 

 

: w 2
. .mnwe: X

we»: .V
L

9»
»

 

 

 

 

 

 

 

 

3|- ‘
. A ‘3. 1M.
:1: a
. 412
. ,. , kn“ ul. irhazl.
.P‘u bR:lZ-I.tr.o I lo L
i czfﬁultvaFaiih.
‘ . l! v
a...» .311.“ £

Huf..!s.ﬂi...ar:§1§m. ,.

 

..v.| :5:-
.. «wad .
4:9.” striaFuiz3 31.33.
. xbszvbknﬁv‘i.! ,
:1 I

 

 

 

94,6
“H! in
.

. , .33.} Jana.»
3.....-

. ‘plwxx P1531111,
{nu}: . " is}
.All‘ 11.... .-

 

, ‘ .. ‘ ‘ ghniiili‘
‘ . . . . h ‘ . Ililluitw: Juno”: I.\ lav)"
" Egg é . V, ‘ V

*. ‘rg‘ ‘ 5

LIBRARY
!.. Michigan State
University

'I

124:7 /
, gird-1’

This‘is to certify that the
thesis entitled

THE EFFECT OF SOCIALIZATION AND ONSET BEHAVIOR:
AN EXAMINATION INTO YOUTHFUL FIREARM
OFFENDING PATTERNS IN MICHIGAN

presented by

Eric L. Grommon

has been accepted towards fulﬁllment
of the requirements for the

MS. degree in Criminal Justice

 

-._ l-— «- v— ‘

 

 

/../ J
" Majorgrofessor’s Signature

Date

 

MSU is an Affirmative Action/Equal Opportunity Institution

_-.—--.—.--—.-._.-- .

 

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
.“JL 2 s 2888
OCT 2 7 2007
Q6819?

 

 

 

 

 

 

 

 

 

 

 

 

 

2/05 cm.mia'

THE EFFECT OF SOCIALIZATION AND ONSET BEHAVIOR: AN
EXAMINATION INTO YOUTHFUL FIREARM OFFENDING PATTERNS IN
MICHIGAN
By

Eric L. Grommon

A THESIS

Submitted to
Michigan State University
in partial fulﬁllment of the requirements
’ for the degree of

MASTER OF SCIENCE
School of Criminal Justice

2005

ABSTRACT
THE EFFECT OF SOCIALIZATION AND ONSET BEHAVIOR: AN
EXAMINATION INTO YOUTHFUL FIREARM OFFENDING PATTERNS IN
MICHIGAN
By

Eric L. Grommon

The purpose of this research is to model theoretical propositions of social
learning theory with developmental or life course theory to examine socialization
pathways into a subculture that utilizes ﬁrearms for criminal purposes. Inherent
within the research is the assumption that cultural and environmental socialization
mechanisms or onset behaviors place an individual at greater risk for involvement in
ﬁrearm offending, which may lead to specialization in offense patterns. Secondary
analyses were conducted on a 1996 study examining the prevalence and incidence of
youthful ﬁrearm ownership, possession, and use for a random sample of incarcerated
male offenders between the ages of 17 to 25 within the state of Michigan. The results
indicate that socialization or social learning variables measuring cultural and
environmental context served as the best predictors of involvement in serious ﬁrearm
offending. Onset behavior variables provided marginal support for the prediction of
ﬁrearm offending seriousness. In efforts to reduce the frequency and proportion of
crime involving ﬁrearms, criminal justice policy and practices should continue to
‘ examine the nexus between gang membership, peer associations, drug sales, and
neighborhood exposure to violence which lead to initiation, persistence, or increased

levels of involvement in ﬁrearm offending.

ACKNOWLEDGEMENT

I would like to thank my thesis committee members Drs. Timothy Bynum,
Edmund McGarrell, and Christina DeJong for their guidance and support throughout
my thesis experience. Your individual and collaborative insights have made a
difﬁcult task seem manageable and I look forward to working with each of you as I
continue my educational experiences in the School of Criminal Justice. I would like
to extend the utmost gratitude to Dr. Christina DeJong for her willingness to
challenge me to learn, interpret, and explain the ins and outs of ordered logit
regression analysis. Your constant availability and assistance made my learning
experience that much easier and I found the challenge to be one of the most
rewarding parts of my thesis experience.

Special thanks also go out to those who have kept me sane throughout the
thesis process. Thank you to my parents for stressing the value of education and
continuing to provide support and the uncanny ability to quell my nerves throughout
my graduate school experience. Thank you to Andrew, my comedic partner in crime
and brother, for providing enough “goofs” to make me laugh to the point of crying
right when I needed them. Thank you to the criminal justice “Friday Night Feasters”
for being there to bounce ideas off of, discuss nonsense, and share food and libations.
Finally, thanks to Jim, Baker Hall’s night custodian, whose unannounced visits during
the wee morning hours in Suite 138 always let me know when it is time to take a
break away from the thesis to discuss nicknames (e.g., “Scooter,” “Jeep,” “Skippy,”

and “Sunshine”) and the potential of Chicagoland sport teams.

iii

TABLE OF CONTENTS

LIST OF TABLES ............................................................................................................... v
CHAPTER I: INTRODUCTION ................................................... 1
Theoretical Background ........................................................................................... 6
Purpose of Study .................................................................................................... 10
CHAPTER H: REVIEW OF LITERATURE ................................................................... 11
Identiﬁcation of Individuals Within a Firearm Subculture .................................... 15
The Dichotomy of Sport Versus Protective Ownership ........................................ 19
Firearm Utilization for Criminal Purposes ............................................................ 24
Research Question ................................................................................................. 29
CHAPTER III: DATA AND METHODS ........................................................................ 31
Dependent Variable ............................................................................................... 32
Independent Variables ........................................................................................... 34
Methods .................................................................................................................. 37
Hypotheses ............................................................................................................. 38
CHAPTER IV: RESULTS ................................................................................................ 40
Univariate Statistics ............................................................................................... 4O
Bivariate Statistics ................................................................................................. 45
Multivariate Modeling Through Ordered Logit Regression .................................. 50
Multivariate Statistics ............................................................................................ 53
Socialization Modeling .......................................................................................... 55
Summary of Socialization Modeling ..................................................................... 6O
Onset Socialization Modeling ................................................................................ 61
Summary of Onset Socialization Modeling ........................................................... 70
CHAPTER V: DISCUSSION ........................................................................................... 72
Missing Variable Responses .................................................................................. 75
Study Limitations ................................................................................................... 77
Conclusions and Policy Recommendations ........................................................... 79
REFERENCES ................................................................................................................ 101
APPENDICES

Appendix A: Items and Reliability for Socialization Independent Variable Scales
Appendix B: Bivariate Contingency Tables

Appendix C: Variable Transformation Illustration

Appendix D: Multinomial Logit Regression Results

Appendix E: Ad Hoc Bivariate ANOVA Results

iv

LIST OF TABLES

Table l: Coding Strategy for the Levels of Involvement in Firearm Offending .............. 34
Table 2: Independent Variables Used for Analyses......... ................................................. 35
Table 3: Univariate Statistics for the Levels of Involvement in Firearm Offending ........ 40
Table 4: Univariate Statistics for Socialization Independent Variables ........................... 41
Table 5: Univariate Statistics for Onset Independent Variables ....................................... 43
Table 6: Univariate Statistics for Control Variables ......................................................... 45

Table 7: Relationship Between Select Independent Variables and Involvement in
Firearm Offending ............................................................................................................. 47

Table 8: Firearm Involvement Regressed on Socialization Variables Including
legitimized Firearm Ownership or Possession ................................................................. 56

Table 9: Firearm Involvement Regressed on Socialization Variables Including
Crirrrinalized Firearm Possession or Use ........................................................................... 59

Table 10: Firearm Involvement Regressed on Ownership Onset and Socialization
Variables Including Legitimized Firearm Ownership or Possession ................................. 62

Table 11: Firearm Involvement Regressed on Ownership Onset and Socialization
Variables Including Criminalized Firearm Possession or Use .......................................... 65

Table 12: Firearm Involvement Regressed on Arrest Onset and Socialization Variables
Including Legitimized Firearm Ownership or Possession ................................................. 67

Table 13: Firearm Involvement Regressed on Arrest Onset and Socialization Variables

Including Criminalized Firearm Possession or Use ........................................................... 69
Table 14: Items and Reliability for Hard Drug Use Scale ................................................ 84
Table 15: Items and Reliability for Neighborhood Exposure to Violence Scale .............. 84

Table 16: Bivariate Relationship of Respondent Age and Involvement in Firearm
Offending ........................................................................................................................... 85

Table 17: Bivariate Relationship of Respondent Race and Involvement in Firearm
Offending ........................................................................................................................... 86

Table 18: Bivariate Relationship of Respondent’s Population of Previous Residence and
Involvement in Firearm Offending .................................................................................... 86

Table 19: Bivariate Relationship of Respondent’s Survey Facility and Involvement in
Firearm Offending ............................................................................................................. 87

Table 20: Bivariate Relationship of Parental Ownership or Possession of Firearms and
Involvement in Firearm Offending ................................................................................... 87

Table 21: Bivariate Relationship of Parental Arrest or Incarceration for Firearm
Possession or Use and Involvement in Firearm Offending ............................................... 88

Table 22: Bivariate Relationship of Peer Ownership or Possession of Firearms and
Involvement in Firearm Offending .................................................................................... 88

Table 23: Bivariate Relationship of Peer Arrest or Incarceration for Firearm Possession
or Use and Involvement in Firearm Offending .................................................................. 89

Table 24: Bivariate Relationship of Gang Membership and Involvement in Firearm
Offending ........................................................................................................................... 89

vi

Table 25: Bivariate Relationship of Hard Drug Use and Involvement in Firearm
Offending ........................................................................................................................... 90

Table 26: Bivariate Relationship of Drug Sales Frequency and Involvement 1n Firearm
Offending ........................................................................................................................... 91

Table 27: Bivariate Relationship of Neighborhood Exposure to Violence and
Involvement in Firearm Offending .................................................................................... 92

Table 28: Bivariate Relationship of Age at First Firearm Ownership and Involvement in
Firearm Offending ............................................................................................................. 93

Table 29: Bivariate Relationship of Age at First Arrest and Involvement in Firearm
Offending ........................................................................................................................... 93

Table 30: Gang Membership Variable Transformation .................................................... 94

Table 31: Multinomial Logit Regression Odds Ratios for Models 9 and 10 in Violation
of the Proportional Odds Assumption ................................................................................ 95

Table 32: Bivariate ANOVA of Missing Responses on the Variable of Neighborhood
Exposure to Violence on Predictors ................................................................................... 98

Table 33: Bivariate ANOVA of Missing Responses on the Variable of Age at First
Firearm Ownership on Predictors ...................................................................................... 99

Table 34: Bivariate ANOVA of Missing Responses on the Variable of Age at First
Arrest on Predictors ......................................................................................................... 100

vii

CHAPTER I: INTRODUCTION

The prevalence and use of ﬁrearms is ingrained into the fabric of American
history, societal organization, and political structure to such an extent that a ﬁrearm
subculture has been and continues to be observed (Hofstadter, 1970; Jacobs, 2002;
Kennett & Anderson, 1975; Kohn, 2004). Contemporary statistics reinforce the
perception of an American ﬁrearm subculture. Although the exact number of ﬁrearms
may never be known (Wright, Rossi, & Daly, 1983), recent estimates have placed the
number of ﬁrearms owned by US. citizens between 230 to 250 million (Jacobs, 2002;
Kleck, 1991; Kleck, 1997). As a point of reference, Wright, Rossi, and Daly (1983)
estimated that there were approximately 100 to 140 million ﬁrearms owned in the US. 27
years ago. National survey data since the early 19608 have consistently estimated that 40
to 50% of US. households own at least one ﬁrearm (Cook & Ludwig, 2000; Kleck, 1991;
Kleck, 1997; Wright & Rossi, 1994; Wright, Rossi, & Daly, 1983). Extrapolating
household survey percentages, there were approximately 42 to 53 million households in
2000 and 37 to 46 million households in 1990 in possession of ﬁrearms (U .S. Census
Bureau, 2000, Table DP-l; US. Census Bureau, 1990, Table DP-l). While there appears
to be a saturation of ﬁrearms in American society, there is also reason to believe that the
number of ﬁrearms in private possession will continue to increase due to market demand.
Four and a half million new ﬁrearms are sold each year in addition to approximately two
million ﬁrearms sold by secondhand means (Bureau of Alcohol, Tobacco, and Firearms,
2000)

Coupled with the growing number of ﬁrearms available in recent years, the US.

continues to have serious violent crime problem although the relative frequency and rate

of violent crime has incrementally subsided since the early 1990s (Jacobs, 2002). Often,
ﬁrearms are involved in a substantial portion of violent crime, resulting in thousands of
injuries or deaths each year (Cook & Ludwig, 2000). The Federal Bureau of
Investigation’s Uniform Crime Report (U CR) indicates that there were approximately
16,500 homicides in 2003 (Federal Bureau of Investigation, 2004) 1. Supplemental
Homicide Report (SHR) data were provided for 14,400, or 87%, of the UCR reported
homicides. The SHR provides comprehensive information on homicide victim and
offender demographics, relations, and weapon involvement and is considered to be the
only crime measure in which a deﬁnitive determination of ﬁrearm use can be made
(Kleck, 1997). The SHR data reveal that approximately 69% of homicide victims were
murdered with ﬁrearms (Federal Bureau of Investigation, 2004, Table 2.10). Nearly 80%
of the ﬁrearms used in the homicides were handguns and approximately 9% were
committed with riﬂes and shotguns (ibid, Table 2.12).

Of the remaining seven violent crimes measured by the UCR, only two — robbery
and aggravated assault — contain supplemental information concerning weapon
involvement. Unlike SHR data, which is completed by law enforcement personnel
during or after a homicide investigation report, weapon involvement in robberies and
aggravated assaults are based upon the communication between the complainant and the
law enforcement agent dictating the complainant’s report. Since both the complainant

and the agent possess variant amounts of discretion affecting the information that is

 

1 Currently UCR reporting processes are being transitioned into a National Incident Based Reporting
System (NIBRS) that captures comprehensive incident details of 22 offense categories encompassing 46
speciﬁc crimes. The Federal Bureau of Investigation (n.d.) reports that the majority of federal, state, and
local law enforcement crime data are submitted through the traditional UCR format. NIBRS data will not
be reported in UCR yearly crime aggregates until the majority of the reports are received via NIBRS
format. Therefore, statistics presented from the 2003 UCR are based upon the traditional format.

reported (Gove, Hughes, & Geerken, 1985; Skogan, 1976), ﬁrearm violence researchers
have acknowledged that a differentiation between ﬁrearm use, display, threaten or attack
carmot be made for UCR aggregate reports of these two crimes (Kleck, 1997). Instead,
terminology such as “ﬁrearm involvement” is preferred in order to capture victim reports
of confrontations with an armed perpetrator.

Of the estimated 413,400 robberies reported through the UCR in 2003,
approximately 42% involved ﬁrearms followed closely by robberies involving personal
weapons such as ﬁsts and feet (Federal Bureau of Investigation, 2004, Table 2.21). Of
the 857,900 aggravated assaults, personal weapons were involved in approximately 27%
of aggravated assaults, while ﬁrearms were involved in nearly 20% (ibid, Table 2.23).
There is also evidence to suggest that the proportion of homicides, robberies, and
aggravated assaults involving ﬁrearms has been relatively static for the past ten years.
Data from the 1993 UCR indicates that approximately 70% of homicides, 42% of
robberies, and 25% of aggravated assaults involved the use of ﬁrearms (Bureau of Justice
Statistics, n.d.).

The rate of violent crime involving ﬁrearms, found through the standardization of
violent offenses per 100,000 in the population, conﬂicts with the static proportionality of
ﬁrearm involvement in speciﬁc UCR violent crimes. The rates of aggravated assaults
and robberies involving ﬁrearms have both reduced by nearly 45% between 1993 and

20032. Additionally, the rate of homicides committed with a ﬁrearm has reduced by

 

2 The rate of aggravated assaults involving ﬁrearms was 110.5 per 100,000 in 1993 and 56.3 in 2003
(Bureau of Justice Statistics, n.d.). The rate of robberies involving ﬁrearms was 108.5 in 1993 and 59.4 in
2003 (Bureau of Justice Statistics, n.d.)

approximately 42% in the 10 year span3 . However, the decrease in the rate of violent
crimes involving ﬁrearms may be confounded by the general reduction in violent crime
since the early 19903. As the frequency of violent crime reduces ﬁom years past and the
national population remains the same or increases slightly, a reduction in the standardized
rate is logically attained. What remains despite the rate decrease is the fact that ﬁrearms
continue to be involved in a large portion of homicide, robbery, and aggravated assault
violent crime categories.

National Crime Victimization Survey (N CVS) data from 2003 have also indicated
a general decline in the frequency and rate of violent crime since the early 19908
(Catalano, 2004). Similar to UCR data capturing weapon involvement for the crimes of
robbery and aggravated assault, ﬁrearms can only be discussed in terms of their
involvement in criminal encounters Since NCVS data cannot provide any delineation
between ﬁrearm use, display, threaten, or attack ﬂﬂeck, 1997). For the approximate ﬁve
million violent crime victimizations reported by respondents in 2003, 7% involved a I
ﬁrearm (Catalano, 2004, Table 10). Approximately one million or 11% of violent crime
victimizations involved a ﬁrearm in 1993. When analyzed by the type of violent crime,
nearly 25% of the 553,000 robbery victimizations involved ﬁrearms (ibid, Table 10).
Additionally, of the four million simple and aggravated assault victimizations,
respondents reported that approximately 5% involved the use of a ﬁrearm (ibid, Table
10). Perkins (2003) analyses of NCVS trends for the years 1993 through 2001 reinforce

statistics from 2003. Approximately 10% of all violent victimizations involved ﬁrearms.

