored.

In coming to the end of the avenue pursued in this study it is helpful to return to
the ﬁrst glimpse provided into the world of slaughterhouse labor, Sinclair’s The Jungle,
as well as the most comprehensive analysis of the community effects of slaughterhouses
to date, Slaughterhouse Blues (2004) by Donald Stull and Michael Broadway. In the
forward to Slaughterhouse Blues Eric Schlosser remarks that our system of meat
production is unique because “[i]t imposes enormous costs upon society that are not
reﬂected in the price of meat.” This study provided an examination of the least readily
explainable of these costs — increases in crime — and demonstrated that compared to
similar industries the effect of slaughterhouses is problematic and unique.

This study also demonstrated that what Stull and Broadway characterize as the
“slaughterhouse blues” in the title of their book cannot be explained away by the
demographic, economic, and social disorganization theories proposed in the literature.
Regarding the “slaughterhouse blues” Broadway states, “the reality is that the Blues
cannot be avoided. They can, however, be ameliorated to some degree by a combination
of enlightened institutional and individual responses to social change” (Broadway 2000,
45). The results of this study indicate that given the current state of production Broadway
is likely correct: ‘the Blues’ cannot be avoided. However, this study also casts some
doubt on the degree to which the manifestation of ‘the blues’ in the form of increased

crime rates can be ameliorated by different responses to the social change experienced in

204

these communities. As in Sinclair’s time one hundred years ago, the problem may be that
‘the blues’ are inherent to the type of work conducted within the walls of
slaughterhouses; that is, ‘the blues’ and the industrialized ‘jungle’ might necessarily go

hand-in-hand.

205

IO

10.

11.

12.

NOTES

By ‘general’ crime rate I am referring to aggregate crime categories, such as the total number of
arrests. In contrast, ‘speciﬁc’ crime refers to individual crime variables, such as arrests for
murder. The bulk of the research thus far on slaughterhouse communities has focused on general
crime rates, most speciﬁcally the total number of arrests.

Other changes in the monopoly structure of the meatpacking industry are worth noting: In 2002,
ConAgra sold the majority of the interest in its Red Meat division and it was renamed Swiﬂ &
Company; and in 2003, the beef and hog processing operation of Farmland Industries were sold to
the world's largest hog producer and processor, Smithﬁeld Foods (Broadway and Stull 2005).

In Britain, slaughtering has not become as concentrated in livestock producing areas as it has in
Canada and the United States. Broadway (2002) speculates that this may be due in part to
slaughterhouses locating in areas where they can obtain government grants, livestock holdings are
less concentrated in Britain in the ﬁrst place, and due to environmental concerns large feedlots are
not as common as they are in Canada and the United States.

Schlosser (2005[2001]) contends that demand by the fast food industry for cheap meat is
responsible, at least in part, for the increased line speeds. He also suggests that a demand by the
Mc Donald’s corporation for slower line speeds would have a signiﬁcant industry-wide effect.

It should be noted that the immigrant workers are able to exercise some degree of agency in these
plants (Grey 1999), although it is certainly constrained.

Boomtown communities are characterized by the following features: they experience
unprecedented population growth within a short amount of time; relatedly, they experience
expanded employment opportunities; and they also experience heavy demands on social services
(Camasso and Wilkinson 1990).

Shaw and McKay considered subcultures an important aspect of neighborhood organization, but
later works focused more on the structural factors, downplaying the cultural, and emphasized the
role of social control. However, the cultural factors have received renewed attention as of late
(Kubrin and Weitzer 2003; Nielsen, Lee, and Martinez 2005; Rice and Smith 2002). Some have
speculated that the weakening of community-based institutions might lead to a view of the
institutions, such as the police, as lacking legitimacy, opening the way for cultural adaptations
(Triplett, Gainey, and Sun 2003).

The deﬁnition of ‘community’ varies according to the source of the deﬁnition and the time period
within which it was put forth (Hillery 1955; Rubin 1969; Wilkinson 1991).

F ink ( 1998) and Schlosser (2005[2001]) both report signiﬁcant sexual harassment of female
workers in the plants.

In the following discussion of the control variables they are described as counts instead of rates.
This is because the statistical techniques described later in the chapter use the counts to convert
them to rates.

As outlined in Appendix B, the County Business Patterns employment data are grouped into
categories, such as 20-99 workers. For the purposes of this study, these categories were assigned
the approximate middle value, such as a value of 60 for the 20-99 category.

1 would have also liked to examine the ﬂuctuations in crime over time in counties where
slaughterhouses open, i.e., the effects one year after opening, two years after opening, etc.

206

13.

14.

15.

16.

Unfortunately, however, such an analysis was not feasible since there were only ﬁve counties in
the sample where large slaughterhouses opened during the period under study.

Since crime rates are fairly stochastic, analyses including them can result in extreme cases arising,
which can exert undue inﬂuence over the analyses. Factor and Principal Components analyses are
particularly susceptible to outliers since they are based on correlations and covariances.
Accordingly, the extreme outliers (approximately eight cases for each variable) were removed
from the variables before the factor and principal components analyses were conducted and were
excluded from the scales created.

The outliers were identiﬁed by graphically plotting the distribution of the variables. The number
of cases removed and the threshold value for removing the cases are outlined for each variable
included in the scales below:

Arrest Rates:

Rape — excluded 6 cases with values greater than 99.

Robbery — excluded 6 cases with values greater than 124.

Burglary — excluded 6 cases with values greater than 500.

Forgery — excluded 9 cases with values greater than 400.

Stolen Property — excluded 7 cases with values greater than 400.

Vandalism — excluded 1 1 cases with values greater than 400.

Other assaults — excluded 12 cases with values greater than 1,499.

Offenses against the family — excluded 3 cases with values greater than 999.
Disorderly conduct — excluded 12 cases with values greater than 1,499.

Report Rates:

Rape — excluded 4 cases with values greater than 150.

Robbery - excluded 12 cases with values greater than 225.

Assault - excluded 5 cases with values greater than 1,500.

Burglary — excluded 13 cases with values greater than 2,200.

Motor vehicle theft — excluded 7 cases with values greater than 500.
Arson — excluded 4 cases with values greater than 200.

It is necessary to note that there is some degree of multicolinearity among the variables.
Speciﬁcally, the total number of males, number of young males, and the number of people in
poverty have VlF values greater than 4 (the values are 19.25, 15.64, and 8.01 respectively). Since
this colinearity is entirely among control variables it has no important effect on the estimates of
the effects of slaughterhouse employment (the VIP value of the lagged slaughterhouse
employment variable is 1.47).