 

3 The rate of homicides committed with ﬁrearms was 6.6 per 100,000 in 1993 and 3.8 in 2003 (Bureau of
Justice Statistics, n.d.).

Twenty-ﬁve percent of robberies, 5% of simple and aggravated assaults, as well as 3% of
rape/sexual assaults during the time span involved ﬁrearms.

In combination and despite their respective deﬁcits (Gove, Hughes, & Geerken,
1985), UCR and NCVS data both indicate that violent crime is decreasing nationally.
The measures also suggest that 69 to 70% of homicides, 42% of robberies, and 25 to 27%
of aggravated assaults reported to police involve ﬁrearms. Additionally, 25% of
robberies and 5% of simple and aggravated assault victimizations involve ﬁrearms.
While one can expect a sizable proportion of violent crimes to involve ﬁrearms, one can
also expect the availability of ﬁrearms to increase by the millions each year. Kleck’s
(1991) production based analyses of manufacturer, import, and export data led to the
conclusion that “fewer than 1% of handguns and well under 1% of all guns will ever be
involved in even a single violent crime” (p. 47-48). Thus, there appears to be a small
subset of ﬁrearms, and by default a small subset of ﬁrearm owners or possessors,
involved in crime. In other words, there appears to be a fragmented and smaller
subculture within the overall ﬁrearm subculture that utilizes ﬁrearms for criminal
purposes.

In an effort to deter ﬁrearm involvement in crime, criminal justice policy and
legislation at the federal, state, and local level has focused on the identiﬁcation of small
subsets of dangerous persons who may be persistently involved in criminal activity as
well as those who have committed current or previous crimes involving ﬁrearms (Bureau
of Justice Statistics, 2004; Jacobs, 2002; Cook & Ludwig, 2000; Kennett & Anderson,
1975; Kleck, 1991; Project Safe Neighborhoods, 2004; Wright & Rossi, 1994; Wright,

Rossi, & Daly, 1983; Zimring, 1975). As suggested by Jacobs (2002), the justiﬁcation

for policy and legislation that seeks to identify high-risk individuals is often rationalized
under the intuitive auspices that “it is easier to regulate a smaller number of persons or
entities than a larger number” (p. 40). However, with nearly 20,000 different ﬁrearm
laws spanning federal, state, and local levels (Wright, Rossi, & Daly, 1983), there is a
question as to the ability of policy and legislation to affect the frequency and rate of
violent crime since the regulatory effort used to deter future criminality typically occurs
in reaction to observed criminal activity (Kleck & Patterson, 1993). Moreover, the ability
to identify and predict individual offending patterns, in terms of frequency and offense, is
fraught with imperfections (Gottfredson & Gottfredson, 1994). To attend for current
deﬁcits in policy and legislation, an examination of individuals within a subculture that
utilizes ﬁrearms for criminal purposes will be made in order to expand knowledge
concerning the etiology of ﬁrearm offending. By doing so, policies and legislation may
be enhanced and redeveloped to reduce not only crimes involving ﬁrearms but overall
crime as well.
Theoretical Background

Social learning theory is commonly used in the discussion of cultural and
subcultural phenomenon such as substance abuse or dependence (Akers et al., 1979),
gang involvement (Cloward & Ohlin, 1963), and violent behavior in general (Anderson,
1998; Wolfgang & Ferracuti, 1967). More importantly, propositions of social learning
theory have been applied to the discussion of legal and illegal ﬁrearm subcultures
through the examination of individual level socialization experiences (Cao, Cullen, &
Link, 1997; Jacobs, 2002;K1eck, 1991; Kleck, 1997; Sheley & Wright, 1995; Wright &

Rossi, 1994; Wright, Rossi, & Daly, 1983). The theoretical perspective contends that

both deviant and conforrnist behaviors are consequences of socially processed learning
experiences. III certain social environments, there are ideologies, norms, and behaviors
transmitted through interpersonal interactions that may differentially associate individuals
or groups into deviant behavior (Sutherland, 1939). The learning of deviant behavior
through differential association rests on the assumption that primary or intimate groups
expose or socialize an individual to ideologies, norms, and behaviors favorable to law
violating behaviors in excess of deﬁnitions unfavorable to crime (Akers, 1985). Since
not every individual will have a simple dichotomy of non-criminal or criminal
associations, but rather variant levels of association between the two, criminal behavior is
most likely to occur when individuals are “exposed ﬁrst, more frequently, for a longer
period of time, and with more intensity to law violating deﬁnitions than to law abiding
deﬁnitions” (Akers & Sellers, 2004, p. 83).

The most inﬂuential associations are the ones an individual comes into contact
with ﬁrst and most often — family and peers (Akers et al., 1979). Beth of these agents of
socialization provide sources of reinforcement and imitation that foster the development
and maintenance of ideologies and norms and also work to inhibit or increase future
instances of behavior. Learning processes are not limited to interpersonal contact but
may also occur through social institutions, such as schools or churches, and from the
environment as a whole (V old, Bernard, & Snipes, 2002). Speciﬁcally, Sutherland
(1939) has presumed that residents of socially disorganized neighborhoods are more at
risk for association with other individuals who encourage criminal behavior relative to
residents from organized areas who have access to established forms of social control.

With an assortment of learning mechanisms available within one’s environment, an

underlying assumption of social learning theory is the dynamic nature of learning.
Deviant behavior may be learned at different ages, at different frequencies for variant
periods of time, through an assortment of social and non-social interactions (Thomberry,
1987; Warr, 1993).

Developmental or life-course theories have attempted to explain individual onset,
continuity, and change in behavior and is considered to be an extension of socialization
based theories such as social learning (Elder, 1994). Commonly, the theoretical
perspective is used to identify individual level specialization or versatility in offending
behavior through the use of onset age (Blumstein et al., 1986; Blumstein, Cohen, &
Farrington, 1988; Dean, Brame, & Piquero, 1996; Loeber & Le Blanc, 1990; Loeber et
al., 1991; Nagin & F arrington, 1992; Piquero et al., 1999). Individuals who have begun
to manifest criminal behavior prior to or during adolescence are more likely to be
involved in continued criminality later in life relative to those who have started their
criminal activities during their early adulthood (Mofﬁtt, 1993). While there may be a
difference between early and late onset in future offending, research suggests that
individualized offending patterns appear to be more continuous or specialized than
variable (Dean, Brame, & Piquero, 1996; Piquero et a1, 1999). In an attempt to explain
individualized criminal careers, the focus is placed on the variable of age and the extent
to which criminal behavior is characteristically static or dynamic with increases in age.

Hirschi and Gottfredson (1983) suggest that the age distribution of criminal
offending does not vary longitudinally across the life course. Instead, there are speciﬁc
categorical groupings of ages, typically mid to late teens, which consistently have high

rates of criminal behavior across offense types relative to any other age group. Once

individuals have progressed beyond age speciﬁc criminality peaks, an aging-out effect
occurs reducing the frequency and rate of offending. Based on this notion, research
attempts describing age of ﬁrst onset, desistance, and criminal careers are futile since the
age distribution of crime is relatively stable across sex, race, type of crime, space, and
time. What is important under this perspective, however, is the age of the offender Since
age will enable one to predict whether or not criminal behavior is manifest before, during,
or after the peak offending period.

Sampson and Laub (1992) concede that there is a disproportionate amount of
criminal behavior within certain age groups that generally declines with age. However,
the authors suggest that criminal behavior is transitory and may continue or discontinue
dependent upon age progression and signiﬁcant life events. Expressed in their age-
graded theory of informal social control, Laub and Sampson (1993) suggest that informal
bonds, such as family and school, may explain a portion of childhood and adolescent
onset of delinquency. In turn, long-terrn delinquent behavioral patterns and continuity of
offense type may progress or desist through adulthood depending upon life events and the
quality or strength of social ties that an individual has. That is, individual pathways “can
change through interaction with key social institutions as they age” (Sampson & Laub,
1992, p. 81). Therefore, Sampson and Laub’s (1992; Laub & Sampson, 1993)
perspective contends that the age of ﬁrst onset, desistance, and criminal careers is
necessary since long-tenn delinquent behavioral patterns are ﬁrst developed during
childhood and may be reinforced or modiﬁed during the transition to adulthood. In this
regard, age enables one to examine the ability of social mechanisms to modify onset

behavioral discourse that may be criminal or conventional in nature.

Purpose of Study

The purpose of this research is to model theoretical propositions of social learning
theory with developmental or life course theory in order to examine socialization
pathways into a subculture that utilizes ﬁrearms for criminal purposes. Speciﬁc interest
is placed upon the individual and combined inﬂuence of peers, family, and the
environment on subsequent ﬁrearm utilization in offending. Age of onset for delinquent
and ﬁrearm possession behavior will also be included in order to examine the inﬂuence
of subcultural socialization variables in combination with developmental variables on
ﬁrearm offending patterns. Policy recommendations will be developed based on data

analyses.

10

CHAPTER H: REVIEW OF LITERATURE

In order to explore a subset of the larger ﬁrearm subculture that utilizes ﬁrearms
for criminal purposes, it is important to ﬁrst grasp the meaning of the term subculture. At
face value, a subculture could be easily deﬁned as a culture within a culture. As deﬁned,
the term appears to be overly broad and ambiguous, which in turn may lead to difﬁculty
in operationalization and empirical study. Despite these difﬁculties, indirect and direct
attempts to study subcultural behavior continue to persist throughout criminological
literature (Akers, 1985; Cao, Adams, & Jensen, 1997; Hagan et al., 1998).

Wolfgang and F erracuti (1967) have provided a detailed conceptualization of
subculture through seven propositions in their subculture of violence thesis. First, and
closely related to the dictionary deﬁnition of the term, the authors suggest that normative
ideologies and values are not homogeneously spread throughout society or amongst all
subcultures. Instead, “some priority allocation is made, that the subcultural variants may
partially accept, sometimes deny, and even construct antitheses of, elements of the
central, wider, or dominant values, yet remain within that cultural system” (p. 99).
Considering the diversity of American society, it is not unreasonable to assume that there
are a large number of ideologies, norms, and values that may be interrelated or mutually
exclusive and categorized as subcultural. For that manner, individuals may actively or
inadvertently participate in a number of different subcultures based upon the context of
their normative ideologies, values, and behavior (Fisher, 1995). However, participation
within multiple subcultures is assumed to have some degree of continuity. As argued by
the authors, subcultural participation “must be complementary or supplementary;

otherwise individual personality might become unintegrated or disintegrated” (p. 104).

11

Second, individuals or groups who share subcultural ideologies, norms, and
values may variably manifest subcultural behavior across a number of environmental and
interpersonal interactions. Closely aligned with the second proposition, the third
proposition contends that the frequency and type of environmental and interpersonal
situations in which the subcultural behavior is manifested may be quantitatively
measured at the individual level. In turn, the measurement provides an indication of “the
extent to which [one] has assimilated the values associated with the subculture” (p. 159).
At face value, the measurement of subcultural behavior appears to be rather simple — a
measure of situational interactions consisting of ideologies, norms, values, and behavior.
However, in order to gauge the extent of differentiation from the dominant culture, the
ideologies, norms, values, and behavior of the dominant culture must be measured and
compared to the subculture. This implies that subcultures deemed inappropriate by the
dominant culture will be easier to identify and empirically study than those subcultures
that are tolerated by the dominant culture.

Fourth, the authors propose that subcultural ideologies and values are “most
prominent in a lirrrited age group, ranging from late adolescence to middle age” (p. 159).
The inclusion of this proposition is aligned with research acknowledging the concentrated
age distribution of crime within a categorical subgroup of ages relative to all other ages
(Hirschi & Gottfredson, 1983; Laub & Sampson, 1993). Instead of simply extending an
age explanation of criminal behavior to an age explanation of involvement in a
subculture, the authors suggest that the subset of ages disproportionately involved in a
subculture may have deﬁcits in appropriate problem-solving mechanisms and resort to

learned subcultural behaviors. Extending onset propositions from developmental theory,

12

individuals who have been exposed to and manifest behavior derived ﬁom subcultural
ideologies, norms, and values at an early age are more likely to continue subcultural
involvement and behavior as time progresses. Moreover, there is also an increased
potential that the subcultural behavior may become specialized and increase in severity
with earlier ages of onset (Mofﬁtt, 1993).

Fifth, ideologies and values that run counter to the subculture could lead to
individual removal from the group. An underlying assumption of this proposition is
based upon the notion that individuals generally participate within subcultures that are
characteristically conformist, cohesive, and interactive in ideologies, norms, values, and
the behavioral expression of learned ideologies, norms, and values. Subcultures absent or
with lesser degrees of these characteristics may not have the ability to remove individuals
with contradictory or conﬂicting ideologies, norms, and values since knowledge of the
complete ideologies, norms, and values for each individual within the subculture may not
be known. As suggested by the ﬁrst proposition, the term subculture includes a wide
degree of variability in the ideologies, norms, and values that can form or be considered a
subculture. Moreover, the ﬁfth proposition suggests that a single subculture can have
considerable within group variability in ideologies, norms, and values of its members.

Sixth, the development of subcultural ideologies, norms, values, and behaviors are
generally learned and conditioned. The authors utilize social learning theory and suggest
that it is the context of an individual’s complete social environment that provides the
transmission and reinforcement of subcultural ideologies, norms, and values, which in
turn produces subsequent subcultural behavior. Parental relationships and imitation

during childhood, peer relationships, age, social status, and spatial or residential location

13

are all considered to be inﬂuential actors within the social environment and may be
responsible for the socialization and creation of subcultural normative ideologies, values,
and behaviors. Moreover, the authors contend that social learning theory explains a
portion of the between and within group variability in ideologies, norms, and values
found amongst subcultures. As such, socialization through social learning processes is
the key to understanding individual involvement in a subculture and the degree to which
normative ideologies and values are tightly or loosely held.

Finally, the authors posit that criminal behavior of individuals or groups within
one subculture is generally legitimized since the criminal act is directed at an individual
or group that shares common subcultural ideologies, norms, and values. The ﬁnal
proposition appears to deviate from prior statements intending to conceptualize
subcultures and more speciﬁcally a subculture of violence. Rather than integrating
earlier propositions into a concluding operationalization of the term, the authors advance
homicide analyses upon which the formulation of a subculture of violence thesis was
based. Speciﬁcally, the authors suggest that those individuals with criminal records are
both the victims and perpetrators of homicide. Based on this notion, the authors assert
that criminal subculture ideologies, norms, and values become engrained to such a degree
that criminal behavioral can become a part of the individual’s or group’s lifestyle, which
in turn leads to interaction with individuals similarly situated within a criminal
subculture. AS such, commonalities in normative ideologies and values between distinct
subcultures can lead to interactions with individuals who are involved in speciﬁed

behaviors.

14

Wolfgang and F erracuti (1967) do not attempt to determine the empirical validity
of their subculture of violence propositions beyond the notation that there is “a potent
theme of violence current in the cluster of values that make up the lifestyle, the
socialization process, the interpersonal relationships of individuals living in similar
conditions” (p. 140). Other than a small discussion of trends for the crime correlates of
sex, age, race, social class, and the ability of the correlates to alter violent subcultures,
there is little application and more speculation of the empirical results the thesis may
ﬁnd. Moreover, the authors suggest that empirical evidence for the thesis is missing or
tautological and may not include many manifestations of criminal behavior.

What the thesis does provide, however, is an invaluable heuristic that has
integrated psychological and sociological concepts for the meaning and empirical
examination of a single subculture or groups of subcultures. The overarching theme
suggests that any study of subculture will be ﬁlled with complexities — subcultures are
not that different from the dominant culture, there may be a variety of ideologies, norms,
and values within a single subculture, and subcultural behavior may be dependent upon
the situation. At the same time, the thesis highlights the importance of applying an
integrated theoretical approach to the examination of subcultural ideologies, norms,
values, and behaviors with social learning theory as a viable starting point.
Identification of Individuals Within a Firearm Subculture

Research examining the empirical validity of Wolgang and F erracuti’s (1967)
subculture of violence thesis has generated a common discourse for the study of
individual or group involvement in any type of subculture. Commonly, the focal concern

is the identiﬁcation of demographical and structural characteristics that may be

15

responsible for the socialization of cultural or subcultural normative ideologies and
values. Characteristics used to test the subculture of violence thesis have included race
(Messner, 1983), socioeconomic status (Heimer, 1997), institutional environment (Felson
et al., 1994), rural versus urban categories of residence (Fisher, 1995), and Southern
regional residencies (Ellison, 1991; Hawley & Messner, 1989; Loﬁin & Hill, 1974;
Messner, 1983). The focus is not directly placed on the context of the normative
ideologies and values held by the subculture, but rather on the social characteristics and
structures that are assumed to have provided both social and nonsocial sources of
learning, reinforcement, and imitation that foster subcultural behavior (Akers, 1985).

Following the discourse, descriptions of individuals or groups of individuals
within a ﬁrearm subculture have generally been construed through demographic and
structural variables measuring ownership and use (Dixon & Lizotte, 1987; Wright, Rossi,
& Daly, 1983). Wright and Marston (1975) found average ﬁrearm owners, both of any
ﬁrearm in general and handguns speciﬁcally, to be predominantly White Protestants of
middle to upper middle class status from the rural South through the use of National
Opinion Research Center’s General Social Survey data. The average owner also owed
longer ﬁrearms, as 60% of the respondents owned a riﬂe, nearly 60% owned a Shotgun,
and approximately 42% owned a handgun.

Extrapolating from the demographic and structural correlates, Wright and
Marston (1975) inferred that ownership in the South and rural areas should be expected
given the land opportunities for ﬁrearm recreation and sport. Moreover, the researchers
infer that the positive relationship between socioeconomic status and the likelihood of

ﬁrearm ownership is associated with monies needed to purchase ﬁrearms in order to

16

participate in recreational and sporting activities. Data were also limited to an
examination of respondents from urban areas under the “land opportunist” assumption
that urban ﬁrearm ownership cannot be used for recreation or sport, but rather some other
motive such as self-protection. Again however, averageurban owners of any type of
ﬁrearm and handguns speciﬁcally were white Protestants of the upper middle class.

The Wright and Marston (1975) study provided the foundation for ﬁrture
examinations of ﬁrearm subculture through the identiﬁcation of speciﬁc social and
structural characteristics thought to disproportionately produce individual ﬁrearm
ownership through learned normative ideologies and values that accept ﬁrearm
possession and use. In replication with multivariate techniques, Kleck (1997) has found
similar characteristics of owners, consisting of married middle-aged whites earning a
relatively high income and residing in a rural area. Again, the average owner was more
likely to own riﬂes and shotguns rather than handguns. Additionally, Kleck (1997) found
ﬁrearm ownership levels to be the lowest among single young black and Hispanics of
lower income residing in urban areas. Kohn’s (2004) ethnographic research of ﬁrearm
enthusiasts in the San Francisco Bay area found ﬁrearm owners and users to be of middle
to upper class status who use their ﬁrearms primarily for recreation or sport.

Yet, the characterization of an average ﬁrearm owner versus the average non-
owner is not as simpliﬁed as the data suggests. Wright and Marston (1975) implied
through categorical comparisons between individual ownership in rural versus urban
areas as well as ownership of any type of ﬁrearm versus ownership of only own
handguns that individuals within these categories may own ﬁrearms and thereby become

involved in the ﬁrearm subculture for a variety of reasons. Beyond the simple

l7

identiﬁcation of persons through bivariate and ethnographic techniques, socio-
demographical information reveals little concerning how an individual from a speciﬁc
social and structural characteristic is socialized into a ﬁrearm subculture or recursively,
how the ﬁrearm subculture captures individuals from narrowly deﬁned social and
structural characteristics.

Marks and Stokes (1976) have been credited as providing one of the ﬁrst attempts
at the examination of socialization into a ﬁrearm subculture through the use of a non-
probability sample of university students dichotomized as being from Southern and non-
Southem states (Wright, Rossi, & Daly, 1983). Overall, the sample had a distinguished
familiarity with ﬁrearms. Nearly 69% of the sample had ﬁred a ﬁrearm at some time in
their life with approximately 59% ﬁring a ﬁrearm prior to their thirteenth birthday. Male
family members, overwhelmingly fathers, introduced most respondents to ﬁrearms and
guided their ﬁrst ﬁring activity.

To account for regional differential socialization effects (Dixon & Lizotte, 1987),
respondents from Southern versus non-Southem residences were compared. Based upon
bivariate analyses, Southern males were more likely to have ever ﬁred a ﬁrearm and ﬁred
a ﬁrearm for the ﬁrst time at an earlier age relative to non-Southerners. Additionally,
both Southern males and females were more likely to have been brought up in a
household with a ﬁrearm or ﬁrearms present.

Consistent with the propositions of social learning theory, the family provides the
initial and continued sources of conditioning, reinforcement, and imitation that lead to the
formulation of normative ideologies, values, and behavior regarding ﬁrearm use (Akers,

1985; Akers & Sellers, 2004; Jensen, 1972; Loeber & Stouthamer-Loeber, 1986). Based

18

upon the transmission of ﬁrearm normative ideologies, values, and behavior, the family
and the household in which an individual was raised becomes an important agent of
socialization not only in the prediction of subsequent ﬁrearm ownership but also for the
onset age in which an individual has handled a ﬁrearm fOr the ﬁrst time (Caetano, 1979;
Diener & Kerber, 1979; Kleck, 1997; O’Connor & Lizotte, 1978).
The Dichotomy of Sport Versus Protective Ownership

Lizotte and Bordua (1980; Lizotte, Bordua, & White, 1981) have expanded the
knowledge of socialization pathways through the recognition that ﬁrearms may be owned
for sporting, protection, and criminal purposes. That is, the larger ﬁrearm subculture
encompasses subsets of individuals who have been socialized into sporting, protective, or
criminal subcultures, with each respective category of ownership having their own
socialized and learned normative ideologies, values, and behaviors regarding use.
Congruent with the subculture of violence thesis (Wolfgang & F erracuti, 1967), the
researchers posit three constructs that provide evidence of an existence of a sporting,
protective, or criminal subculture. First, normative ideologies, values, and behaviors will
be observed and should, to a variable extent, be different from the dominant society.
Second, there must be a mechanism of generational transmission that socializes an
individual into normative ideologies, values, and behavior of the subculture. Finally,
group membership within the subculture should be identiﬁable based upon contact and
association with similarly situated persons.

Overall, the research provided strong evidence for the existence of a sporting
subculture and partial evidence for the existence of a protective ownership subculture.

The existence of a criminal subculture utilizing ﬁrearms to further criminal behavior

19

could not be made since the sample represented 764 telephone interviews with heads of
households that are assumed to be lawful owners. A key aspect that explained a portion
of the differential socialization pathways into general ﬁrearm ownership is the effect of
cultural and environmental socialization. Cultural socialization processes encompass
learned normative ideologies and values that are exchanged within a distinct social group
(Kleck, 1991; Luckenbill & Doyle, 1989). Firearm ownership in this regard, is
considered to be a product of differential association among speciﬁc individuals or
groups that generally accept ﬁrearm ownership and use behaviors in excess of alternative
ideologies, norms, and values that condone ﬁrearm ownership and use (Cao, Cullen, &
Link, 1997). Environmental socialization processes, on the other hand, are related to the
actual and perceived levels of criminality as well as the extent of inequality found within
a measure of residence generally operationalized at the neighborhood level (Kleck, 1991;
Luckenbill & Doyle, 1989). Ownership in this regard is considered a reactionary product
of self-preservation needs in an apparent hostile environment. While there appears to be
two separate and easily categorized socialization mechanisms available, the difference
between cultural and environmental socialization processes is not clear-cut and may, in
fact, be interrelated. Both processes are grounded upon distinct social groups with
cultural processes examining the effect of small intact groups and environmental
processes examining the effect of the larger spatial area. Further, both processes provide
continued sources of individual learning and behavior adaptation through normative
ideology and value transmission, reinforcement, modiﬁcation, modeling, and imitation.
F arnilial socialization was found to be an important predictor of ﬁrearm

ownership for sport. Parental ownership and household residency within presumably

20

rural counties, as measured by counties with large concentrations of hunting licenses,
were correlated with younger ages of ﬁrearm ownership. In combination, parental
ownership, rural residence, and the youthfulness of ﬁrst ﬁrearm ownership were key
variables in the path model prediction of sporting ownership. Based upon these ﬁndings,
the researchers ﬁnd support for their hypothesis that parents who own ﬁrearms for
hunting or recreational purposes socialize their children into the acceptable, and socially
approved, sporting use of ﬁrearms at an early age, which developmentally inﬂuences
sporting ownership later in life.

Comparatively, environmental socialization was found to be an important
predictor of ownership for protective purposes. The rate of violent crime within the
respondent’s county as well as the respondent’s perception of crime within the county,
their fear of crime, and their victimization experiences were moderate predictors of
protective ownership. Moreover, the level of perceived and actual crime found within the
respondent’s environment was correlated with urban rather than rural populations.

Based on these ﬁndings, protective ownership is a function or a reaction to the presence
or perceived presence of crime with urban environments explaining a large portion of the
link between crime and protective ownership. Yet, the researchers did not completely
rule out the effect of cultural socialization on ﬁrearm ownership for protection. Familial
socialization, in terms of parental ownership for sport, enhanced the predictive power of
the protection model. Thus, parental ownership of a ﬁrearm has the ability to produce
both sport and protective ﬁrearm owners.

The protective path model included an additional measure of socialization that

was not included in the sporting model in the form of peer association. Peer ownership

21

for protective purposes was found to be a strong predictor of subsequent ﬁrearm
ownership for protection. Based upon the strength of the correlation between peers and
protective ﬁrearm ownership, peer association is hypothesized as being the key agent of
socialization that modiﬁes ownership for sport into ownership for protection. That is,
familial socialization forms the foundation for subsequent ﬁrearm ownership, which can
then be altered into sporting or protective ownership based upon the socialization
processes within an individual’s peer group. Although a measure of peer association was
not included in the sporting model, it is not unreasonable to assume under the
propositions of social learning theory and the subculture of violence thesis that
association with peers within a sporting subculture may validly predict sporting
ownership. Unfortunately, there appears to be a dearth of information regarding sporting
subculture socialization beyond familial inﬂuence and rural residence correlates.

Research continues to expand Lizotte and Bordura’s (1980; Lizotte, Bordura, &
White, 1981) socialization dichotomy of ﬁrearm ownership for either sport or protection,
which national surveys have consistently shown to be the primary and secondary motives
for ownership (Kleck, 1991; Kleck, 1997; Wright, Rossi, & Daly, 1983). A number of
characteristics have been used to differentiate between the two types of ownership.
Firearm ownership for sport is considered to be more legitimate and behaviorally of
lower risk to self and society relative to protective ﬁrearm ownership (Lizotte & Bordua,
1980; Lizotte et al, 1994). Generally, ﬁrearm ownership for sport is considered a
throwback to American frontier traditions of the self-sufﬁcient and adept ﬁrearm owner
who utilizes ﬁrearms for hunting and enhancing hunting skills through recreational

activities (Kennett & Anderson, 1975; Kohn, 2004). The type of ﬁrearms owned

22

contributes to the hunting utility of the sporting subculture. Riﬂes and shotguns are the
predominant type of ﬁrearm owned with the frequency of carrying such weapons
paralleling state aggregates of seasonal hunting days (Bjerregaard & Lizotte, 1995;
Kleck, 1991; Kleck, 1997; Lizotte et al., 1994). I

Comparatively, ownership for protection is oﬁen considered to be reﬂective of
high-risk behaviors due to its theorized and implicit association with criminal activity
(Lizotte & Bordua, 1980; Lizotte, Bordua, & White, 1981). Wright and Rossi (1994)
have argued that ﬁrearm ownership for protection, in addition to any other self-
preservation motives, includes “protection in the context of the commissions of crimes”
(p. 139). For instance, Sheley and Wright (1995) found readiness to defend oneself, the
potential that the victim would be armed, and the ability of a ﬁrearm to ease escape from
a criminal encounter to be the most important reasons for possessing a ﬁrearm during the
commission of a crime. Thus, in a rationalistic sense, protective ownership is thought to
encompass a subset of criminal offenders that utilize ﬁrearms in order to protect
themselves from a resistant, and sometimes armed victim, in addition to citizens who
legitimately own ﬁrearms for the sole purpose of self and family preservation.

Protective ﬁrearm ownership has been strongly correlated with individual
participation in drug markets and gangs (Bjerregaard & Lizotte, 1995; Lizotte et al.,
2000; Lizotte et al., 1994). Research examining individual involvement in drug markets,
whether for purchase and subsequent use or for sales, have suggested that ﬁrearms are
threatened, shown, or used in the criminal obtainment of monies or capital to support
habits, to settle disputes between purchasers and dealers, and to diffuse the potential for a

violent encounter between the purchaser and dealer during a drug transaction (Blumstein,

23

1995; F agan & Wilkinson, 1998). As a source of informally regulated revenue, gangs
have become prominent in the drug market and utilize ﬁrearms to advance
entrepreneurial business interests (F agan & Wilkinson, 1998; Howell & Decker, 1999;
Spergel, 1990). In addition to participation in a drug market economy, gang members
threaten, show, or utilize ﬁrearms to manifest identity in a speciﬁc gang or sect of a
larger gang, to defend territory, to participate in criminal behaviors, and for interactions
with other members of the same gang or members of rival gangs (Spergel, 1990).
While the nexus between drug market participation, gang membership, and
protective ﬁrearm ownership or use is expected, research suggests that the agents of
socialization may provide differential effects across ages. Overall, individual
involvement in drug sales has been identiﬁed as the strongest predictor of subsequent
ﬁrearm ownership for protection controlling for variables capturing frequencies of drug
use and purchase as well as gang membership (Bjerregaard & Lizotte, 1995; Lizotte et
al., 2000; Lizotte et al., 1994). When analyses have examined the time-order sequence
between drug market and gang participation, Lizotte et a1. (2000) found gang
involvement to be the most inﬂuential predictor of ﬁrearm carrying for protection for 14
to 16 year olds. Drug dealing and association with peers who own ﬁrearms for protective
purposes were found to be the most inﬂuential predictor for 17 to 20 year olds. Thus,
individual involvement in drug sales may provide the strongest predictor of protective
ﬁrearm ownership, possession, and use, but gang participation provides an early
socialization exposure to ﬁrearms in the context of ownership, possession, or use.

Firearm Utilization for Criminal Purposes

24

While ﬁrearm ownership for protective purposes has been associated with
criminal behaviors, most of the data available for the examination of the criminal use of
ﬁrearms is derived from surveys of arrestees and felons measuring ownership and
utilization. K1eck(1991) has suggested that arrestee data provides a glimpse into lower
level or ﬁrst time offending, while felon data highlights serious and relatively active
offending patterns. Purposive sampling of arrestees and inmates captures not only
generalized criminal activity, but also criminal behavior involving ﬁrearms.

The pervasiveness of ﬁrearm ownership and use among arrestees is so widespread
that it has led some researchers to conclude that the “possession and use of guns is not
only common, but also tolerated and accepted as the norm” (Decker, Pennell, &
Caldwell, 1997, p. 1). In order to understand the normative ideology of ﬁrearm
possession and use among the arrestees, it is important to consider the arrestees’
participation in high-risk behaviors in high-risk environments. Many of the arrestees
interviewed by Decker, Pennell, and Caldwell (1997) as well as Treatment Alternatives
for Safe Communities (1997) staff indicated gang and drug sale involvement, which was
correlated with an increased likelihood for frequent ﬁrearm carrying and utilization in
crime. Many of the arrestees valued the respect and the ability to retaliate that only a
ﬁrearm could provide them, which may be explained by the ﬁnding that arrestees were
often found to be both the victims and perpetrators of ﬁrearm offenses (Decker, Pennell,
& Caldwell, 1997). Not surprisingly, most of the arrestees own or possess a ﬁrearm for
protection or self-defense, with a handgun as the ﬁrearm of choice.

Additionally, many arrestees perceived the need to carry their ﬁrearm routinely in

order to protect themselves within their own neighborhoods. Arrestee interviews

25

conducted within the city of Chicago (Treatment Aitematives for Safe Communities,
1997) found 70% of arrestee ﬁrearm owners agreed that there were a large number of
ﬁrearms available within their neighborhood and 40% agreed that it is important to have
a ﬁrearm for protection within their neighborhood. In cOmbination with the high-risk
behaviors of involvement in gang membership and drug sales, arrestees appear to
consider ﬁrearm ownership and possession a necessity “as much to protect themselves
against the uncertainties of an unfriendly environment as to prey upon the larger
population” (Wright & Rossi, 1994, p. 15).

Wright and Rossi (1994) have provided one of the most comprehensive
examinations into the link between ﬁrearm ownership and criminal behavior through
their non-probability survey of 1,874 felons in 10 states. Similar to arrestee information,
felons’ ownership and use of ﬁrearms during crimes is just as pervasive. Nearly 75% of
the sample had owned some type of a ﬁrearm, with more than 85% of the owners owning
handguns. A majority of the respondents carried a ﬁrearm all of the time or during
perceived instances in which a ﬁrearm may be necessary. In terms of involvement in
ﬁrearm offending, approximately half of the sample had indicated that they had
committed at least one crime with a ﬁrearm.

Given the carrying prevalence and levels of involvement in ﬁrearm offending
amongst the sample, the authors developed four categorical typologies of ﬁrearm
offenders based upon criminal history record information for comparison purposes. The
“one-time ﬁrearm user” consisted of felons who have used any type of ﬁrearm in the
commission of a single criminal act. The “sporadic handgun user” used only handguns in

the commission of relatively few criminal acts. The “handgun predator” used only

26

handguns in the commission of nearly every criminal act that they had been involved in.
Finally, the “shotgun predator” only utilized shotguns or riﬂes for more than one criminal
act.