1 am aware of Allison and Waterman’s (2002) suggestion that dummy variables for the units be
used to model the ﬁxed effect instead of using the ﬁxed effects option in Stata. However, due to
excessive multi-collinearity introduced by the county dummy variables (with VlF values as high
as 825) this technique could not be employed with these data.

The results from the TSCS negative binomial regression discussed here are taken from the
analysis of the entire time period, since it entailed analyzing more cases and resulted in more of
the variables reaching the threshold of signiﬁcance. However, it is worth noting that due to the
large sample size used in these analyses, as well as the analyses using the arrest rate and report
rate scales as the dependent variables, some of the statistically signiﬁcant coefﬁcients are
substantively rather small.

207

APPENDIX A

Counties Included in the Study:

STATE

Alabama
Alabama
Alabama
Alabama
Alabama
Alabama
Alabama
Arizona

Arkansas
Arkansas
Arkansas
Arkansas
Arkansas
Arkansas
Arkansas
Arkansas
Arkansas
Arkansas
Arkansas
Arkansas
Arkansas
Arkansas
Arkansas
Arkansas
Arkansas
Arkansas
Arkansas
Arkansas
Arkansas
Arkansas
Arkansas
Arkansas
Arkansas
Arkansas
Arkansas
Arkansas
Arkansas
Georgia

COUNTY NAME
Choctaw
Clarke
Conecuh
Covington
Lamar
Marengo
Monroe
Greenlee
Ashley
Baxter
Boone
Calhoun
Chicot
Clark
Clay
Columbia
Drew
Fuhon
Howard
Independence
lzard
Marion
Monroe
Nevada
Newton
Ouachita
Phillips
Pike

Polk

Pope
Randolph
Searcy
Sevier
Sharp
Stone
Union
Woodruff

Appling

208

Georgia
Georgia
Georgia
Georgia
Georgia
Georgia
Georgia
Georgia
Georgia
Georgia
Georgia
Georgia
Georgia
Georgia
Georgia
Georgia
Georgia
Georgia
Georgia
Georgia
Georgia
Georgia
Georgia
Georgia
Georgia
Idaho
ldaho
ldaho
ldaho
Idaho
Idaho
ldaho
ldaho
ldaho
ldaho
ldaho
Idaho
Idaho
lowa
Iowa
Iowa
Iowa
Iowa

Atkinson
Bacon
Ben Hill
Candler
Clay
Coffee
Dodge
Emanuel
Glascock
Hancock
lrwin

Jeff Davis
Johnson
Montgomery
Quitman
Rabun
Stephens
Telfair
Toombs
Towns
Treutlen
Union
Washington
Wheeler
Wilcox
Bear Lake
Blaine
Boundary
Camas
Cassia
Custer
Gooding
Jerome
Lemhi
Lincoln
Minidoka
Teton
Twin Falls
Adams
Appanoose
Buena Vista
Calhoun
Carroll

209

lowa
lowa
Iowa
Iowa
lowa
lowa
lowa
Iowa
Iowa
Iowa
lowa
lowa
lowa
lowa
Iowa
Iowa
Iowa
lowa
lowa
lowa
lowa
lowa
lowa
lowa
lowa
lowa
Iowa
Iowa
lowa
Iowa
Iowa
Iowa
lowa
lowa
lowa
Iowa
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas

Cerro Gordo
Clay

Davis
Decatur
Des Moines
Dickinson
Emmet
Floyd
Franklin
Hancook
Henry
Howard
Humboldt
Jefferson
Kossuth
Lee
Mahaska
Mitchell
Monroe
O'Brien
Osceola
Page

Palo Alto
Pocahontas
Poweshiek
Ringgold
Sac

Taylor
Van Buren
Wapello
Wayne
Webster
Winnebago
Winneshiek
Worth
Wright
Allen
Barber
Barton
Chautauqua
Cheyenne
Clark

Clay

Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas

Cloud
Comanche
Decatur
Dickinson
Edwards
Ellis
Ellsworth
Finney
Ford
Geary
Gove
Graham
Grant
Gray
Greely
Hamilton
Haskell
Hodgeman
Jewell
Keamy
Kiowa
Labette
Lane
Lincoln
Logan
Lyon
Marshall
Meade
hdhcheﬂ
Montgomery
Morris
Morton
Neosho
Ness
Norton
Osborne
Ottawa
Pawnee
Phillips
Pratt
Rawlins
Republic
Rice

211

Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Kansas
Louisiana
Louisiana
Louisiana
Louisiana
Louisiana
Louisiana
Louisiana
Louisiana
Mississippi
Mississippi
Mississippi
Mississippi
Mississippi
Mississippi
Mississippi
Mississippi
Mississippi
Mississippi
Mississippi
Mississippi
Mississippi
Mississippi
Mississippi

Riley
Rooks
Rush
Russell
Saline
Scott
Seward
Sheridan
Sherman
Smith
Stafford
Stanton
Stevens
Thomas
Trego
Wallace
Washington
Wichita
Wilson
Woodson
Catahoula
Claiborne
Concordia
East Carroll
Franklin
Madison
Tensas
West Carroll
Adams
Alcom
Bolivar
Calhoun
Carroll
Chickasaw
Choctaw
Clarke
Clay
Coahoma
Franklin
Grenada
Humphreys
lssaquena
ltawamba

212

Mississippi
Mississippi
Mississippi
Mississippi
Mississippi
Mississippi
Mississippi
Mississippi
Mississippi
Mississippi
Mississippi
Mississippi
Mississippi
Mississippi
Mississippi
Mississippi
Mississippi
Mississippi
Mississippi
Mississippi
Mississippi
Mississippi
Mississippi
Mississippi
Mississippi
Mississippi
Mississippi
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska

Jasper
Jefferson
Kemper
Lauderdale
Lee
Leﬂore
Lowndes
Monroe
Montgomery
Neshoba
Newton
Noxubee
Oktibbeha
Pike
Pontotoc
Prentiss
Sharkey
Sunﬂower
Tallahatchie
Tippah
Union
Walthall
Washington
Wayne
Webster
Winston
Yalobusha
Adams
Antelope
Arthur
Banner
Blaine
Boone
Box Butte
Boyd
Brown
Buffalo
Cedar
Chase
Cherry
Cheyenne
Clay
Colfax

213

Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska

Cuming
Custer
Dawes
Dawson
Deuel
Dundy
Fillmore
Franklin
Frontier
Fumas
Garden
Garﬁeld
Gosper
Grant
Greeley
Hall
Hamilton
Harlan
Hayes
Hitchcock
Holt
Hooker
Howard
Jefferson
Kearney
Keith
Keya Paha
Knox
Lincoln
Logan
Loup
Madison
McPherson
Merrick
Morrill
Nance
Nemaha
Nuckolls
Pawnee
Perkins
Phelps
Pierce
Platte