Across the typologies, the primary offense leading to current incarceration was
robbery. Homicides and aggravated assaults followed as secondary and tertiary crimes
for the one-time ﬁrearm user, while burglary and aggravated assault or weapons offenses
were the secondary and tertiary crimes for the Sporadic handgun user, handgun predator,
and shotgun predator. Based on these ﬁndings, it is apparent that institutionalized
ﬁrearm offenders specialize in robbery offenses and also have a tendency to commit
homicides, aggravated assaults, and burglaries.

Wright and Rossi (1994) also captured information concerning ﬁrearm
socialization. In general the family was credited as providing the initial introduction to
ﬁrearms for the sample, with approximately half of the sample indicating that their father
had provided their ﬁrst shooting experience. On average, the ﬁrst'ﬁring experience
occurred at the age of 13, with subsequent ownership of their ﬁrst shotgun or riﬂe at 15
and handgun at age 18. Additionally, many of the respondents had family members who
had received intervention by the criminal justice system. Unfortunately, statistics were
only presented at the univariate level without any bivariate or multivariate analyses
linking parent or sibling criminality to ﬁrture ﬁrearm ownership or utilization in the
commission of criminal acts. Nonetheless, approximately 25% of the sample had a father
who had been arrested at some point in time, with 18% having a father who had served a

jail or prison sentence. Additionally, over half of the sample had a sibling that had been

27

arrested, with 40% of the sample having a sibling who had served jail or prison
sentences.

While the family provided the initial introduction into a ﬁrearm subculture for the
felons, peers were found to have provided the socialization link between legitimate
ownership and the criminal utilization of a ﬁrearm. Nearly 90% of the sample had
associated with peers who owned or possessed a ﬁrearm. Association with these same
peers was moderately correlated with younger ages of ﬁrst ﬁrearm crime and handgun
obtainment, increased numbers of ﬁrearms owned, and the likelihood of a ﬁrearm being
involved in the conviction offense and the frequency of ﬁrearm carrying.

Sheley and Wright (1995) provided a partial replication of Wright and Rossi’s
(1994) study using survey data from a non-probability sample of juveniles ﬁ’om inner-
city high schools and correctional institutions in four states. Of interest in this discussion
is the extent of ﬁrearm involvement in crime for the sample of 835 institutionalized
juveniles. Family, peers, and the environment as a whole were again found to be
important agents of socialization. Nearly 80% of the sample grew up in a household with
male family members owning ﬁrearms, 90% of the sample was associated with peers
who owned ﬁrearms and possessed them on person frequently, and over 80% of the
sample indicated that they had been threatened or shot at with a ﬁrearm at some point in
their life.

The socialization nexus of gang membership, drug use and sales, ﬁrearm
ownership and use was also apparent for the sample. Approximately 60% of the sample
indicated afﬁliation with a gang. Forty percent of the sample indicated an experience

with cocaine use, 30% indicated crack use, and 20% indicated a use of heroin. While the

28

minority of the sample was involved in drug use, over 70% was involved in drug sales.
Nearly 90% of the sample had owned a ﬁrearm at some point in time, with handguns
being the most commonly owned ﬁrearm. Ahnost all of the respondents who indicated
ownership carried their ﬁrearm occasionally and a majority of the sample had used a
ﬁrearm for a criminal act, with 70% of the criminal actions occurring, on average, prior
to the respondent’s sixteenth birthday.
Research Question

Consistent with Wolfgang and Ferracuti’s (1967) subculture of violence thesis,
legitimate ﬁrearm subcultures for sport do not appear to be that much different from
subcultures that utilize ﬁrearms for criminal purposes in terms of the initial exposure to
ﬁrearms. The family provides the initial ﬁrearm socialization process at an early age
through the presence of a ﬁrearm within the household or by the process of parents ﬁring
ﬁrearms with their children. Yet, there are empirical differences between the two groups.
Individuals within the ﬁrearm subculture of sport are more likely to reside in rural areas
and own and use longer guns such as riﬂes and shotguns. Urban residence and utilization
of handguns are characteristics commonly identiﬁed with subcultures that utilize ﬁrearms
for criminal purposes. A number of measures have been identiﬁed in the literature that
may explain differential socialization processes into a subculture that utilizes ﬁrearms for
criminal purposes beyond the initial familial socialization. Peer association, gang
membership, drug use or sales, and the context of one’s environment are correlated with
individualized use of ﬁrearms in crime and younger ages of ﬁrearm ownership,
possession, and use for arrestees and incarcerated individuals. As such, there is a need to

examine the effect of cultural and environmental socialization measures on individuals

29

who specialize their offending patterns, not in terms of speciﬁc crimes, but through their

utilization of ﬁrearms during criminal activities.

30

CHAPTER IH: DATA AND METHODS

The data used for this research were initially collected for a 1996 study examining
the prevalence and incidence of youthful ﬁrearm ownership, possession, and use for a
random sample of incarcerated male offenders between the ages of 17 to 25 within the
state of Michigan. In terms of the sampling technique for the original study, 3 out of 26
state managed correctional facilities were ﬁrst selected based upon their inmate
populations and security levels. Of the three facilities selected for data collection,
Institution One was an initial processing center for male offenders under the age of 25
and housed inmates across all security levels. Institution Two was classiﬁed as a medium
security facility housing male inmates over the age of 17. The third and ﬁnal institution
housed male inmates over the age of 14 and was classiﬁed as a medium security facility.

Inmates were selected from the three correctional institutions based upon the
criterions of age (inmates between the ages of 17 to 25) and commitment date (inmates
committed to the institution within the past seventeen months). A. simple random sample
was selected from a Department of Corrections management information system list of
inmates who matched age and prison commitment criterion. In turn, the inmate list of
offenders selected through the simple random sample formed the pool of potential
participants to be included in the study.

In terms of the research design, surveys were administered to small groups of 10
to 15 inmates. T wo research assistants were present to monitor and administer the
survey. At the beginning of each session, inmates were informed that their participation
was voluntary and their identity would be kept anonymous. Respondents were not asked

to provide their name, prison identiﬁcation number, or any other information that could

31

allow for future identiﬁcation. Furthermore, the respondents were not given any
incentive for their participation beyond their participation and break from the routine.

In the administration of the survey, each question and potential response were
read aloud in order to maximize inmate comprehension. The results of the individual
surveys were only available to members of the research team, as individualized survey
results were aggregated for reporting purposes. Further, none of the individual responses
were or could be singled out for reporting purposes as unique and personal identiﬁers are
not available.

The ﬁnal data set contains self-report survey information from a total of 501
respondents reﬂecting an assortment of variables concerning offender demographics,
familial characteristics, employment history, ﬁrearm use and acquisition patterns,
attitudes and perceptions of ﬁrearm use and violence, drug use history, as well as the
nature and level of involvement with crime and peers". Supplemental current offense,
criminal history, and drug and alcohol use and treatment history information for the
respondents were captured from the respondent’s pre-sentence investigation report and
included within the data sets. As such, the data provides a wealth of information
concerning the acquisition, possession, and use of ﬁrearms amongst a sample of
relatively serious youthful offenders incarcerated during or relatively soon after their
theorized peak offending period (Hirschi & Gottfredson, 1983; Laub & Sampson, 1993).

Dependent Variable

 

‘ The response rate for the initial 1996 study was 57%, with 929 inmates selected for participation based
upon the criterions of age and corrrrnitrnent date and 404 of the inmates refusing participation (Huebner,
Bynum, & Hinduja, 2001). With 24 incomplete surveys, the ﬁnal analyses were reduced to 501.

5 Michigan Compiled Law 771.14 states that a pre-sentence investigation report be prepared for the
sentencing judge in every case in which a person is charged with and convicted of a felony.

32

As mentioned previously, the objective of this research was to determine the
effects of socialization and developmental variables on levels of involvement in ﬁrearm
offending. In order to capture data regarding individual involvement in ﬁrearm
offending, a typology of ﬁrearm offending behavior was~ adapted from Wright and Rossi
(1994). The operationalization of the typology is based upon whether or not the
respondent had committed any crimes involving a ﬁrearm as well as the frequency in
which the respondent carried a ﬁrearm prior to incarceration. The four measures used to
develop the dependent variable include:

(a.) “Did you ever use a gun to commit a crime?”

(b.) “Have you ever committed any other crime using a gun?”

(c.) “Was the offender in possession of a weapon at the time of the offense?”

(d.) “While on the street, how often did you carry a gun?”

The ﬁrst two measures are based upon self-report survey responses and include a
dichotomous response set of “yes” or “no” describing prior ﬁrearm involvement in crime.
The third measure is used to verify self-report survey responses through the use of pre-
sentence investigation report information. In the circumstance in which a respondent
provides a “no” response for questions of “ever use a gun to commit a crime” and “ever
committed any other crime with a gun,” pre-sentence investigation report indications of
ﬁrearm involvement will serve the purpose of overriding a “no” response to a “yes”
response. The ﬁnal measure is used to capture the respondent’s frequency of ﬁrearm

carrying based upon self-report survey information and includes a discrete response set

9’ 6‘ ’9 6‘

consisting of six categories: “never, couple of times a year, once a month,” “couple

of times a month,” “once a week,” and “every day.”

33

Table 1 provides the coding strategy used to formulate a rank ordered dependent
variable typology consisting of three levels ranging ﬁom low (e. g., “no involvement in
ﬁrearm offending”) to high (e. g., “serious involvement in ﬁrearm offending”) degrees of
ﬁrearm offending seriousness. In general, the coding strategy is contingent upon the
frequency of ﬁrearm carrying, which is based upon research suggesting that the
distinguishing feature between individuals who use ﬁrearms to commit crimes and those
who do not is the frequency of ﬁrearm carrying (Wright & Rossi, 1994). For example, if
a respondent provides a “no” or missing response for the variable of prior utilization of a
ﬁrearm for a crime, but does provide an indication of ﬁrearm carrying behavior, then the
subsequent coding is one of moderate involvement.

Table 1: Coding Strategy for the Levels of Involvement in Firearm Offending

 

Firearm Cm‘ng Frequenevl'

In Crime Never Couple Once a Couple Once a Every Missing
Times a Month Times a Week Day

 

 

Year Month
Yes M1 M1 MI MI SI ' SI MI
No NI MI M1 M1 MI M1 N1
Missing NI MI MI M1 M1 M1 Missing

 

TKey: NI=No Involvement, MI=Moderate Involvement, SI=Serious Involvement.

Independent Variables

Ten variables were selected from the self-report survey data in order to Observe
their effect on levels of involvement in ﬁrearm offending. Table 2 presents an overview
of the independent variables as well as their levels of measurement. Eight of the
independent variables were used to observe the effect of socialization on the respondent’s
level of involvement in ﬁrearm offending. The socialization variables rely upon the
theoretical perspective of social learning theory with emphasis on the differential

association processes of parents, peers, gangs, drug markets, and the overall

34

neighborhood environment as a whole. The remaining two independent variables are
used to observe the effect of onset behaviors on the level of involvement in ﬁrearm
offending.

Table 2: Independent Variables Used for Analyses

 

Independent Variables Level of Measurement

 

Socialization/Social Learning Variables
Parents Own or Possess Firearms Dichotomous (Y es/No)

Arrested/Incarcerated for Possession/Use Dichotomous (Yes/No)

Peers Own or Possess F irearrns Dichotomous (Y es/No)
Arrested/Incarcerated for Possession/Use Dichotomous (Y es/No)

Gang Membership Dichotomous (Y es/No)

Hard Drug Use Scale Ordinal Scale (0-5)

Drug Sales Frequency Ordinal Range (1—6)

Neighborhood Exposure to Violence Scale Ordinal Scale (0-8)

Onset Variables

Age of First Firearm Ownership Continuous

Age of First Arrest for Any Crime Continuous

 

Familial socialization consists of a dichotomous measure of parental ownership or
possession of ﬁrearms and a dichotomous measure of parental arrest or incarceration for
ﬁrearm possession or use. Similarly, peer socialization consists of a dichotomous
measure of peer ownership or possession of ﬁrearms and a dichotomous measure of peer
arrest or incarceration for ﬁrearm possession or use. The inclusion of multiple measures
for familial and peer socialization variables will be for comparative purposes, as bivariate
and multivariate modeling will attempt to observe the effect of legitimized ownership or

possession (i.e., parents and peers own or possess ﬁrearms) versus criminalized

35

possession or use (i.e., parents and peers arrested or incarcerated for ﬁrearm possession
or use) on the level of involvement in ﬁrearm offending.

The remaining socialization variables include gang membership, drug market
participation, and environmental violence. Gang membership is a dichotomous measure,
which has been validated based upon a secondary response indicating whether or not the
gang was organized. Drug market participation consists of separate measures for drug
use and sales. A hard drug use scale was gleaned from Sheley and Wright (1995) and
measures the respondent’s experimentation with heroin, crack, cocaine, PCP, and
methamphetamines. The continuum of the scale ranges from no hard drug use (0) to use
of every hard drug included in the scale (5). The drug sales frequency measure provides
a continuous range of sale activity from “never sold drugs” (0) to sold drugs “almost
every day” (6). Finally, environmental violence is measured by a scale of neighborhood
exposure to violence consisting of prior neighborhood victimization experiences as well
as attitudinal responses to the statements of “in your neighborhood, there are many
ﬁrearms on the street” and “in your neighborhood, it is important to have a ﬁrearm for
protection.” Along a continuum, the scale ranges from no exposure to violence (0) to
maximum exposure (8). More information regarding the creation of hard drug use and
neighborhood exposure to violence scales are contained in Appendix A, Tables 13 and
14.

The two developmental or onset variables are continuous measures of self-
reported age at ﬁrst ﬁrearm ownership and age at ﬁrst arrest for any crime. Similar to
familial and peer socialization measures discussed previously, the use of two separate

measures of onset behavior will be used to compare their effect on the level of

36

involvement in ﬁrearm offending. Four additional variables will be used in order to
control for demographic differences. Control variables include respondent’s age, race,
and population of previous residence6. The ﬁnal control measure includes the
respondent’s institution from which they were sampled.

Methods

Univariate statistics will be provided for all independent and control variables as
well as the dependent variable. Bivariate statistics, through the use of contingency tables,
will be provided for each of the independent and control variables in order to assess
individual relationships with the dependent variable. Multivariate modeling through the
use of ordered logit regression analyses will be used to examine the individual and
collaborative relationships of the control variables and selected independent variables on
the level of involvement in ﬁrearm offending. Two sets of multivariate models will be
produced. One set will consist of restricted and full models measuring the effect of
socialization variables. The second set of models will consist of restricted and full
models measuring the effect of onset variables in addition to socialization variables.

The purpose behind the use of restricted and full modeling is to observe the
changes in the overall ﬁt of the model and logit coefﬁcients with the addition of the
theoretically relevant socialization variables of gang membership and drug market
participation to a baseline model that only contains variables controlling for the effect of
demographic characteristics, familial socialization, peer socialization, and environmental

socialization. Therefore, the purpose is theoretical in nature by attempting to develop

 

6 Survey respondents were asked to provide the name of the city in which they resided prior to
incarceration. Based upon their response, US Census Bureau (2000) population information was captured
and used to develop the respondent’s population of previous residence.

37

parsimonious models that include variables empirically used to predict participation in a
subcu1ture that utilizes ﬁrearms for criminal purposes. Restricted modeling in this
manner is different from “reduced” modeling in criminological literature, which attempts
to improve the overall ﬁt of a model for the purpose of maximizing the model’s ability to
explain the variance in the dependent variable. Reduced modeling ﬁrst develops a full
model including theoretically relevant variables and selects only those variables found to
be statistically signiﬁcant for inclusion in a reduced model. Insigniﬁcant variables are
removed.
Hypotheses

Six alternative hypotheses are derived ﬁom the literature and data available for
analyses:
Hal: There is a relationship between family ﬁrearm ownership, possession, and use and
individual levels of involvement in ﬁrearm offending.
Hazi There is a relationship between peer ﬁrearm ownership, posSession, and use and
individual levels of involvement in ﬁrearm offending.
Ha3: There is a relationship between gang membership and individual levels of
involvement in ﬁrearm offending.
H34: There is a relationship between drug market participation, in terms of use and sales,
on individual levels of involvement in ﬁrearm offending.
Has: There is a relationship between the levels of violence found within one’s
environment and individual levels of involvement in ﬁrearm offending.
H36: There is a relationship between the age of ﬁrearm ownership onset and the age of

ﬁrst arrest on individual levels of involvement in ﬁrearm offending.

38

Hypotheses 1 though 4 reﬂect cultural agents of socialization (Kleck, 1991;
Luckenbill & Doyle, 1989) that have been empirically shown to inﬂuence legitimate and
criminal ﬁrearm ownership, possession, and use behaviors. Hypothesis 5 builds upon
prior research suggesting that environmental socializatiOn processes (Kleck, 1991;
Luckenbill & Doyle, 1989) have an effect on ownership, possession, and use. Finally,
hypothesis 6 predicts, under the auspices of developmental or life-course theories, that
earlier onsets of ownership or criminal activity may lead to an increased risk of persistent
and specialized ﬁrearm offending (Dean, Brame, & Piquero, 1996; Mofﬁtt, 1993;

Piquero et al, 1999).

39

CHAPTER IV: RESULTS
Univariate Statistics

Table 3 provides the univariate statistics for the ordinal dependent variable level
of involvement in ﬁrearm offending consisting of three categories of increasing
seriousness. Overall, the majority of the sample has some level of involvement in ﬁrearm
offending behavior with only 22% lacking involvement. The modal category of serious
involvement suggests that the sample captures a large portion of individuals who utilize
ﬁrearms for criminal purposes and carry their ﬁrearm on a regular basis. In fact, the
overall distribution of the dependent variable appears to be biased toward the severe end
of the continuum.