214

Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nebraska
Nevada
Nevada
Nevada
Nevada
Nevada
Nevada
Nevada

North Carolina
North Carolina
North Carolina
North Carolina
North Carolina
North Carolina
North Carolina
North Carolina
North Carolina
North Carolina
North Carolina
North Carolina
North Carolina
North Carolina
North Carolina
North Carolina
North Carolina
North Carolina
North Carolina
North Dakota
North Dakota

Polk

Red Willow
Richardson
Rock
Scotts Bluff
Sheridan
Sherman
Sioux
Stanton
Thayer
Thomas
Valley
Webster
Wheeler
York

Elko
Esmeralda
Eureka
Humboldt
Lander
Mineral
White Pine
Alleghany
Ashe
Bertie
Cherokee
Chowan
Clay
Craven
Dare
Graham
Hertford
Hyde
Macon
Mitchell
North Hampton
Pamlico
Pasquotank
Perquimans
Tyrrell
Washington
Adams
Benson

215

North Dakota
North Dakota
North Dakota
North Dakota
North Dakota
North Dakota
North Dakota
North Dakota
North Dakota
North Dakota
North Dakota
North Dakota
North Dakota
North Dakota
North Dakota
North Dakota
North Dakota
North Dakota
North Dakota
North Dakota
North Dakota
North Dakota
North Dakota
North Dakota
North Dakota
North Dakota
North Dakota
North Dakota
North Dakota
North Dakota
North Dakota
North Dakota
North Dakota
Oklahoma

Oklahoma

Oklahoma

Oklahoma

Oklahoma

Oklahoma

Oklahoma

Oklahoma

Oklahoma

Oklahoma

Billings
Bottineau
Bowman
Burke
Cavalier
Dickey
Divide
Dunn
Eddy
Foster
Golden Valley
Griggs
Hettinger
LaMoure
Logan
McHenry
McIntosh
McKenzie
Mountrail
Pembina
Pierce
Ramsey
Renville
Rolette
Sargent
Sheridan
Slope
Stark
Stutsman
Towner
Ward
Wells
Williams
Alfalfa
Atoka
Beaver
Beckham
Carter
Choctaw
Cimarron
Coal
Custer
Dewey

216

Oklahoma
Oklahoma
Oklahoma
Oklahoma
Oklahoma
Oklahoma
Oklahoma
Oklahoma
Oklahoma
Oklahoma
Oklahoma
Oklahoma
Oklahoma
Oklahoma
Oklahoma
Oklahoma
Oklahoma
Oklahoma
Oklahoma
Oklahoma
Oklahoma
Oklahoma
Oklahoma
South Carolina
South Carolina
South Dakota
South Dakota
South Dakota
South Dakota
South Dakota
South Dakota
South Dakota
South Dakota
South Dakota
South Dakota
South Dakota
South Dakota
South Dakota
South Dakota
South Dakota
South Dakota
South Dakota
South Dakota

Ellis
Garﬁeld
Grant
Greer
Harmon
Harper
Hughes
Jackson
Johnston
Kay
Latimer
Love
Major
McCurtain
Murray
Pittsburg
Pontotoc
Roger Mills
Seminole
Texas
Washita
Woods
Woodward
Bamberg
Beaufort
Aurora
Beadle
Bennett
Bon Homme
Brookings
Brown
Brule
Buffalo
Campbell
Charles Mix
Clarke
Codington
Corson
Day
Davison
Deuel
Dewey
Douglas

217

South Dakota
South Dakota
South Dakota
South Dakota
South Dakota
South Dakota
South Dakota
South Dakota
South Dakota
South Dakota
South Dakota
South Dakota
South Dakota
South Dakota
South Dakota
South Dakota
South Dakota
South Dakota
South Dakota
South Dakota
South Dakota
South Dakota
South Dakota
South Dakota
South Dakota
South Dakota
South Dakota
South Dakota
South Dakota
Tennessee
Tennessee
Tennessee
Tennessee
Tennessee
Tennessee
Tennessee
Tennessee
Tennessee
Tennessee
Tennessee
Tennessee
Tennessee
Tennessee

Edmunds
Fall River
F aulk
Grant
Gregory
Hamlin
Hand
Harding
Hughes
Jerauld
Jones
Kingsbury
Lyman
McPherson
Marshall
Mellette
Perkins
Potter
Roberts
Sanbom
Shannon
Spink
Stanley
Sully
Todd
Tripp
Walworth
Yankton
Ziebach
Benton
Cumberland
Decatur
Dyer
Fentress
Henry
Lake
Moore
Obion
Overton
Pickett
Van Buren
Weakley
White

218

Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas

Anderson
Angelina
Bailey
Borden
Brewster
Briscoe
Brown
Childress
Cochran
Collingsworth
Comanche
Cottle
Crockett
Culberson
Dallam
Dawson
Duval
Edwards
Erath
Foard
Freestone
Gaines
Gillespie
Hall
Hansford
Hardeman
Hartley
Hemphill
Houston
Howard
Jeff Davis
Kenedy
Kent
Kimble
Kingsbury
Kinney
Knox
Lipscomb
Llano
Loving
Mason
Maverick
McCulloch

219

Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Texas
Utah
Utah
Utah
Utah
Utah
Utah
Utah
Utah
Utah
Utah
Utah
Virginia
Virginia
Virginia
Virginia
Virginia
Virginia
Virginia
Virginia

Mills
Mitchell
Nacogdoches
Ochiltree
Parmer
Pecos
Presidio
Real
Reeves
Roberts
Sabine
San Augustine
San Saba
Scurry
Sherman
Stephens
Sunon
Terrell
Throckmorton
Titus
Uvalde
Val Verde
Yoakum
Zavala
Beaver
Carbon
Emery
Garﬁeld
Grand
Millard
Piute

San Juan
Sevier
Uintah
Wayne
Accomack
Bath
Buchanan
Dickenson
Grayson
Highland
Lancaster
Lunenburg

220

Virginia

Virginia

Virginia

Virginia

Virginia

Virginia

Virginia

Virginia

Wyoming
Wyoming
Wyoming
Wyoming
Wyoming
Wyoming
Wyoming
Wyoming
Wyoming
Wyoming
Wyoming
Wyoming
Wyoming
Wyoming
Wyoming
Wyoming
Wyoming
Wyoming
Wyoming

Mecklenburg
Northhampton
Northumberland
Richmond
Tazewell
Westmoreland
Wise

Norton City
Big Horn
Campbell
Carbon

Crook
Freemont
Goshen

Hot Springs
Johnson
Lincoln
Niobrara

Park

Platte
Sheridan
Sublette
Sweetwater
Teton

Uinta
Washakie
Weston

221

APPENDIX B

 

 

 

 

 

 

 

 

 

 

Variable Operationaliz— Source Measure- Mean (or Standard
Name ation ment percent- Deviation
ages in the
case of
nominal
variables)
INDEPENDENT VARIABLES
Shw Number of County 0 = 0 64.26 402.36
slaughterhouse Business 10=1-19;
workers in county Patterns 60:20-99;
175=100-249;
375=250-499;
750=500-999;
1750:1000-
2499;
3750=2500-4999
7500 = 5000+
She Number of County 0, l, 2..., N 0.37 0.65
slaughterhouse Business
establishments in Patterns
county
lrw Number of iron and County 10=l-l9; 0.95 15.95
steel forging Business 60:20-99;
workers in county Patterns 175=100-249;
375=250-499;
750:500-999;
1750:1000-
2499;
3750=2500-4999
lre Number of iron and County 0, 1, 2..., N 0.01 0.09
steel forging Business
establishments in Patterns
coun
Tm Number of truck County 10=1—19; 5.76 45.25
trailer Business 60:20-99;
manufacturing Patterns 175=100-249:
workers in county 375=250~499;
750=500-999;
1750=1000-
2499;
3750=2500-4999
Tre Number of truck County 0, l, 2..., N 0.06 0.28
trailer Business
manufacturing Patterns
establishments in
ceiﬂty
Mow Number of motor County 10=l-l9; 0.96 14.3

 

 

 

 

 

 

222

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

vehicle metal Business 60:20-99;
stamping workers Patterns 175=100-249;
in county 375=250-499;
750=500-999;
1750=1000-
2499;
3750:2500-4999
Moe Number of motor County 0, l. 2..., N 0.01 0.09
vehicle metal Business
stamping Patterns
establishments in
coun
Siw Number of sign County 10:1-19; 5.27 52.64
manufacturing Business 60=20-99;
workers in county Patterns 175=100-249;
375=250-499:
750=500-999:
1750:1000-
2499;
3750:2500-4999
Sie Number ofsign County 0, 1. 2..., N 0.19 0.61
manufacturing Business
establishments in Patterns
county
Law Number of County 10: l - l 9; 2.73 13.65
industrial launderer Business 60:20-99;
workers in county Patterns 175=100-249;
375:250-499:
750:500-999;
1750:1000-
2499;
3750:2500-4999
Lae Number of County 0. 1, 2..., N 0.08 0.31
industrial launderer Business
establishments in Patterns
cow
CONTROL VARIABLES
Povno Number of people Small 0, l, 2..., N 2239.99 2548.04
in the county in Area
poverty lncome
(determined and
according to the Poverty
characteristics of Estimates.
the household, such US.
as income, age of Census
residents, etc.) Bureau
Une Annual Local 0% - 100% 5.02 3.06
unemployment rate Area
(the number of Unemploy
unemployed as a ment
percent of the labor Statistics,
force), not Bureau of
seasonally Labor
adjusted, for the Statistics
county

 

 

 

 

223

 

 

Pop

Total county
population

Populatio
n
Estimates
(US
Census
Bureau)

13859.88

14010.09

 

lmm

Net international
immigration into
the county

Populatio
n
Estimates
(US
Census
Bureau)

-N..., -l, 0,1...,

15.95

47.48

 

Net internal
migration into the
county

Populatio
n
Estimates
(US
Census
Bureau)

-N..., -l, 0,1...,

-33.84

260.96

 

Tom

Total number of
males in the county

Adapted
from
several
separate
categories
from
Populatio
n
Estimates
(US
Census
Bureau)

0,1, 2..., N

6816.09

6908.21

 

Popden

Population density
(number of people
per square mile)

Calculated
by
dividing
the total
population
by area

0,1,2..., N

22.2

32.09

 

TNwH

Total number of
non-whites and/or
Hispanics

Calculated
by
subtractin
g the
number of
non-
Hispanic
white
males and
non-
Hispanic
white
females
from the
total
population

0,1, 2..., N

3077.25

5449.53

 

 

YngM

 

Young men age 15
to 34

 

Calculated
by
summing
the

 

O, l,2...,N

 

1988.33

 

2394.04

 

224

 

 

following

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

categories
: males
age 15-19,
males age
20—24,
males age
25-29.
males age
30-34
DEPENDENT VARIABLES
Tot Total number of Uniform 0, l. 2 , N 592.67 856.02
arrests. Includes Crime (rate = (for rate =
non-Index and Reports 3348.05 / 2605.23)
Index offenses 100,000)
Vio Number of arrests Uniform 0, I, 2.... N 18.43 32.04
for violent crimes. Crime (rate = 99.07) (for rate =
Sum of variables Reports 1 I 1.06)
Murder through
Aggravated Assault
Mur Number arrests for Uniform 0, 1, 2 , N 0.64 2.08
murder Crime (rate = 3.56) (for rate =
Reports I 1.64)
Rap Number arrests for Uniform 0, l, 2..., N 1.18 2.28
rape Crime (rate = 6.44) (for rate =
Reports 1 1.76)
Rob Number of arrests Uniform 0, I, 2..., N 1.91 4.59
for robbery Crime (rate = 8.3) (for rate =
Reports 15.95)
AgA Number of arrests Uniform 0, 1, 2.... N 14.66 26.29
for aggravated Crime (rate = 80.21) (for rate =
assault Repoirts 94.96)
Bur Number of arrest Uniform 0, I. 2 , N l 1.16 18.09
for burglary Crime (rate = 64.21) (for rate =
Reports 71.05)
OthA Number ofarrests Uniform 0, 1, 2..., N 56.38 99.17
for other types of Crime (rate = 289.45) (for rate =
assault lgports 283.35)
For Number of arrests Uniform 0, l, 2..., N 6.67 16.16
for forgery Crime (rate = 31.75) (for rate =
Reports 48.33)
St] Number ofarrests Uniform 0, I, 2..., N 3.15 7.1
for possessing Crime (rate = 17.36) (for rate =
stolen property Reports 39.31)
Van Number of arrests Uniform 0, l, 2.... N 6.64 15.95
for vandalism Crime (rate = 35.49) (for rate =
Reports 56.52)
Sex Number of arrests Uniform 0. 1, 2 , N 2.47 5.61
for sex offenses, Crime (rate = 14.74) (for rate =
excluding rape and Reports 23.63)
prostitution
Fam Number ofarrests Uniform 0, 1, 2 , N 8.75 25.9
for ‘offenses Crime (rate = 48.87) (for rate =
against the family Reports 92.21)

 

and child’