Due to the self-report nature of the data and potential for respondent non-
response, the munber or percent of missing cases will be provided throughout the
univariate results section. Accordingly, the dependent variable is missing for
approximately 2% of the sample.

Table 3: Univariate Statistics for the Levels of Involvement in Firearm Offending

 

 

(n=501).

Variables Frequency (n) Percentage
No Involvement 112 22%
Moderate Involvement 146 29%
Serious Involvement 235 47%
Missing 8 2%

 

Table 4 provides the univariate statistics for the socialization independent

variables used in the analyses.

40

Table 4: Univariate Statistics for Socialization Independent Variables (n=501).

 

 

Variables Frequency (Percent) Percent Missing Mean (s.d.)

Parents Own/Poss. 501 0% .40 (.49)
Parents Arr./Incarc. 500 <1% .07 (.25)
Peers Own/Poss. 488 3% .87 (.34)
Peers Arr./Incarc. 489 2% .72 (.45)
Gang Membership 472 6% .30 (.46)
Hard Drug Use“ 433 14% .90 (1.42)

Never Used (0) 267 (62%)

Mad. User (1-2) 100 (23%)

Serious User (3-5) 66 (15%)
Drug Sales Frequency* 480 4% 3.98 (2.17)

Never Sold (1 ) I 33 (28%)

Few Life/Year (2-3) 55 (I I %)

Few Mo/Week (4-5) 78 (16%)

Every Day (6) 214 (45 %)
Exposure to Violence“ 455 9% 4.65 (2.22)

No Exposure (0) 30 (7%)

Low Exposure (1—3) 1 00 (22 %)

Mod. Exposure (4-6) 223 (49%)

Max. Exposure (7-8) 1 02 (22%)

 

‘Hard drug use scale, frequency of drug sales, and neighborhood exposure to violence scale will be utilized
as ordinal measures for bivariate and multivariate analyses. The nominal groupings shown here are

provided for presentation purposes.

Approximately 40% of the sample had parents who owned or possessed ﬁrearms,

with the majority of the respondents lacking parents who owned or possessed ﬁrearms.

Few of the respondents, approximately 7%, had parents who were arrested or

incarcerated for ﬁrearm possession or use. Most of the respondents were associated with

peers who owned or possessed ﬁrearms and were associated with peers who had been

arrested or incarcerated for ﬁrearm possession or use. The ﬁnal dichotomous

41

socialization measure indicates that 30% of the respondents identiﬁed themselves as gang
members. In terms of the overall variability in response for the dichotomous variables,
parental ownership or possession of ﬁrearms had the most variance in response, as
exempliﬁed by the closest approximation to an even split amongst the dichotomous
categories, followed by gang membership, association with peers arrested or incarcerated
for ﬁrearm possession or use, association with peers who own or possess ﬁrearms, and
parents arrested or incarcerated for frrearrn possession or use respectively.

For the ordinal socialization independent variables, the hard drug use scale ranges
in value from “never used” (0) to “serious hard drug user” (5) and has a mean of .90 and
a standard deviation of 1.42. The statistics appear to be concentrated toward the lower
end of the scale and suggest that many of the respondents are not partaking in hard drug
use or use only one type of substance. Sixty-two percent of the respondents indicated
that they had never used heroin, crack, cocaine, PCP, or methamphetamines. Twenty-
three percent indicated that they had used one or two hard drugs, while 15% indicated
that they had used three or more of the hard drugs. Furthermore, the hard drug use scale
appears to have a substantial proportion of missing cases, with approximately 14% of the
respondents failing to provide a response.

The second ordinal socialization measure, frequency of drug sales, ranges in value
from “never sold” (1) to “sold every day” (6) and has a mean just below 4 with a standard
deviation of 2. 1 7. The marginal percentages of the variable indicate that approximately
72% of the sample reported some experience in drug sales with the remaining 28%
indicating that they have never sold drugs. Of the respondents indicating experience in

drug sales, approximately 45% sold almost every day.

42

Finally, the neighborhood exposure to violence scale ranges in value ﬁ'om “no
exposure to violence” (0) to “maximum exposure to violence” (8) with a mean just below
5 and a standard deviation of 2.22. Very few of the respondents indicated that they had
not been exposed to any violence in their neighborhood. Most, in fact, have been
exposed to some form of violence within their neighborhood as both the marginal
percentages and mean statistic suggest that the respondents are exposed to moderate
amounts of violence in their neighborhood. Approximately 9% of the respondents also
failed to provide a response for the neighborhood exposure to violence scale, which
represents the second highest proportion of missing cases for the socialization
independent variables.

Table 5 provides the univariate statistics for the onset independent variables used

in the analyses.

 

Table 5: Univariate Statistics for Onset Independent Variables (n=501).

 

Variables n Percent Mln Max . Mean (s.d.)

' Missing
First Firearm 386 23% 5 21 13.83 (2.66)
First Arrest 472 6% 7 23 15.31 (2.71)

 

The average age of ﬁrst ﬁrearm ownership is nearly 14 years old with a standard
deviation of 2.66. The youngest self-reported age in which the respondents ﬁrst owned a
ﬁrearm was 5 years of age and the oldest self-reported age was 21. Overall, the standard
deviation statistic is quite low in comparison to the mean, which suggests an approximate
normal distribution of the variable as the dispersion of ages is relatively centered near the
mean. The age of ﬁrst ﬁrearm ownership also contained the highest proportion of

missing cases among all the variables included in the analyses. For the second onset

43

independent variable, the average age of ﬁrst arrest for any crime is 15 years of age with
a standard deviation of 2.71. The youngest self-reported age of ﬁrst arrest was 7, while
the oldest age of ﬁrst arrest was 21. Similar to the age of ﬁrst ﬁrearm ownership, the age
of ﬁrst arrest for any crime approximates a normal distribution as revealed by the
standard deviation statistic.

Univariate statistics for the control variables used in the analyses are presented in
Table 6. The average respondent is approximately 20 years of age with a standard
deviation of 1.86. Considering that the sample was restricted to respondents between the
ages of 17 to 25, the observed average is just below the midpoint of the range. In terms
of the average respondent’s race or ethnicity, the marginal percentages indicate that 49%
of the respondents were black, 38% were white, 7% were Hispanic, and 6% provided an
“other” indication of race. The variable capturing the respondent’s population of
previous residence indicates that the sample includes a blend of individuals from sparsely
populated as well as heavily populated areas7. The marginal percentage mode was at the
low end of the continuum, with 38% of the respondents previously residing in cities with
populations less than 50,000 inhabitants, which is followed by 27% of the respondents at
the high end of the continuum from populations in excess of 250,000. Interestingly
enough, all of the respondents from populations in excess of 250,000 persons indicated

residence within the city of Detroits. Finally, the majority of the respondents,

 

7 According to the US. Census Bureau (2000), an urban area is generally operationalized as densely settled
census blocks with total populations of at least 2,500 for urban clusters, or at least 50,000 for urbanized
areas. Alternatively, rural areas are generally considered areas that are not urban. Since census block
infommion could not be captured and differentiations between urban clusters and urbanized areas cannot
be made based upon the self-report nature of the survey, there is an assumption that respondents from cities
with populations less than 50,000 are presumably more “rural” than those respondents from populations of
50,000 or more.

8 US. Census Bureau (2000, Summary File 1) information suggests that the city of Detroit (population of
951,270) accounted for approximately 10% of Michigan’s total population (9,938,444).

44

approximately 58%, were sampled from Institution two, which was a medium security
facility housing male inmates over the age of 17.

Table 6: Univariate Statistics for Control Variables (n=501).

 

 

Variables Frequency (Percent) Percent Missing Mean (s.d.)
Respondent Age 498 <1% 20.43 (1.86)
Respondent Race 477 5% 1.70 (.85)

Black 235 (49%)
White 1 79 (38%)
Hispanic 33 (7%)
Other 30 (6%)
R. Pop. of Residence 493 2% 2.34 (1.23)
<50, 000 186 (38%)
50, 001-100, 000 86 (I 7%)
100,001-250,000 89 (18%)
250,001+ 132(27%)
Respondent Facility 501 0% 1.65 (.82)
Institution 1 I 12 (22%)
Institution 2 288 (58%)

Institution 3

101 (20%)

 

Bivariate Statistics

Chi-square tests of independence are used to examine bivariate relationships

between two discrete variables containing dichotomous or multinomial distributions

(Agresti, 1989; Agresti, 2002; Bachman & Paternoster, 2004). Speciﬁcally, the chi-

square test examines the joint occurrence of two discrete variables and the probability

that the co-occurrence of the two variables is above and beyond the product of their

separate probabilities (Bachman & Paternoster, 2004). In order to test for the

independence of a selected independent variable and the level of involvement in ﬁrearm

45

offending, separate chi-square analyses will be produced for each control and
independent variable.

Three independent variables measuring age are collapsed into discrete categories
in order to complete the chi-square analyses. All three Of the variables were collapsed
into dichotomous categories based upon the variable’s mean. The control independent
variable of respondent age had a mean of 20 years of age and was collapsed into two
categories consisting of 17 to 20 year olds and 21 to 25 year olds. The age of ﬁrst
ﬁrearm ownership had a mean of 14 years of age and was collapsed into categories of 0
to 13 years old and 14 years of age and above. Finally, the age of ﬁrst arrest for any
crime had a mean of 15 years of age and was collapsed into 0 to 14 years of age and 15
years of age and above. The remaining ordinal measures, hard drug use scale, drug sales
frequency range, and neighborhood exposure to violence scale, were kept in their original
ordinal level of measurement.

Table 7 provides the chi-square and measure of association statistics for the
control, socialization, and onset independent variablesg. For presentation purposes,
Table 7 includes the percentage not involved and the percentage involved (i.e., the
combination of moderate and serious involvement) in ﬁrearm offending. Separate
contingency tables for each of the control (Tables 16 through 19), socialization (Tables
20 through 27), and onset (Tables 28 and 29) independent variables across the three

categories of the ordinal dependent variable are provided in Appendix B.

 

9 According to Bachman and Paternoster (2004) lambda measures of association (A) are used for nominal
variables and garmna measures (7) are used for ordinal variables. For Table 7, the use of the lambda or
gamma statistic will be dependent upon the speciﬁc independent variable’s level of measurement. Thus,
nominal independent variables will result in a lambda statistic discussion while ordinal independent
variables will result in a gamma statistic discussion.

46

Table 7: Relationship Between Select Independent Variables and Involvement in

 

 

Firearm Offending.

Variables Percent Not Percent f Measure of

Involved Involved Association

§0_ntr_01
R’s Age 23% 77% 6.51* -.19
R’s Race 23% 77% 17.95“ .00
R’s Pop of 22% 78% 2595*" .28
Residence
R’s Facility 23% 77% 10.06* .00
Socialization
Parents WP 23% 77% 1.30 .00
Parents M 23% 77% 9.06" .00
Peers O/P 23% 77% 6660*" .13
Peers AH 22% 78% 119.56*** .21
Gang Membership 23% 77% 60.34*** .00
Hard Drug Use 22% 78% 15.36 .18
Drug Sales Freq. 23% 77% 123.15*** .58
Violent Exposure 23% 77% 157.9l*'** .58
_(_)_ns_et
First Firearm 11% 89% 9.28" .28
First Arrest 22% 78% 5.59 .13

 

*p < 0.05, up < 0.01, mp < 0.001

All of the control variables are signiﬁcantly related to the level of involvement in
ﬁrearm offending (a < .05). Despite the signiﬁcance of the control variables, lambda
statistics for the respondent’s race and facility from which they were sampled are well
under one percent (A = .00) revealing a relatively non-existent level of association with
the dependent variable. Gamma statistics for the respondent’s age (7 = -.19) and
population of previous residence (7 = .28) suggest a weak relationship with the level of

involvement in ﬁrearm offending. Respondent’s age is negatively related to the

47

dependent variable, suggesting that as age increases (ceiling limit of 25) the level of
involvement in ﬁrearm offending decreases. Reciprocally, as age decreases (ﬂoor limit
of 17) the level of involvement in ﬁrearm offending increases. In terms of previous
residence, respondents from areas of larger populations. are more likely to be involved in
serious ﬁrearm offending than respondents from smaller populated areas.

Two socialization variables, parental ownership or possession of ﬁrearms and
hard drug use, produced insigniﬁcant chi-square statistics revealing a lack of a
relationship between the variables and involvement in ﬁrearm offending. In fact, for the
variable parental ownership or possession of ﬁrearms, the probability of obtaining a chi-
square statistic as large as 1.30 by chance with repeated sampling of a similar sample size
is quite large (a = 0.52). Beyond these two variables, the remaining socialization
variables were signiﬁcantly related to the level of involvement in ﬁrearm offending (a <
.05). The six Signiﬁcant socialization variables include: parental arrest or incarceration
for ﬁrearm possession or use, association with peers who own or possess ﬁrearms,
association with peers who have been arrested or incarcerated for ﬁrearm possession or
use, gang membership, frequency of drug sales, and the level of neighborhood exposure
to violence.

Parental arrest or incarceration for ﬁrearm possession or use and gang
membership lambda statistics suggest that the utilization of these two socialization
variables does little to reduce errors in the prediction of involvement in ﬁrearm offending
(A = .00). On the other hand, errors in the prediction of involvement in ﬁrearm offending
can be reduced by approximately 13% through the use of the peer ownership or

possession of ﬁrearms (A = .13). Moreover, errors in the prediction of the dependent

48

variable can be reduced by approximately 21% with the inclusion of the peer arrest or
incarceration for ﬁrearm possession or use (it = .21). Both frequency of drug sales (7 =
.58) and the level of neighborhood exposure to violence (7 = .5 8) gamma statistics
revealed moderately positive relationships with involvement in ﬁrearm offending.
Frequent drug sale activity is more likely to lead to serious involvement in ﬁrearm
offending relative to individuals whose drug sale activity is rather infrequent.
Additionally, increased exposure to neighborhood violence is more likely to lead to
serious involvement than lower levels of exposure to neighborhood violence.

Finally, the onset variable of age at ﬁrst ﬁrearm ownership was signiﬁcantly
related to involvement in ﬁrearm offending (a < .05), while the second onset variable,
age at ﬁrst arrest for any crime was insigniﬁcant. The gamma statistic for the age of ﬁrst
ﬁrearm ownership variable reveals a weak positive relationship with involvement in
ﬁrearm offending ('y = .28). That is, as the age of ﬁrst ﬁrearm ownership increases the
likelihood for serious involvement in ﬁrearm offending also increases.

It is important to note the number of cases used for each of the contingency tables
presented in Appendix B. For the control independent variables, the sample size used for
analysis ranged from 470 to 493 cases. The sample size ranged from 427 to 493 cases for
the socialization variables and the sample size ranged from 384 to 464 cases for the onset
variables. The variability of the sample size used for each of the separate independent
variables is due to the use of complete case chi-square analyses (Brame & Paternoster,
2003). That is, chi-square analyses were performed on respondent information that was
complete for the independent and dependent variable, with missing cases excluded from

analyses. Therefore, observations that contain substantial proportions of missing cases

49

(i.e., hard drug use scale, neighborhood exposure to violence scale, age of ﬁrst ﬁrearm
ownership and age of ﬁrst arrest) may bias the chi-square results as chi-square estimates
are heavily inﬂuenced by sample size (Bachman & Paternoster, 2004). Multivariate
modeling will attempt to control for the effect of missing cases for variables with a large
number of missing cases.
Multivariate Modeling Through Ordered Logit Regression

Ordered logit regression is a common multivariate statistical procedure for ordinal
level response categories as opposed to linear regression or binary logistic regression
modeling (Agresti, 1989; Agresti, 2002; Long, 1997; McKelvey & Zavoina, 1975;
Winship & Mare, 1984). Linear regression considers response categories as interval or
continuous variables having constant and measurable intervals between increasing and
decreasing variable values. Based on the assumption of linearity, linear regression
attempts to ﬁt a straight line through observed data and examine the change in the
dependent variable produced by a change in the independent variable (Berk, 2004).
Comparatively, ordered logit acknowledges the lack of constant intervals between the
categories of an ordinal variable and controls for the limited distribution of the variable.
Regression coefﬁcients are produced and interpreted as the amount of change in the
dependent variable on its underlying scale continuum (e. g., less to more severe) brought
about by a unit change in the independent variable holding other variables constant
(McKelvey & Zavoina, 1975). Moreover, ordered logit reduces the number of
assumptions held by linear regression (i.e., homoscedascity and normal distribution of

errors) since the residuals produced by the model are assumed to be correlated with the

50

values of the dependent variable, thus producing more reliable estimates (W inship &
Mare, 1 984).

Binary logistic regression is another multivariate procedure that may be used to
analyze ordinal level response categories. For instance, the dependent variable level of
involvement in ﬁrearm offending may be collapsed into three separate binary logistic
regressions: moderate involvement (1) relative to no involvement (0), serious
involvement (1) relative to no involvement (0), and moderate involvement (1) relative to
serious involvement (0). While the creation of separate binary logistic regressions is
feasible, numerous ﬁrearm researchers (Bjerregaard & Lizotte, 1995; Cao, Cullen, &
Link, 1997; Lizotte et al., 1994) have expressed caution that the creation of separate
binary logistic regressions will systematically exclude at least one category of the
dependent variable unless two categories are combined into one variable that may or may
not be theoretically relevant to the research at hand (e. g., no involvement and moderate
involvement could be combined and coded 1 for analysis relative to serious involvement
coded 0). The pragmatic danger in excluding a single category of the dependent variable
is the reduction of the original sample size, which increases the potential for under or
over-estimated logit coefﬁcients due to inﬂations in standard error. Moreover, the binary
logistic regression, as presented above, produces three separate sets of regression
coefﬁcients that must be interpreted relative to their respective binary categories.
Ordered logit, on the other hand, simultaneously estimates multiple logits through the use
of all ordinal categories, which therefore reduces the standard error of prediction, and
produces one set of coefﬁcients (Agresti, 2002). As such, ordered logit is able to

efﬁciently produce parsimonious and powerful models without the loss of observed data

51

(Agresti, 1989; Agresti, 2002; Long, 1997; McKelvey & Zavoina, 1975; Winship &
Mare, 1984).

In order to produce simultaneous estimates or cumulative logits for all of the
ordinal categories of the dependent variable, ordered logit rests on the assumption of
proportional odds. Since the ordinal dependent variable values are considered arbitrary
thresholds of an underlying continuum of ﬁrearm offending seriousness, the y-intercepts
used for the maximum likelihood estimation should vary across the values of the
observed and predicted values of the dependent variable while the logit coefﬁcients for
each of the independent variables should remain the same for each of the maximum
likelihood estimates (Agresti, 1989; Long, 1997). Therefore, the proportional odds
assumption contends that the cumulative algorithms used to produce maximum likelihood
estimates (i.e., contrasts between no involvement, moderate involvement, and serious
involvement) are not signiﬁcantly different from estimates that would be obtained if the
algorithms were obtained separately (i.e., contrasts only made between any two of the
three categories of the dependent variable).

A Score Test using a chi-square distribution is commonly applied to test for the
assumption of proportional odds (Long, 1997). Ifnon-signiﬁcant, the odds ratios can be
interpreted as constant across all levels of the dependent variable since the maximum
likelihood lines of prediction will be essentially parallel. If the test is signiﬁcant,
however, the interpretation of odds ratios as constant may be erroneous as the maximum
likelihood lines of prediction are non—parallel with logit coefﬁcients taking on different

values across the distribution of the dependent variable. In the event that the assumption

52

of proportional odds is rejected, Long (1997) has suggested that multinomial logit models
should be used since they do not adhere to the assumption of proportional odds.
Multivariate Statistics

A number of the control independent variables were collapsed into dummy
variables for the ordered logit regression analysis. The dummy variable for the
respondent’s race (“black”) will examine the level of involvement in ﬁrearm offending
for black individuals relative to the reference category of all other races or ethnicities.
The selection of “black” as the racial or ethnic category was purposely made in order to
control for the frequency in whiCh black respondents completed surveys relative to other
racial or ethnic groups (see Table 4). The dummy variable “Detroit” will exarrrine the
level of involvement in ﬁrearm offending for individuals from the city of Detroit relative
to the reference category of all other geographic and lesser-populated locations within the
state of Michigan"). Finally, the respondent’s survey facility (“Institution 2”) will
examine the level of involvement in ﬁrearm offending for those individuals who were
surveyed in the institution providing the most surveys relative to the reference category
of the remaining institutions.

Variable transformations were also made in order to control for the effect of
missing data on the independent and dependent variables. Univariate statistics presented
earlier indicated that 0 to 23% of the distribution of any one independent or dependent
variable may be missing information. Due to the amount of missing cases, bivariate

results indicated a large variability in the sample sizes between any one relationship

 

’0 The logic behind the inclusion of a durrrrny variable speciﬁcally for Detroit is based upon the frequency
of violence found within the city and federal initiatives to control ﬁrearms through mandatory sentencing
(Heumann, Loﬁin, & McDowall, 1982; Lofu'n & McDowall, 1981; McDowall, 1991).

53

between a speciﬁed independent variable and the dependent variable. Moreover, the
reduction in sample size may be quite high when all of the independent variables are
added into a regression model. In order to maintain a sample size that will produce
reliable logit regression estimates, individual independent variables missing 5 or more
percent of their univariate distribution were selected for transformation based upon their
univariate statistics; they include: the control independent variable of “black,” the
socialization independent variables of hard drug use scale, level of neighborhood
exposure to violence scale, and gang membership, and the onset independent variables of
age of ﬁrst ﬁrearm ownership and age of ﬁrst arrest. For the remaining independent and
dependent variables missing 4 or fewer percent of their univariate distribution, the
missing cases were excluded from the multivariate analyses.

For the six variables selected for transformation, a new independent variable was
created that mirrored the distribution of the original variable, however missing data was
included into the overall distribution and coded 01 I. An additional missing data indicator
variable was created for each independent variable and coded 0 to reﬂect valid or non-
missing data and 1 to represent missing data (Brame & Paternoster, 2003). When the two
new variables are entered into the ordered logit regression analysis, “the missing data
[indicator] takes the uncertainty associated with the missing survey data into account and
adjusts the standard errors of the parameter estimates accordingly” (Brame & Paternoster,
2003, p. 74). Regression coefﬁcients produced for both the new independent variable
including the missing data and the missing data indicator variable will allow for the

comparison of signiﬁcant differences between missing and non-missing data on the

 

“ An example of the variable transformation is provided in Appendix C.

54

dependent variable. Most importantly, the transformation produces a constant sample
size of 453, which encompasses 90% of the original sample size.

Multivariate models are presented additively, with the addition of one new
theoretically relevant socialization variable (i.e., gang membership and drug market
participation) made to a baseline or restricted model (i.e., control, familial socialization,
peer socialization, and environmental socialization variables) in order to reach a ﬁll]
model that contains all the theoretically relevant variables. Additionally, secondary
models will be created that exchange primary variables measuring legitimized ﬁrearm
socialization inﬂuences (i.e., parents and peers own or possess ﬁrearms) with variables
measuring criminalized ﬁrearm socialization inﬂuences (i.e., parents and peers have been
arrested for ﬁrearm possession or use).

Socialization Modeling
Table 8 provides the multivariate results for socialization models including

variables of legitimized parental and peer ﬁrearm ownership or p0ssession.

55

Table 8: Firearm Involvement Regressed on Socialization Variables Including
Legitimized Firearm Ownership or Possession (n=453).

 

 

Model 1 Model 2 Model 3 Model 4

ML Estimates
Constant 3 .73 .43 -1.00 -1.02
(SE) (1.15) (1.18) (1.23) (1.23)
Constant 2 2.51“ 2.26 .98 .96
(SE) (1.16) (1.18) (1.23) (1.23)
Odds Ratios
R’s Age .81*** .82*** .84M .84”
BlackNew 1.25 1.31 1.28 1.29
MissBlack 1.56 1.54 1.36 1.40
Detroit 1.64 1.73* 1.82* 1.82*
Institution 2 .91 .93 .90 .91
Parents O/P 1.01 1.03 1.06 1.06
Peers O/P 375*" 339*" 2.36” 2.38“
V. ExposeNew 1.61*** 1.54*** 1.49“" 1.49“”
MissVE 645*" 5.51*** 4.56*** 449*“
GangNew 2.96"'M 253*" 2.51***
MissGang .85 .76 .75
Drug Sales Freq. 136*“ 136*“
H. Drug UseNew 1.02
MissI-IDU 1.16
Score Test )8 12.33 14.06 13.14 19.46

(9 dt) (11 df) (12dt) . (14 (if)
-2LL 763.02 741.06 702.53 702.30
LR )6 179.64*** 201 .61*** 240.13*** 240.36***

(9 df) (11 dl) (12 (11) (14 d1)

 

*p < 0.05, “p < 0.01, ***p < 0.001

Across all of the socialization models, the reduction of the —2 log likelihood (-
2LL) value and signiﬁcance of the likelihood ratio (LR) chi-squared test implies that each
of the independent variables used in the models provides an adequate measure of the
level of involvement in ﬁrearm offending. That is, the -2LL and LR chi-square statistics
reveal that the inclusion of the selected independent variables in the models provide an
enhanced ability to predict the levels of involvement in ﬁrearm offending than the
constant alone. Moreover, the two statistics reveal that at least one of the regression

coefﬁcients for the independent variables used in each model is non-zero.

56

A number of highly signiﬁcant independent variable predictors of involvement in
ﬁrearm offending were obtained (p < 0.001). Individuals who provided missing
responses on the variable of neighborhood exposure to violence relative to those who
provided a non-missing response were more likely to be involved in serious ﬁrearm
offending controlling all other variables across models. Interestingly enough, missing
responses on the variable of neighborhood exposure to violence provided the largest
effect of the variables considered across models. Gang membership increases the
likelihood of involvement in serious ﬁrearm offending by a factor of approximately 2.5 to
3 controlling for all other variables across models. Increased exposure to violence in
one’s neighborhood increases the likelihood for involvement in serious offending, while
frequent activity in drug sales was also found in increase the likelihood of involvement in
serious ﬁrearm offending.

Association with peers who own or possess ﬁrearms was highly signiﬁcant in
Models 1 and 2 and maintained a strong signiﬁcance (p < 0.01) through Model 4.
Controlling for all other variables across models, association with peers who own or
possess ﬁrearms increases the likelihood for involvement in serious ﬁrearm offending.
Similarly, respondent age was also found to be highly signiﬁcant in restricted models (p
< 0.001) and maintained signiﬁcance in the full model including all of the theoretically
relevant variables (p < 0.01). Considering that the sample was restricted to respondents
between the ages of 17 to 25, older respondents were less likely to be involved in serious
ﬁrearm offending. The dummy variable “Detroit” was also found to be signiﬁcant (p <
0.05) with the inclusion of additional theoretically relevant variables to the baseline

Model 1. Although the addition of variables to a misspeciﬁed ordered logit regression

57

model may increase the standard errors of the logit coefﬁcients and produce signiﬁcant
results, the signiﬁcance of the dummy variable “Detroit” in Model 4 is attributable to the
theoretical relevance of the predictors (Allen, 1997). Respondents from Detroit were
more likely to be involved in serious ﬁrearm offending. relative to the remainder of the
respondents who resided in other locations within the state of Michigan while controlling
for all other variables. The nonsigrriﬁcant logit coefﬁcients for parental ﬁrearm
ownership or possession and hard drug use indicate no relationship with involvement in
ﬁrearm offending.

A second set of socialization models (Models 5 through 8) were produced that
include variables of criminalized parental and peer ﬁrearm socialization in the form of

arrest or incarceration for ﬁrearm possession or use. Table 9 provides the results.

58

Table 9: Firearm Involvement Regressed on Socialization Variables Including
Criminalized Firearm Possession or Use (n=453).

 

 

Model 5 Model 6 Model 7 Model 8

ML Estimates
Constant 3 .96 .73 -1.02 -1.03
(SE) (1.13) (1.17) . (1.22) (1.22)
Constant 2 2.83" 2.66* 1.05 1.04
(SE) (1.14) (1.17) (1.22) (1.22)
Odds Ratios
R’s Age .82*** .82*** .86" .86“
BlackNew 1.12 1.19 1.14 1.15
MissBlack 1.20 1.18 1.07 1.09
Detroit 167* 1.76* 1.87* 1.88*
Institution 2 .88 .89 .87 .88
Parents M 1.59 1.38 1.75 1.76
Peers All 445*“ 420*" 332*" 333*"
V. ExposeNew 1.48*** 1.43*** 1.38*** 138*"
MissVE 4.56*** 3.96" 330* 3.24”
GangNew 293*” 2.43""""I 2.41“"
MissGang .85 .78 .77
Drug Sales Freq. 1.36*** 136*"
H. Drug UseNew 1.01
MissHDU 1.19
Score Test A? 9.13 10.58 11.05 18.12

(9 dt) (11 (If) (12 dt) (14 df)
-2LL 741.32 720.85 684.47 684.19
LR )8 201.34*** 221.82*** 258.19*** 258.47***

(9 (if) (11 df) (12 d1) (14 d1)

 

*p < 0.05, ”p < 0.01, *"p < 0.001

Overall, Models 5 through 8 that include criminalized parental and peer ﬁrearm
socialization variables mirror the prior socialization model including legitimized
variables. The —2LL and LR chi-square values indicate that each of the independent
variables used in the models provide an adequate measure of the level of involvement in
ﬁrearm offending. The effect of respondent age, neighborhood exposure to violence,
gang membership, and ﬁ'equency of drug sales is relatively similar to Models 1 through 4
in terms of odds ratio value, directionality, and signiﬁcance across all four of the models.

Moreover, the variables capturing parental ﬁrearm socialization and hard drug use were

59

again found to be insigniﬁcant. Therefore, controlling for all other variables in the
models, parental arrest or incarceration for ﬁrearm possession and use as well as hard
drug use are unrelated to involvement in ﬁrearm offending.

A few of the variables in the second set of models were slightly different from the
models including legitimized parental and peer ﬁrearm agents of socialization. Peer
association was again found to be a highly signiﬁcant predictor of involvement in ﬁrearm
offending and also provided the largest effect of the variables considered in the full
model (p < 0.001). That is, association with peers who had been arrested or incarcerated
for ﬁrearm possession or use increases the likelihood of involvement in serious ﬁrearm
offending by a factor of approximately 3 to 4.5 controlling for all other variables across
models. The prior variable providing the largest effect, missing responses on the variable
of neighborhood exposure to violence, was highly signiﬁcant in the reduced Model 1 (p <
0.001), but was found to reduce signiﬁcance in subsequent models (ranging from p <
0.05 to p < 0.01) that included additional theoretically relevant variables. Finally, the
dummy variable “Detroit” was consistently signiﬁcant (p < 0.05) across all of the models.
Summary of Socialization Modeling

Gang membership, increased exposure to neighborhood violence, and active
participation in drug sales were generally found to be strong predictors of serious
involvement in ﬁrearm offending across models that include legitimized and criminalized
parental and peer ﬁrearm ownership, possession, and use variables. Peer association,
measured both by association with peers who own or possess ﬁrearms and association
with peers who had been arrested or incarcerated for ﬁrearm possession or use, was also

found to be a signiﬁcant predictor of serious ﬁrearm offending involvement.

60

Based on these ﬁndings, there is strong evidence in support of Hypothesis 2
predicting a relationship between peer ﬁrearm ownership, possession, and use and
individual levels of involvement in ﬁrearm offending. Hypothesis 3 is strongly
supported, as the relationship between gang membership and individual levels of
involvement in ﬁrearm offending has been observed. The predicted relationship between
drug market participation (Hypothesis 4) and involvement in ﬁrearm offending is only
partially supported as the frequency of drug sale activity is found to be a signiﬁcant
predictor of involvement in levels of ﬁrearm offending, whereas the use of hard drugs is
insigniﬁcant. Hypothesis 5, predicting a relationship between the levels of violence
found within one’s environment and individual levels of involvement in ﬁrearm
offending, was also strongly supported. A relationship between parental ﬁrearm
ownership, possession, and use and involvement in levels of ﬁrearm offending was
unfounded. Therefore there is no evidence to accept Hypothesis 1.

Onset Socialization Modeling

Subsequent models consider the effect of onset age at ﬁrst ﬁrearm ownership
(Models 9 through 16) and ﬁrst arrest for any crime (Models 17 through 24) in
combination with the socialization variables used in the prior models. In order to avoid
the potential for variable confound and misspeciﬁed regression estimates for onset age
variables, the respondent’s age was removed from all onset analyses. Table 10 provides
the results for ownership onset and socialization models including variables of

legitimized parental and peer ﬁrearm ownership or possession.

61

Table 10: Firearm Involvement Regressed on Ownership Onset and Socialization
Variables Including Legitimized Firearm Ownership or Possession (n=453).

 

 

Model 9 Model 10 Model 11 Model 12
ML Estimates
Constant 3 -.98 -1.16 -2.00** -2.12**
(SE) (.79) (.79) (.82) (.83)
Constant 2 1.07 .95 .26 .13
(SE) (.79) (.79) (.81) (.82)
Odds Ratios
First FirearmNew .91* .92* .92* .92
MissFF .02*** 02*" .03*** .03***
BlackNew 1.52 1.59 1.55 1.57
MissBlack 1.61 1.62 1.45 1.52
Detroit 196* 2.02” 2.16" 2.18“
Institution 2 .70 .7 3 .72 .74
Parents O/P .66 .67 .72 .72
Peers O/P 233* 2.18* 1.62 1.67
V. ExposeNew 1.47*** 1.41 *** 138*" 1.37***
MissVE 683*" 5.72*** 4.77" 4.55”
GangNew 277*" 246*" 244*“
MissGang 1 .04 .98 .96
Drug Sales Freq. 1.33*** 132*"
H. Drug UseNew 1.01
MissHDU 1.58
Score Test )8 2034* 20.85“ 18.98 22.25

(10 df) (12 d1) (13 d1) (15 dt)
-2LL 693.63 676.38 646.66 644.94
LR x’ 249.04*** 266.28*** 296.00*** 297.72***

(10 (ll) (12 dt) (13 dt) (15 dt)

 

‘p < 0.05, ”p < 0.01, mp < 0.001

Models 9 through 12 provide an indication of model ﬁt as the —2LL values
decrease across models and LR chi-square values are signiﬁcant. Irnportantly however,
Models 9 and 10 are found to be in violation of the proportional odds assumption
according to the signiﬁcant Score Test chi-square values. As such, caution is needed in
the interpretation of the odds ratios since the logit lines of prediction are non-parallel with
regression coefﬁcients taking on different values across the distribution of the dependent
variable. Alternative multinomial logit regression models for these two models are

presented in Appendix D. Overall, the ordered and multinomial logit regression models

62

produced similar results in terms of the odds ratio coefﬁcient values and directionality.
Slight differences were found in terms of the signiﬁcance of the logit coefﬁcients for the
age of ﬁrst ﬁrearm ownership (only signiﬁcant in ordered logit models), the dummy
variable “Black” (only signiﬁcant in multinomial logit models), and the dummy variable
“Detroit” (only signiﬁcant in ordered logit models) which may be attributable to the
comparatively higher standard error values in multinomial logit models.