 

 

 

 

 

 

225

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Dis Number of arrests Uniform 0, I, 2..., N 25.06 49.79
for disorderly Crime (rate = 134.18) (for rate =
conduct Reports 193.69)
InR Total number of Uniform 0, l, 2. . ., N 380.26 709.51
reports of Index Crime (rate = 1827.9) (for rate =
crimes, excluding Reports 1611.92)
arson. The sum of
variables Murder
through Motor
Vehicle theft
MurR Number of Uniform 0, l, 2..., N 0.56 1.4
reported murders Crime (rate = 3.25) (for rate =
Reports 8.82)
RapR Number of Uniform O, l, 2..., N 2.96 6.67
reported rapes Crime (rate = 14) (for rate
Reports =20.88)
RobR Number of Uniform 0, 1, 2..., N 4.42 14.14
reported robberies Crime (rate = 15.74) (for rate =
Reports 31.55)
AssR Number of Uniform 0, 1, 2..., N 28.47 59.51
reported Crime (rate = 142.1) (for rate =
aggravated assaults Reports 181 .13L
BurR Number of Uniform 0, 1, 2..., N 84.53 160.48
reported burglaries Crime (rate = 429.53) (for rate =
Reports 390.19)
MvR Number of Uniform 0, 1, 2..., N 18.09 36.26
reported motor Crime (rate = 87.83) (for rate =
vehicle theﬁs Reports 88.16)
ArsR Number of Uniform 0, 1, 2..., N 2.15 4.94
reported arsons Crime (rate = 1 1.83) (for rate =
Reports 22.51)

 

226

 

REFERENCES

Adams, Carol. 1991. The Sexual Politics of Meat: A F eminist- Vegetarian Critical
Theory. New York: Continuum.

Albrecht, Stan L. 1982. "Commentary." Pacific Sociological Review 25: 297-306.

Allen, Leana C. 2000. "The Inﬂuence of Military Training and Combat Experience on
Domestic Violence." Pp. 81-103 in Battle Cries on the Homeﬁont, edited by P.
Mercier and J. Mercier. Springﬁeld: Charles C. Thomas.

Allison, Paul D. and Richard P. Waterman. 2002. "Fixed-Effects Negative Binomial
Regression Models." Sociological Methodology 32: 247-265.

Arluke, Arnold and Clinton R. Sanders. 1996. Regarding Animals. Philadelphia: Temple
University Press.

Bacon, David. 1999. "Ethnic Cleansing Hits Immigrant Workers, Organizers, in Midwest
Meatpacking: INS Declares War on Labor." The Nation 268: 18-23.

Baron, Larry and Murray A. Straus. 1988. "Cultural and Economic Sources of Homicide
in the United States." The Sociological Quarterly.

Bataille, Georges. 1997. "Slaughterhouse." Pp. 22 in Rethinking Architecture: A Reader
in Cultural Theory, edited by N. Leach. New York: Routledge.

Beck, N. and J. N. Katz. 1996. "Nuisance vs. Substance: Specifying and Estimating
Time-Series-Cross-Section Models." Political Analysis 6: 1-36.

Beck, Nathaniel. 2001. "Time-Series-Cross-Section Data: What Have we Learned in the
Past Few Years?" Annual Review of Political Science 4: 271-293.

Beime, Piers. 2004. "From Animal Abuse to Interhuman Violence? A Critical Review of
the Progression Thesis." Society & Animals 12: 39-65.

Bender, Thomas. 1978. Community and Social Change in America. New Brunswick:
Rutgers University Press.

Ben'y, E.H., R.S. Krannich, and T. Greider. 1990. "A Longitudinal Analysis of
Neighboring in Rapidly Changing Rural Places." Journal of Rural Studies 6: 175-
186.

Bonanno, Alessandro, Lawrence Busch, William H. F riedland, Lourdes Gouveia, and
Enzo Mingione. 1994. From Columbus to ConAgra: The Globalization of
Agriculture and Food. Lawrence: University Press of Kansas.

227

Braverman, Harry. 1974/1998. Labor and Monopoly Capital: The Degradation of Work
in the Twentieth Century. New York: Monthly Review Press.

Broadway, Michael. 1990. "Meatpacking and its Social and Economic Consequences for
Garden City, Kansas in the 19803." Urban Anthropology 19: 321-344.

—. 1995. "From City to Countryside: Recent Changes in the Structure and Location of
the Meat- and Fish-Processing Industries." Pp. 17-40 in Any Way You Cut It:
Meat Processing and Small Town America, edited by D. D. Stull, M. Broadway,
and D. Grifﬁth. Lawrence: University Press of Kansas.

Broadway, Michael J. 1994. "What Happens when the Meatpackers Come to Town?"
Small Town 24: 24-28.

—. 1998. "Where's the Beef? The Integration of the Canadian and American
Beefpacking Industries." Prairie Forum 23: l9-30.

—. 2000. "Plarming for Change in Small Towns or Trying to Avoid the Slaughterhouse
Blues." Journal of Rural Studies 16: 37-46.

—. 2001. "Bad to the Bone: The Social Costs of Beefpacking's move to Rural Alberta."
Pp. 39-51 in Writing off the Rural West: Globalization, Governments, and the
Transformation of Rural Communities, edited by R. Epp and D. Whitson.
Edmonton: University of Alberta Press.

—. 2002. "The British Slaughtering Industry: A Dying Business." Geography 87: 268-
280.

Broadway, Michael J. and T. Ward. 1990. "Recent Changes in the Structure and Location
of the United States Meatpacking Industry." Geography 75: 76-79.

Broadway, Michael and Donald D. Stull. 2005. "Meat Processing and Garden City, KS:
Boom and Bust." Journal of Rural Studies.

Brueggemann, J and C Brown. 2003. "The Decline of Industrial Unionism in the
Meatpacking Industry: Event-Structure Analyses of Labor Unrest, 1946-1987."
Work and Occupations 30: 327-360.

Bureau of Labor Statistics, Local Area Unemployment Statistics. 2005. vol. 2005.

Bureau of Labor Statistics, US. Department of Labor. 2004. "Table SNROl: Highest
Incidence Rates of Total Nonfatal Occupational Injury and Illness Cases, Private
Industry, 2003." vol. 2005.

—. 2004. "Table SNR06: Highest Incidence Rates of Total Nonfatal Occupational Injury
Cases, Private Industry, 2003." vol. 2005.

228

Caldwell, Wayne J. 1998. "Land-Use Planning, the Environment, and Siting Intensive
Livestock Facilities in the let Century." Journal of Soil and Water Conservation
53: 102-106.