Gang membership, neighborhood exposure to violence, the frequency of drug
sales, and missing responses on the variable of age of ﬁrst ﬁrearm ownership were found
to be highly signiﬁcant predictors of involvement in ﬁrearm offending across the models
that were not in violation of the proportional odds assumption (p < 0.001). Gang
membership not only increased the likelihood of involvement in serious ﬁrearm
offending controlling for all other variables, it also provided the strongest effect for all
the highly signiﬁcant variables. Frequent activity in drug sales and increased exposure to
violence within one’s neighborhood increased the likelihood of involvement in serious
ﬁrearm offending. Individuals who provided missing responses on the variable of age of
ﬁrst ﬁrearm ownership were less likely to be involved in serious ﬁrearm offending,
which is a ﬁnding that seems to be somewhat intuitive. Individuals that have never
owned or possessed a ﬁrearm may not have indicated an age of ﬁrst ownership and
would be considered less at risk for involvement in ﬁrearm offending.

Missing responses on the variable of neighborhood exposure to violence reduced
in its level of signiﬁcance while the dummy variable “Detroit” gained in its level of
signiﬁcance with the addition of theoretically relevant variables. However, the increase

or decrease in levels of signiﬁcance may be due to the proportional odds assumption

63

violation and potential for erroneous regression coefﬁcient estimates (see Appendix D for
multinomial logit regression results). Models 11 and 12 indicate that both of the
variables were positively related to the dependent variable (p < 0.01). Individuals who
provided missing responses on the variable of neighborhood exposure to violence were
more likely to be involved in serious ﬁrearm offending, while those respondents from the
city of Detroit were also more likely to be involved in serious offending.

The age of ﬁrst ﬁrearm ownership was signiﬁcant (p < 0.05) in Models 9 through
11. Since the sample had been restricted to 17 to 25 year olds, presumably the ceiling
age in which an individual could own a ﬁrearm would be 25. Interpreting the odds ratio,
individuals who gained ownership of a ﬁrearm at older ages were less likely to be
involved in serious ﬁrearm offending. However, the variable was insigniﬁcant for the
ﬁnal model that includes all of the theoretically relevant variables. Peer association was
found to be signiﬁcant (p < 0.05) in the models that were found to be in violation of the
proportional odds assumption, but was not found to be signiﬁcant in Models 11 and 12
that did not violate the assumption (See Appendix D for multinomial logit regression
results). Finally, parental ownership or possession of ﬁrearms and hard drug use were
again found to be independent of ﬁrearm offending involvement.

Table 11 provides the results for alternative ownership onset and socialization
models (Models 13 through 16) that include variables of criminalized parental and peer

ﬁrearm socialization.

64

Table 11: Firearm Involvement Regressed on Ownership Onset and Socialization

Variables Including Criminalized Firearm Possession or Use (n=453).

 

 

Model 13 Model 14 Model 15 Model 16
ML Estimates
Constant 3 -.88 -1.07 -2.03** -2.12**
(SE) (.70) (.70) (.73) (.74)
Constant 2 1.29 1.16 .33 .24
(SE) (.70) (.70) (.65) (.73)
Odds Ratios
First FirearmNew .90** .90* .91 * .91*
MissFF .02*** .02*** .O3*** .O3***
BlackNew 1.55 1.62* 1.57 1.56
MissBlack 1.35 1.34 1.21 1.25
Detroit 1.98* 2.05" 2.20” 2.21“
Institution 2 .70 .73 .72 .75
Parents M 1.43 1.29 1.52 1.56
Peers M 4.17*** 399*" 3.25*** 3.28***
V. ExposeNew 1.34*** 128*” 1.26*** 126*"
MissVE 4.38" 3.70" 3.14* 298*
GangNew 2.70*** 238*" 2.36***
MissGang 1.13 1.07 1.07
Drug Sales Freq. 1.30*** 1.30***
H. Drug UseNew 1.00
MissHDU 1.61
Score Test 1’ 17.34 17.73 17.00 20.68

(10 df) (12 d1) (13 df) ' (15 (If)
-2LL 669.73 654.13 628.81 626.83
LR 16 272.93*** 288.53*** 313.86*** 315.83***

(10 df) (12 df) (13 d1) (15 df)

 

*p < 0.05, "p < 0.01, mp < 0.001

Models 13 through 16 are relatively similar to Models 9 through 12, with the —
2LL and LR chi-square values providing a good indication of model ﬁt. Comparatively,
models including criminalized parental or peer ﬁrearm socialization in the form of arrest
or incarceration for ﬁrearm possession or use were not found to be in violation of the
proportional odds assumption as the Score Test chi-square values were insigniﬁcant. The
effect of gang membership, neighborhood exposure to violence, frequency of drug sales,
the dummy variable “Detroit,” missing responses on the variable for age of ﬁrst ﬁrearm

ownership, and missing responses on the variable for neighborhood exposure to violence

65

are comparable to Models 9 through 12 that include legitimized ﬁrearm agents of
socialization in terms of odds ratio value, directionality, and signiﬁcance across all four
of the models. Moreover, variables capturing parental ﬁrearm socialization and hard
drug use were again found to be insigniﬁcant.

Unlike the previous model, however, peer association was found to be a highly
signiﬁcant predictor of involvement in ﬁrearm offending and provided the largest effect
of all the variables found to be highly signiﬁcant in the full model (p < 0.001). Thus,
association with peers who had been arrested or incarcerated for ﬁrearm possession or
use increased the likelihood of involvement in serious ﬁrearm offending, net all other
variables included in the models. The onset age of ﬁrst ﬁrearm ownership was also
negatively associated with involvement in ﬁrearm offending across all of the models.
Although the variable decreased in its level of signiﬁcance with the inclusion of
additional variables, the signiﬁcance remained throughout the ﬁnal model (p < 0.05).
Finally, Model 14 indicates a somewhat anomalous ﬁnding that has yet to be observed in
any of the previous models. Speciﬁcally, black individuals are more likely to be involved
in serious ﬁrearm offending relative to non-blacks. It is important to note, however, that
the effect for the dummy variable “black” is not found in Models 15 and 16 that
additively include all the theoretically relevant variables.

Table 12 provides the results for arrest onset and socialization models including
variables of legitimized parental and peer ﬁrearm ownership or possession. Whereas
Tables 10 and 11 presented the effect of the age at ﬁrst onset of ﬁrearm ownership, the
subsequent models will examine the age of ﬁrst arrest for any crime in conjunction with

socialization variables.

66

Table 12: Firearm Involvement Regressed on Arrest Onset and Socialization Variables
Including Legitimized Firearm Ownership or Possession Q1=453).

 

 

Model 17 Model 18 Model 19 Model 20
ML Estimates
Constant 3 -2.57*** -2.78*** -3.93*** -3.96***
(SE) (.74) (.75) - (.79) (.80)
Constant 2 -.84 -.98 -1.98** -2.01**
(SE) (.73) (.74) (.77) (.78)
Odds Ratios
First ArrestNew .94 .95 .97 .97
MissFA .19* .20* .44 .44
BlackNew 1.19 1.25 1.24 1.25
MissBlack 1.47 1.46 1.38 1.41
Detroit 168* 1.78* 1.84* 1.85*
Institution 2 .78 .81 .82 .82
Parents O/P 1.01 1.03 1.07 1.07
Peers O/P 4.05*** 3.63*** 2.47** 2.49"
V. ExposeNew 1.59*** 153*" 1.47*** l.47***
ETVMiss 643*" 5.35*** 4.40” 4.35"
GangNew 3.10*** 262*” 261*"
GangMiss 1.01 .88 .87
Drug Sales Freq. 137*" 1.37***
H. Drug UseNew 1.01
MissI-IDU 1.1 1
Score Test )7 14.83 16.33 15.06 21.48

(10 (if) (12 d1) (13 df) (15 df)
-2LL 772.98 749.45 709.95 709.84
LR )8 169.68*** 193.22*** 232.71*** 232.83***

(10 df) (12 d1) (13 dt) (15 dt)

 

sup < 005, “up < 0.01, asap < 0001

Models 17 through 20 again provide a good indication of model ﬁt for the

variables used according to the reduction in —2LL values and signiﬁcance of LR chi-

squares. Controlling for all the variables used, gang membership, neighborhood exposure

to violence, and frequency of drug sales were highly signiﬁcant predictors and positively

associated with involvement in ﬁrearm offending across all of the models (p < 0.001).

Additionally, gang membership provides the largest effect among the highly signiﬁcant

variables used in the ﬁnal model. The dummy variable “Detroit” was also found to be

67

signiﬁcant (p < 0.05), as individuals from the city of Detroit were more likely to be
involved in serious ﬁrearm offending.

The level of signiﬁcance was reduced for missing responses on the variable of
neighborhood exposure to violence and association with peers who own or possess
ﬁrearms with the inclusion of additional variables. However, both of variables remained
signiﬁcant and positively associated with involvement in ﬁrearm offending through
Model 20 (p < 0.01). Missing responses on the variable of age of ﬁrst arrest was found to
be signiﬁcant (p < 0.05) for Models 17 and 18, therefore indicating that individuals who
provided missing responses on the age of their ﬁrst arrest were less likely to be involved
in serious ﬁrearm offending. However, the variable was found to be insigniﬁcant in
models that include additional theoretically relevant variables.

The onset measure, age of ﬁrst arrest for any crime, and socialization measures,
parental ownership or possession of ﬁrearms and hard drug use, were found to be
insigniﬁcant and therefore unrelated to involvement in ﬁrearm Offending.

Criminalized parental and peer ﬁrearm socialization variables in the form of arrest
or incarceration for ﬁrearm possession or use are included for alternative models of arrest

onset and socialization. Table 13 provides the results.

68

Table 13: Firearm Involvement Regressed on Arrest Onset and Socialization Variables

Including Criminalized Firearm Possession or Use (n=453).

 

 

Model 21 Model 22 Model 23 Model 24
ML Estimates
Constant 3 -2.20** -2.39*** -3.79*** -3.82***
(SE) (.70) (.71) (.77) (.77)
Constant 2 -.36 -.49 -1.74* -1.78*
(SE) (.69) (.70) (.75) (.75)
Odds Ratios
First ArrestNew .95 .95 .98 .98
MissFA .25 .26* .57 .58
BlackNew 1.07 1.13 1.10 1.11
MissBlack 1.17 1.15 1.09 1.10
Detroit 1.71 * 1.81* 1.89* 1.89*
Institution 2 .76 .79 .79 .80
Parents M 1.93 1.62 2.06 2.07
Peers All 450*" 424*“ 336*" 3.37***
V. ExposeNew 1.47*** 1.41*** 1.37*** 1.37***
MissVE 450*" 3.85" 3.19“ 3.14*
GangNew 302*" 2.48*** 246*"
MissGang 1.00 .89 .88
Drug Sales Freq. 1.37*** 137*”
H. Drug UseNew 1.01
MissHDU 1.15
Score Test x2 10.71 11.80 12.19 19.49

(10 d1) (12 d1) (13 df) (15 dt)
-2LL 750.77 729.30 690.82 690.64
LR )8 191.89*** 213.36*** 251.85*** 252.02***

(10 df) (12 (If) (13 d1) (15 df)

 

*p < 0.05, ”p < 0.01, *“p < 0001

The model ﬁt statistics produce four signiﬁcant LR chi-square values across the
models. Gang membership, neighborhood exposure to violence, frequency of drug sales,
the dummy variable “Detroit,” and missing responses on the variable for neighborhood
exposure to violence produce similar odds ratio values, directionality, and levels of
signiﬁcance across all four of the models as compared to Models 17 through 20 that
utilize legitimized parental and peer ﬁrearm socialization variables. Moreover, the
variables capturing onset age of ﬁrst arrest for any arrest, parental ﬁrearm socialization,

and hard drug use were again found to be insigniﬁcant.

69

Compared to the previous regression results from Table 10, peer association
provided the largest effect of all the variables used and was also found to be a highly
signiﬁcant predictor (p < 0.001) when criminalized parental and peer ﬁrearm agents of
socialization are exchanged with legitimized variables. Missing responses on the variable
for onset age of ﬁrst arrest was only signiﬁcant in the restricted Model 22 and was not
found to be signiﬁcant in the subsequent full model.

Summary of Onset Socialization Modeling

Based on the ﬁndings, there is marginal to no support for Hypothesis 6 predicting
a relationship between the onset age at ﬁrst ﬁrearm ownership and ﬁrst arrest on
involvement in levels of ﬁrearm offending. The onset age of ﬁrst ﬁrearm ownership was
a signiﬁcant predictor of involvement across all the models when criminalized parental
and peer ﬁrearm possession or use variables were used. However, when legitimized
parental and peer ownership or possession variables were used, the onset age of ﬁrst
ﬁrearm ownership was signiﬁcant for only the restricted models and not the full model
including all the theoretically relevant variables. The age of ﬁrst arrest was insigniﬁcant
across all models.

Adding to the consistency of socialization modeling, the onset socialization
models strongly support Hypotheses 2, 3, and 5 and supports the drug sales premise of
Hypothesis 4. Onset socialization models produced strongly signiﬁcant predictors among
the variables of gang membership, exposure to neighborhood violence, ﬁ'equency of drug
sale activity, and association with peers who had been arrested or incarcerated for ﬁrearm
possession or use. Association with peers who own or possess ﬁrearms was also

signiﬁcant across all models including the onset measure of ﬁrst arrest, but was only

70

signiﬁcant in the restricted Models 9 and 10 that include the onset measure of ﬁrst
ﬁrearm ownership”. Hypothesis 1 and the drug use premise of Hypothesis 4 were

unsupported across all of the onset socialization models.

 

‘2 It is irrrportant to note that multinomial logit regression Models 9 and 10 (Appendix D) found the
likelihood for serious involvement in ﬁrearm offending to be increased by association with peers who
owned or possessed ﬁrearms

71

CHAPTER V: DISCUSSION

The identiﬁcation of individual offenders who utilize ﬁrearms for criminal
purposes and the prediction of speciﬁc offending patterns is an arduous process full of
inherent difﬁculty. The purpose of this research was to model theoretical propositions of
social learning theory with developmental or life course theory in order to examine
socialization pathways into a subculture that utilizes ﬁrearms for criminal purposes. This
research assumes that speciﬁc socialization mechanisms or onset behaviors place an
individual at a greater risk for initial and persistent involvement in ﬁrearm offending,
which may therefore lead to specialization in ﬁrearm offending patterns. Numerous
models were created that allowed for comparisons between socialization versus onset
models and legitimized (i.e., ownership or possession of ﬁrearms) versus criminalized
(i.e., arrest or incarceration for ﬁrearm possession or use) parental and peer ﬁrearm
socialization variables. Despite the creation of these comparative models, nearly every
model produced similar results.

In general, socialization or social learning variables measuring cultural and
environmental context served as the best predictors of involvement in serious ﬁrearm
offending. Variables measuring gang membership, peer association, and participation in
drug sales all suggest that differential association and socialization within distinct social
groups account for a signiﬁcant portion of serious ﬁrearm offending involvement.
Regarding peer association speciﬁcally, association with criminal peers provided the
strongest predictor of serious involvement in ﬁrearm offending. When the alternate

variable of association with peers who own or possess ﬁrearms was left in the models,

72

gang membership, neighborhood exposure to violence, and drug sales typically provided
larger effects than the measure of peer association.

In conjunction with the cultural agents of socialization, spatial socialization by
means of speciﬁc neighborhood environments places an individual more at risk for
involvement in ﬁrearm offending. The environmental socialization measure,
neighborhood level of exposure to violence, was signiﬁcant and positively related to
involvement in ﬁrearm offending across all of the models. Moreover, the control dummy
variable of “Detroit,” which was purposely selected due to the frequency of violence
found within the city of Detroit and federal initiatives to curtail ﬁrearm offending
(Heumann, Loﬁin, & McDowall, 1982; Loﬁin & McDowall, 1981; McDowall, 1991),
was also found to be signiﬁcant across all of the models. Informed by the work of
Wright and Rossi (1994), the environment inﬂuences the levels of involvement in ﬁrearm
offending and may produce ﬁrearm offenders. That is, individuals from hostile and
violent environments may become seriously involved in ﬁrearm offending due to the
increased need for self-preservation and the utility of a ﬁrearm to fulﬁll those needs
(Decker, Pennell, & Caldwell, 1997; Wright & Rossi, 1994).