Camasso, M. J. and K. P. Wilkinson. 1990. "Severe Child Maltreatment in Ecological
Perspective: The Case of the Western Energy Boom." Journal of Social Service
Research 13: 1-18.

Cohen, Lawrence E. and Kenneth Land. 1987. "Age structure and crime symmetry versus
asymmetry and the projection of crime rates through the 19903." American
Sociological Review 52: 170-182.

Conley, Dalton and Kristen W. Springer. 2001. "Welfare state and infant mortality."
American Journal of Sociology 107: 768-897.

Cortes, Kalena. 2005. "Do immigrants beneﬁt from an increase in the Minimum Wage
rate? An analysis by immigrant industry concentration." University of California,
Berkeley.

Cronon, William. 1991. Nature ’3 Metropolis: Chicago and the Great West. New York;
London: WW Norton and Company.

Currie, J. and D. Thomas. 1995. "Does head start make a difference." American
Economic Review 85: 341-364.

Dalla, Rochelle L., Amy Ellis, and Sheran C. Cramer. 2005. "Immigration and Rural
America: Latinos' perceptions of work and residence in three meatpacking
communities." Community, Work & Family 8: 163-185.

Dee, Thomas S., David C. Grabowski, and Michael A. Morrisey. 2005. "Graduated driver
licensing and teen trafﬁc fatalities." Journal of Health Economics 24: 571-5 89.

DeLind, Laura. 1998. "Parma: A Story of Hog Hotels and Local Resistance." Pp. 23-38
in Pigs, Proﬁts, and Rural Communities, edited by K. M. Thu and E. P.
Durrenberger. Albany: State University of New York Press.

DeVellis, Robert F. 1991. Scale Development: Theory and Applications, vol. 26.
Newbury Park: Sage Publications.

Eisnitz, G. 1997. Slaughterhouse: The shocking story of greed, neglect, and inhumane
treatment inside the US. meat industry. New York: Prometheus.

F abrigar, Leandre R., Duane T. Wegener, Robert C. MacCallum, and Eric J. Strahan.
1999. "Evaluating the Use of Exploratory Factor Analysis in Psychological
Research." Psychological Methods 4: 272-199.

Fehn, Bruce. 1998. "African-American Women and the Struggle for Equality in the
Meatpacking Industry, 1940-1960." Journal of Women 's History 10: 45-69.

229

Fink, Deborah. 1998. Cutting into the Meatpacking Line: Workers and Change in the
Rural Midwest. Chapel Hill: University of North Carolina Press.

F inkel, Steven E. 1995. Causal Analysis with Panel Data. Thousand Oaks; London; New
Delhi: Sage Publications.

Finsterbusch, Kurt. 1982. "Commentary: Boomtown disruption thesis: Assessment of
current status." Paciﬁc Sociological Review 25: 307-322.

Firebaugh, G. and F. Beck. 1994. "Does economic growth beneﬁt the masses? Growth,
dependence, and welfare in the third world." American Sociological Review 59:
631-653.

Flynn, Clifton P. 2002. "Hunting and Illegal Violence against Humans and Other
Animals: Exploring the Relationship." Society and Animals 10: 137-154.

Fording, R. 1997. "The conditional effect of violence as a political tactic: Mass
insurgency, welfare generosity, and electoral context in the American states."
American Journal of Political Science 41: 1-29.

F reudenberg, William R. 1982. "Commentary: Balance and bias in boomtown research."
Pacific Sociological Review 25: 323-338.

Freudenberg, William R. and Robert E. Jones. 1991. "Criminal behavior and rapid
community growth: Examining the evidence." Rural Sociology 56: 619-645.

Freudenberg, WR. 1981. "Women and men in an energy boomtown: adjustment,
alienation and adaptation." Rural Sociology 46: 220-244.

—. 1986. "The density of acquaintanceship: an overlooked variable in community
research?" American Journal of Sociology 92: 27-63.

Gale, Richard P. 1982. "Commentary." Paciﬁc Sociological Review 25: 339-348.

Giles, M. and K. Hertz. 1994. "Racial threat and partisan identiﬁcation." American
Political Science Review 88: 317-326.

Gimbel, Cynthia and Alan Booth. 1994. "Why does military combat experience adversely
affect marital relations?" Journal of Marriage and the Family 56: 691-703.

Gold, Raymond L. 1982. "Commentary." Pacific Sociological Review 25: 349-356.

Goldschmidt, Walter. 1998. "The urbanization of rural America." Pp. 183—198 in Pigs,
Proﬁts, and Rural Communities, edited by K. M. Thu and E. P. Durrenberger.
Albany: State University of New York Press.

230

Gordon, Rachel A., Benjamin B. Lahey, Eriko Kawai, Rolf Loeber, Magda Stouthhamer-
Loeber, and David P. Farrington. 2004. "Antisocial Behavior and Youth Gang
Membership: Selection and Socialization." Criminology 42: 55-87.

Gouveia, Lourdes and Donald D. Stull. 1995. "Dances with Cows: Beefpacking's impact
on Garden City, Kansas and Lexington, Nebraska." Pp. 85-107 in Any Way You
Cut It: Meat Processing and Small-Town America, edited by D. D. Stull, M.
Broadway, and D. Grifﬁth. Lawrence: University of Kansas Press.

—. 1997. "Latino Immigrants, Meatpacking, and Rural Communities: A Case Study of
Lexington, Nebraska." Julian Samora Research Institute, Michigan State
University, East Lansing.

Gozdziak, EM and MN Bump. 2004. "Poultry, apples, and new immigrants in the rural
communities of the Shenandoah Valley: An ethnographic case study."
International Migration 42: 149-164.

Greenhouse, Steven. 2005. "Meat Packing Industry Criticized on Human Rights
Grounds." in New York Times. New York.

Grey, Mark. 1995. "Pork, poultry and newcomers in Storm Lake, Iowa." Pp. 109-127 in
Any Way You Cut It: Meat Processing and Small- Town America. Lawrence:
University Press of Kansas.

—. 1999. "Immigrants, migration and worker turnover at the Hog Pride Pork Packing
Plant." Human Organization 58: 16-27.

Grey, Mark A. 1998. "Meatpacking in Storm Lake, Iowa: A Community in Transition."
Pp. 57-72 in Pigs, Profits, and Rural Communities, edited by K. M. Thu and E. P.
Durrenberger. Albany: State University of New York Press.

Grey, Mark A. and Anne C. Woodrick. 2002. "Unofﬁcial Sister Cities: Meatpacking
Labor Migration between Villachuato, Mexico, and Marshalltown, Iowa." Human
Organization 61: 364-3 76.

Grieco, Elizabeth M. 2001. "The White Population: 2000."

Grinols, Earl L. and David B. Mustard. 2006. "Casinos, Crime, and Community Costs."
The Review of Economics and Statistics 88: 28-45.

Hagan, Jacqueline Maria. 2004. "Contextualizing immigrant labor market incorporation."
Work and Occupations 31: 407-423.

Hake, ER and MB King. 2002. "The Veblenian credit economy and the corporatization of
American meatpacking." Journal of Economic Issues 36: 495-505.

Halaby, Charles N. 2004. "Panel models in sociological research: Theory into practice."
Annual Review of Sociology 30: 507-544.

231

Halpem, Rick. 1997. Down on the Kill Floor: Black and White Workers in Chicago 's
Packinghouses, 1904-1954. Urbana; Chicago: University of Illinois Press.

Henson, Robin K. and J. Kyle Roberts. 2006. "Use of Exploratory Factor Analysis in
Published Research: Common Errors and Some Comment on Improved
Practice." Educational and Psychological Measurement 66: 393-416.

Hillery, G. A. 1955. "Deﬁnitions of community: Areas of agreement." Rural Sociology
55: Ill-123.

Horowitz, Roger. 1997. "'Where Men Will Not Work': Gender, Power, Space, and the
Sexual Division of Labor in America's Meatpacking Industry, 1890—1990."
Technology and Culture 38: 187-213.

—. 2005. "Book Review: Slaughterhouse Blues, by D. Stull and M. Broadway."
Enterprise & Society 6: 546-548.

Horowitz, Roger and Mark J. Miller. 1999. Immigrants in the Delmarva Poultry
Processing Industry: The changing face of Georgetown, Delaware, and
Environs. The Julian Samora Research Institute, Michigan State University, East
Lansing.

Hunter, Lori M., Richard S. Krannich, and Michael D. Smith. 2002. "Rural migration,
rapid growth, and fear of crime." Rural Sociology 67: 71-89.

ICPSR, Inter-university Consortium for Political and Social Research -. 2004. "Uniform
Crime Reporting Program Data [United States]: County-Level Detailed Arrest
and Offense Data, 2002." University of Michigan, Ann Arbor, MI.

Iowa State University Extension. 1998. Handbook for Creating Sustainable Multiethnic
F ood-Producing Communities. Ames: Iowa State University Extension.

J ablonsky, Thomas T. 1993. Pride in the Jungle: Community and Everyday Life in Back
of the Yards Chicago. Baltimore; London: The Johns Hopkins University Press.

Jacob, Brian A. and Lars Lefgren. 2003. "Are idle hands the devil's workshop?
Incapacitation, concentration, and juvenille crime." The American Economic
Review 93: 1560-1577.

Kalof, Linda. forthcoming. Looking at Animals. Reaktion.

Kalton, Graham and Constance F. Citro. 2000. "Panel Surveys: Adding the Fourth
Dimension." Pp. 36-53 in Researching Social and Economic Change: The Uses
of Household Panel Studies, edited by D. Rose. London; New York: Routledge.

Kawachi, Ichiro, Bruce P. Kennedy, and Richard G. Wilkinson. 1999. "Crime: Social
disorganization and relative depravation." Social Science and Medicine 48: 719-
731.

232

Kim, Jae-On and Charles W. Mueller. 1978. Introduction to Factor Analysis. Newbury
Park: Sage Publications.

King, Gary. 1997. A Solution to the Ecological Inference Problem: Reconstructing
Individual Behavior from Aggregate Data. Princeton: Princeton University Press.

Krannich, R.S., E.H. Berry, and T. Greider. 1989. "Fear of crime in rapidly changing
rural communities: A longitudinal analysis." Rural Sociology 54: 195-212.

Krivo, Lauren J. and Ruth D. Peterson. 2004. "Labor market conditions and violence
crime among youths and adults." Sociological Perspectives 47: 485-505.

Kubrin, Charis E. and Ronald Weitzer. 2003. "New directions in social disorganization
theory." Journal of Research in Crime and Delinquency 40: 374-402.

Leary, Mark. 1991. Introduction to Behavioral Research Methods. Belmont: Wadsworth.

Lee, Matthew T., Ramiro Martinez, and Richard Rosenﬁeld. 2001. "Does immigration

increase homicide? Negative evidence from three border cities." The Sociological
Quaterly 42: 559-580.

Lee, MR and GC Ousey. 2001. "Size Matters: Examining the link between small
manufacturing, socioeconomic deprivation, and crime rates in nonmetropolitan
communities." Sociological Quarterly 42: 581-602.

Long. J. Scott and Jeremy F reese. 2006. Regression Models for Categorical Dependent
Variables Using Stata. College Station, Texas: Stata Press.

Long, S. J. 1997. Regression Models for Categorical and Limited Dependent Variables.
Thousand Oaks: Sage.

Markus, Erik. 2005. Meat Market: Animals, Ethics, and Money. Boston: Brio Press.

Marshall, Amy D., Jillian Panuzio, and Casey T. Taft. 2005. "Intimate partner violence
among military veterans and active duty servicemen." Clinical Psychology
Review 25: 862-876.

Marshall, David and Marilyn McShane. 2000. "First to ﬁght: Domestic violence and the
subculture of the Marine Corps." Pp. 15-29 in Battle Cries on the Homefront,
edited by P. Mercier and J. Mercier. Springﬁeld: Charles C. Thomas.

Martin, Phillip, J. Edward Taylor, and Michael Fix. 1996. Occasional Paper No. 21 -
Immigration and the Changing Face of Rural America: Focus on the Midwestern
States. Julian Samora Research Institute, East Lansing.

Mercier, Peter J. 2000. "Violence in the military family." Pp. 3-11 in Battle Cries on the
Homefront, edited by P. Mercier and J. Mercier. Springﬁeld: Charles C. Thomas.

233

Miles-Doan, Rebecca. 1998. "Violence between spouses and intimates: Does
neighborhood context matter?" Social Forces 77: 623-645.

Murdock, Steve H. and F. Larry Leistritz. 1982. "Commentary." Pacific Sociological
Review 25: 357-366.

National Farmers Union News. 2002. "Farmers Union Members Testify on Negative
Impacts of Livestock Concentration." p. 3.

National Right to Work Legal Defense Foundation. 2005. "Right to Work States."

Nerlove, Marc. 2002. Essays in Panel Data Econometrics. Cambridge: Cambridge
University Press.

Nibert, David. 2002. Animal Rights/Human Rights: Entanglements of Oppression and
Liberation. Lanham; Boulder; New York; Oxford: Rowman and Littleﬁeld
Publishers, Inc.

Nielsen, Amie, Matthew Lee, and Ramiro Martinez. 2005. "Integrating race, place and
motive in social disorganization theory: Lessons from a comparison of Black and

Latino homicide types in two immigrant destination cities." Criminology 43: 83 7-
871.

Nielsen, Francois. 1999. "Odum Institute Short Course: Analysis of Pooled Time Series
Cross Sections." vol. 2005: The Odum Institute.

Olsson, Karen. 2002. "The Shame of Meatpacking." The Nation 275: 11-16.

Osgood, D. Wayne. 2000. "Poisson-based regression analysis of aggregate crime rates."
Journal of Quantitative Criminology 16: 21-43.

Patterson, Charles. 2002. Eternal Treblinka: Our Treatment of Animals and the
Holocaust. New York: Lantern Books.

Philo, Chris. 1998. "Animals, geography, and the city: Notes on inclusions and
exclusions." Pp. 51-71 in Animal Geographies: Place, Politics, and Identity in the
Nature-Culture Borderlands, edited by J. Wolch and J. Emel. London; New York:
Verso.

Rice, Kennon J. and William R. Smith. 2002. "Socioecological models of automotive
theft: Integrating routine activity and social disorganization approaches." Journal
of Research in Crime and Delinquency 39:304-336.

Rose, David. 2000. "Household panel studies: An overview." Pp. 3-35 in Researching

Social and Economic Change: The Uses of Household Panel Studies, edited by
D. Rose. New York: Routledge.

234

Rosen, Leora N., Robert J. Kaminski, Angela Moore Parmley, Kathryn H. Knudson, and
Peggy F ancher. 2003. "The effects of peer group climate on intimate partner

violence among married male US. army soldiers." Violence Against Women 9:
1045-1071.

Rubin, Israel. 1969. "Function and structure of community: Conceptual and Theoretical
Analysis." International Review of Community Development: 111-119.

Sampson, Robert. 1987. "Urban black violence: The effect of male joblessness and
family disruption." American Journal of Sociology 93: 348-3 82.

Sampson, Robert and W. Byron Groves. 1989. "Community structure and crime: Testing
social disorganization theory." American Journal of Sociology 94: 774-802.

Schlosser, Eric. 2005[2001]. Fast Food Nation: The Dark Side of the All-American
Meal. New York: Houghton Mifﬂin.

Serpell, James. 1986. In the Company of Animals: A Study of Human-A nimal
Relationships. Oxford; New York: Basil Blackwell.

Skaggs, J immy M. 1986. Prime Cut: Livestock Raising and Meatpacking in the United
States, 1607-1983. College Station: Texas A & M University Press.

Skinner, Chris. 2000. "Dealing with measurement error in panel analysis." Pp. 113-125 in
Researching Social and Economic Change: The Uses of Household Panel
Studies, edited by D. Rose. London; New York: Routledge.

Smith, Michael D., Richard S. Krannich, and Lori M. Hunter. 2001. "Growth, decline,
stability, and disruption: A longitudinal analysis of social well-being in four
western rural communities." Rural Sociology 66: 425-450.

Smith, Mick. 2002. "The 'ethical' space of the abattoir: On the (in)human(e) slaughter of
other animals." Human Ecology Review 9: 49-58.

StataCorp. 2005. Stata Statistical Software: Longitudinal/Panel Data. College Station:
StataCorp LP.

Stull, D. D. 1994. "Of Meat and (Wo)Men: Meatpacking's Consequences for
Communities." Kansas Journal of Law and Public Policy 3: 1 12-1 18.

Stull, D. D. and Michael J. Broadway. 1990. "The effects of restructuring on beefpacking
in Kansas." Kansas Business Review 14:10-16.

Stull, Donald and Michael Broadway. 2004. Slaughterhouse Blues: The Meat and Poultry
Industry in North America. Toronto: Wadsworth.

Stull, Donald D. and Michael Broadway. 1995. "Killing them softly: Work in
meatpacking plants and what it does to workers." Pp. 61-83 in Any Way You Cut

235

It: Meat Processing and Small-T own America, edited by D. D. Stull, M.
Broadway, and D. Grifﬁth. Lawrence: University Press of Kansas.

Thomas, Keith. 1983. Man and the Natural World: A History of the Modern Sensibility.
New York: Pantheon Books.

Tonnies, Ferdinand. 1983. "Gemeinschaft and Gesellschaft." Pp. 7-16 in New
Perspectives on the American Community, edited by R. Warren and L. Lyon.
Homewood, Illinois: The Dorsey Press.

Triplett, Ruth A., Randy R. Gainey, and Ivan Y. Sun. 2003. "Institutional strength, social
control and neighborhood crime rates." Theoretical Criminology 7: 439-467.

US. Census Bureau. 2003. "Labor Force, Employment, and Earnings."

US. Census Bureau, EPCD, County Business Patterns. 2006. "County Business
Patterns."

US. Census Bureau, "Estimates and Projections Area Methdology." vol. 2006.

US. Census Bureau, Population Estimates. vol. 2005.

US. Census Bureau, Population Estimates. "State & County Terms and Deﬁnitions."
US. Census Bureau, Small Area Income and Poverty Estimates.

US. Department of Commerce, US. Census Bureau, County Business Patterns.

U. S. Department of Agriculture, Economic Research Service. 2005. "Rural-urban
continuum codes."

US. Department of Justice and Federal Bureau of Investigation. 2004. Uniform Crime
Reporting Handbook.

Unknown author. 2005. "What Meat Means." in The New York Times. New York.

Wagenaar, Theodore. 1981. Readings for Social Research. Belmont: Wadsworth
Publishing Co.

Wilkinson, Ken. 1991. The Community in Rural America. New York; Westport; London:
Greenwood Press.

Wilkinson, Kenneth R, James G. Thompson, Robert R. Jr. Reynolds, and Lawrence M.
Ostresh. 1982. "Response." Paciﬁc Sociological Review 25: 367-376.

Wilkinson, KP, R. Reynolds, J .G. Thompson, and L.M. Ostresh. 1984. "Violent crime in
the western energy development region." Sociological Perspectives 27: 241-256.

236

Wilkinson, KP, J .G. Thompson, R. Reynolds, and L.M. Ostresh. 1982. "Local disruption
and western energy development: A critical review." Pacific Sociological Review
25: 275-296.

Wirth, Louis. 1938. "Urbanism as a way of life." The American Journal of Sociology 44:
1-24.

Yaffee, Robert. 2003. "A Primer for Panel Data Analysis." in Social Sciences, Statistics
and Mapping, vol. 2005: New York University, Information Technology
Services.

York, Richard. 2004. "Humanity and Inhumanity: Toward a Sociology of the
Slaughterhouse." Organization and Environment 17: 260-265.

237