There is a noteworthy, and unexpected, exception to the notion of cultural
socialization variables providing the best predictors of involvement in ﬁrearm offending.
Based upon the literature, the family has been identiﬁed as providing the initial
orientation into the mechanics and use of ﬁrearms among both criminal (Sheley &
Wright, 1995; Wright & Rossi, 1994) and non-criminal samples (Lizotte & Bordua,
1980; Lizotte, Bordua, & White, 1981; Marks & Stokes, 1976). Subsequently, the initial

familial exposure to ﬁrearms is considered to produce a ﬁrearm owning adult that has

73

been socialized into ﬁrearm ownership, possession, and use. The current results fail to
provide any support for the notion of parental socialization and its effect on involvement
in ﬁrearm offending. At the bivariate level, parental arrest or incarceration for ﬁrearm
possession or use was found to be signiﬁcant with a very weak measure of association.
When added to multivariate modeling controlling for the effect of other theoretically
relevant variables, the measure of parental arrest or incarceration for ﬁrearm possession
or use was found to be insigniﬁcant across all of the restricted and full models. The
measure of parental ownership or possession of a ﬁrearm was also insigniﬁcant across all
the bivariate and multivariate analyses. Therefore, the exchange of learned normative
ideologies and values between parent and child does not predispose an individual toward
involvement in ﬁrearm offending. Instead, the signiﬁcant agents of socialization that
shape involvement in ﬁrearm offending are beyond the immediate family - gangs, peers,
individuals involved in drug sales, and the larger environment as a whole.

Onset variables were marginal to non-existent predictors of involvement in
ﬁrearm offending in comparison to the socialization variables. The age of ﬁrst arrest
provided no support for the prediction of involvement in ﬁrearm offending as the
bivariate and multivariate modeling results found the variable insigniﬁcant across all of
the analyses. On the other hand, the age of ﬁrst ﬁrearm ownership was found to be a
signiﬁcant predictor of serious involvement in restricted models that controlled for
legitimized parental or peer ﬁrearm ownership or possession and across all models that
controlled for criminalized parental or peer ﬁrearm possession or use. Thus, there is
some evidence that younger ages of ﬁrst ﬁrearm ownership increased the likelihood of

serious involvement in ﬁrearm offending. Inadvertently, the variable of respondent age

74

validated the onset age of ﬁrst ownership ﬁndings. Across all of the socialization
models, older respondents (ceiling limit of 25) were signiﬁcantly less likely to be
involved in serious ﬁrearm offending, while younger respondents (ﬂoor limit of 17) were
more likely to be involved in serious offending.

Extending developmental theory to the current research context (Hirschi &
Gottﬁedson, 1983; Mofﬁtt, 1993; Sampson & Laub, 1992), those individuals who have
obtained their ﬁrst ﬁrearm at earlier ages would have more of an opportunity to utilize
their ﬁrearm for legitimate or criminal purposes with the progression of age through the
theorized peak offending period, which may explain the negative relationship between
the onset variable of age at ﬁrst ﬁrearm ownership and the control variable of respondent
age on the level of involvement in ﬁrearm offending. Of interest to criminal justice
policy is this population; those individuals that have obtained a ﬁrearm early in life and
have committed crimes involving ﬁrearms prior to or during adolescence since they are
most at risk for continued ﬁrearm offending (Mofﬁtt, 1993).

Missing Variable Responses

In general, literature has suggested that prisoner self reports can provide valid and
reliable measures of prior criminal behavior, especially compared to the information
captured in ofﬁcial arrest and conviction reports (Marquis, 1981). However, some degree
of incomplete information is often captured in self-reports since the respondents have the
opportunity to selectively answer questions or reactively chose not to answer sensitive
questions (Singleton & Straits, 1999). In the current research, a number of unintended
results occurred that are of relevance to the methodological utility of prisoner self-report

survey data. Missing responses on the variable of neighborhood exposure to violence, on

75

the variable of age at ﬁrst ﬁrearm ownership, as well as on the variable of age at ﬁrst
arrest were signiﬁcant predictors in socialization and onset socialization modeling. Ad
hoc bivariate analyses through the use of AN OVA were conducted in order to observe
any patterns in the missing data that may bias results. The results are provided in
Appendix E.

Missing responses on the variable of neighborhood exposure to violence increased
the likelihood of involvement in serious ﬁrearm offending across all of the socialization
and onset socialization models. The ad hoc analyses found ﬁve signiﬁcant relationships
with the missing responses. On average, black respondents were more likely to have
provided a missing response relative to all other races or ethnicities and the onset age of
ﬁrst ﬁrearm ownership was approximately two years younger for the respondents that
provided a missing response. The remaining variables suggest that those respondents
who provided missing responses for the variable of neighborhood exposure to violence
also provided omissions for race or ethnicity, gang membership, and the age of ﬁrst
ﬁrearm ownership. Therefore, there is reason to believe that certain respondents only
partially completed the survey or selectively chose to answer non-personal questions.

Missing responses on the variable of ﬁrst ﬁrearm ownership decreased the
likelihood of involvement in ﬁrearm offending. As mentioned previously, respondents
who have never owned or possessed a ﬁrearm may not have indicated an age of ﬁrst
ownership but rather glanced over the question. Seven variables were signiﬁcantly
related to missing responses for the ad hoc analyses. Black respondents were again more
likely to have provided a missing response relative to all other races or ethnicities.

Respondents were also more likely to have provided a missing response on the variable

76

of neighborhood exposure to violence if they failed to respond to the variable of ﬁrst
ﬁrearm ownership. Beyond these two relationships, the remaining ﬁve variables
reinforced the notion that the respondents who provided missing responses were more
likely to have lacked involvement in ﬁrearm offending based upon the ﬁndings. Those
who have provided missing responses were also less likely to have parents who owned or
possessed ﬁrearms, parents who had been arrested or incarcerated for ﬁrearm possession
or use, associations with peers who owned or possessed ﬁrearms, been involved in gangs,
and less exposure to violence within their neighborhoods relative to those who provided a
non-missing response. Therefore, respondents that chose not to provide a response on the
variable of ﬁrst ﬁrearm ownership do not appear to be systematically shielding patterns
of criminal behavior. Instead, they appear to have skipped the age of ﬁrst ﬁrearm
ownership due to its irrelevance with their background.

Finally, missing responses on the onset variable of ﬁrst arrest provided one
signiﬁcant relationship. Those respondents that have provided ‘a missing response on the
age of ﬁrst arrest were less likely to be involved in frequent drug sale activity. Unlike the
previous two variables capturing missing responses, the variable capturing missing
responses on the variable of ﬁrst arrest does not provide any room for speculation as to
why some respondents opted not to respond.

Study Limitations

The effect of non-response was omnipresent throughout the research. Bivariate
analyses produced variable sample sizes that were contingent upon the frequency of
missing cases found within the control, independent, and dependent variables. Variable

transformations were made to the multivariate models in order to control for the effect of

77

missing cases. However, the creation and addition of two similar measures into a
multivariate model induces additional interrelation or multicollinearity among the
predictors that existed prior to the variable transformation”. As a result, the multivariate
models suffer from rnisspeciﬁcation and result in biased logit coefﬁcients from which
inference and results were discussed (Fox, 1991). Although ad hoc analyses suggested
that missing cases were a product of general omission by a proportion of the respondents
as well as omission on a speciﬁc variable by respondents least likely to be involved in
ﬁrearm offending, the prevalence of missing cases adds a degree of ambiguity to the
validity of the results.

Future research may wish to include alternate measures of neighborhood exposure
to violence and drug use. 111 the current research, the measure of neighborhood exposure
to violence scale blended survey questions that could be separated into theoretical
measures of victimization and perceptions of violence. That is, there may be a strong
relationship between the measure of prior victimization and invOlvement in ﬁrearm
offending and an insigniﬁcant relationship between the perception of ﬁrearm density in
the respondent’s neighborhood and need for protection. Alternatively, the perceptual

measures may provide the strong relationship with involvement in ﬁrearm offending

 

'3 Correlation matrices, tolerance and variance inﬂation factors (VIF), as well as collinearity diagnostics
were produced for all of the control, socialization, and onset independent variables used in the full
multivariate analyses (Models 4, 8, 12, 16, 20, and 24). All of the relationships between the six new
variables created that include missing information and the corresponding six missing data indicators were
signiﬁcant (p < .001) ranging in absolute value from r = .20 to r = .92, tolerance and VIP factors failed to
approximate the value 1, and the condition index and variance proportions did not load strongly on a single
concept. Beyond the six new variables, multicollinearity amongst the remaining independent variables
continued to be problematic. There were 36 signiﬁcant relationships (p < .001) ranging in absolute value
from r = .14 to r = .43, 10 relationships (p < .01) ranging in absolute value from r = .11 to r = .14 , and 13
relationships (p < .05) ranging in absolute value from r = .05 to r = .12. Tolerance factors, VIF factors,
condition index, and variance proportions values suggest that the remaining variables are not providing a
unique variation and may be measuring the same or larger construct (e.g., ﬁrearm offending subculture or
lifestyle).

78

while the victimization measures are found to be insigniﬁcant. Either way, the
combination of the two constructs into one scale may produce signiﬁcant ﬁndings when
one of the measures provides more of an effect than the other. Therefore, future research
should attempt to control for the effect of victimizatiOn on perceptual measures.

Previous research involving ﬁrearm ownership, possession, and use has suggested
that drug use is an insigniﬁcant predictor of involvement in ﬁrearm activity (Bjerregaard
& Lizotte, 1995; Decker, Pennell, & Caldwell, 1997; Lizotte et al., 2000; Lizotte et al.,
1994; Sheley & Wright, 1995; Treatment Alternatives for Safe Communities, 1997).
Drug use in this research has been operationalized as use of harder or more illicit drugs
(i.e., experimentation with heroin, crack, cocaine, PCP, and methamphetamines).
Individuals who are frequent users or abusers of “softer” drugs (e.g., marijuana and
alcohol) may be associated with increased levels of involvement in ﬁrearm offending,
since the types of drugs used and frequency of use has been associated with persistent
offending (Chaiken & Chaiken, 1990; Nurco et al., 1988). Future research should
incorporate measures of soft drug use for comparison purposes with hard drugs in order
to observe their effect on involvement in ﬁrearm offending.

Finally, the results are also limited in their generalizability. The sample includes
a limited age group of 17 to 25 year old males from Michigan that have been incarcerated
during or relatively soon after their theorized peak offending period of mid to late teens
(Hirschi & Gottfredson, 1983). Based on this notion, the sample may be biased toward
the severe end of an offending continuum and should not be considered representative of
all 17 to 25 year old males in Michigan nor all institutionalized persons.

Conclusions and Policy Recommendations

79

The ﬁndings of this research suggest that the nexus between gang membership,
associations with peers who own or possess ﬁrearms, associations with peers who had
been arrested or incarcerated for ﬁrearm possession or use, drug sales, and neighborhood
exposure to violence increases the risk for serious involvement in ﬁrearm offending.
When these cultural and environmental socialization agents converge upon a youthful
individual who may or may not have already obtained their ﬁrst ﬁrearm, there is a
signiﬁcant potential for the individual to begin, persist, or increase their level of
involvement in ﬁrearm offending. They key for criminal justice policy and practices is to
reduce the ﬁequency and proportion of crime involving ﬁrearms among at risk
populations in order to reduce the overall frequency of crime within an area.

Extrapolating ﬁ'om the ﬁndings of the current research, there is an implicit notion
that standalone general deterrence strategies may have marginal effects. The
respondents, by virtue of being incarcerated in a state institution, have not been deterred
from prior criminal involvement. Moreover, the respondent’s associates also do not
appear to be deterred from participation in criminal activity. A large majority of the
sample, approximately 72% (see Table 2), has some degree of association with peers who
have been arrested or incarcerated for ﬁrearm possession or use. Therefore, a targeted or
directed approach is needed.

At the ﬁont end of the criminal justice system, problem-orientated policing
strategies through the identiﬁcation of “hot spots,” or spatial locations within a
community or neighborhood that are known to produce a disproportionate amount of
violent crime, have been used to combat criminogenic areas prone to ﬁrearm violence

that include individuals most at risk for involvement in ﬁrearm (Braga et al., 1999;

80

McGarrell et al., 2001; Sherman & Rogan, 1995), gang (Kennedy, Piehl, & Braga, 1996;
Braga et al., 2001), and drug sale (Weisburd & Green, 1995) activities”. Optimally, a
layered approach in the identiﬁcation of a concentrated ﬁrearm hot spot is desired; where
the rates of ﬁrearm violence, gang activity, and drug markets within a community or
neighborhood are highest. From the identiﬁcation of the hot spot, a directed effort could
be made to increase police presence in the targeted areas and enforce ﬁrearm carrying
statutes in order to decrease the frequency of ﬁrearm carrying among the infrequent and
routine ﬁrearm carriers.

At one end of the continuum, the directed effort should deter the infrequent carrier
as the certainty of arrest and perception of risk increases (Braga et al., 1999), which
therefore may reduce the potential for opportunistic violent crimes involving ﬁrearms
(Kleck, 1997). Since the infrequent carrier also will be deterred from carrying a ﬁrearm
for protection, the likelihood for entering portions of the ﬁrearm hot spot area and
participation in high-risk activities (e. g., drug sales and gang activity) is decreased. At
the other end of the continuum, the routine ﬁrearm carrier may also be deterred from the
frequency in which they have previously carried ﬁrearms due to the increased police
presence. However, there is some suggestion that the routine carrier’s needs for
protection outweigh the costs of apprehension (Wright & Rossi, 1994). In these
instances, the enforcement of ﬁrearm carrying statutes in high crime areas may serve the
purpose of incapacitating those individuals most likely to utilize their ﬁrearm for criminal

purposes (McGarrell et al., 2001). That is, the incapacitation effect through the use of

 

'4 It is inrportant to note that Kennedy, Piehl, and Braga (1996) do not consider Operation Ceaseﬁre to be a
hot spot intervention although the acknowledgement of concentrated gang areas drove the research
methodology. The authors note, “unlike most efforts to attack such [gang] hot spots through place-focused
interventions, our network-focused intervention will not necessarily be aimed at hot spots” (p. 181).

81

ﬁrearm carrying statutes may remove the routine carrier through formal criminal justice
processing beginning at arrest and leading to incarceration for a set number of years.
Figuratively, the incapacitation effect may only lead to arrest and ﬁrearm seizure, which
would therefore “incapacitate” the routine carrier ﬁom having the ability to utilize their
ﬁrearm. Although routine carriers are generally able to procure a new ﬁrearm in a
relatively short amount of time (Wright & Rossi, 1994), the time lapse between arrest and
ﬁrearm seizure to release back into the community may provide enough of a “cooling-
off” period to lessen the probability of committing a ﬁrearm involved crime.

Criminal justice policy and practices at the tail end of the criminal justice system
may also provide a focused deterrent effect for individuals that have been previously
involved in ﬁrearm, gang, and drug sale activity. In large part however, the general idea
is similar to that of problem-orientated policing of ﬁrearm hot spots — deter the marginal
ﬁrearm carrier and deter or incapacitate the routine carrier who is most likely to be
involved in future ﬁrearm offending. Utilization of “lever pulling” strategies (Kennedy,
Piehl, & Braga, 1996) and a network of partnerships across the law enforcement,
prosecutorial, and correctional institutions within a community could enhance the parole
supervision efforts for known ﬁrearm offenders, gang members, and drug sellers in
conjunction with mandatory state or federal penalties for new offenses and technical
offenses involving ﬁrearms (see Project Safe Neighborhoods, 2004). Through the pre-
parole identiﬁcation of prisoners who have committed a ﬁrearm, gang, or drug sales
offense leading to incarceration and the increased communication and subsequent action
alluding to the notion that ﬁrearm offending will not be tolerated, the use of enhanced

parole supervision may provide enough of a deterrent effect to reduce the likelihood of

82

ﬁrearm offending among the most at risk population for ﬁrture involvement. When the
deterrence strategies fail to deter, incapacitative efforts have the ability to revoke parole
supervision and place at risk persons back into institutional supervision.

In the larger picture, the criminal justice policies and practices discussed should
not been seen as panaceas. Policy and practices regarding ﬁrearms will continue to
change with the progression of time and evaluative efforts may reinforce or dismantle the
utility of various federal, state, and local initiatives. As suggested by Kleck (1997),
“solving the [ﬁrearm] violence problem will have to involve reductions in economic
inequality, injustice, and the social disorder these generate” (p. 396). Therefore, until
larger social change is made to at risk communities, the maintenance of criminal justice

policy and practices regarding ﬁrearm violence will continue to provide temporary relief.

83

APPENDIX A

ITEMS AND RELIABILITY FOR SOCIALIZATION INDEPENDENT VARIABLE

 

 

SCALES
Table 14: Items and Reliability for Hard Drug Use Scale (n=433).
Hard Drug Use Scale Items Mean Alpha if Item Removed

1. Have you ever tried heroin? .12 .79
2. Have you ever tried crack? .12 .78
3. Have you ever tried cocaine? .33 .77
4. Have you ever tried PCP? .16 .76
5. Have you ever tried metharnphetarnines? .15 .77

Reliability Coeﬂicient (Alpha) = .81

 

Table 15: Items and Reliability for Neighborhood Exposure to Violence Scale (n=455).

 

 

Neighborhood Exposure to Violence Scale Items Mean Alpha if Item

Removed
1. Ever been threatened with a ﬁrearm? .76 .68
2. Ever been shot at? .76 .67
3. Ever been injured by gunshot? ' .30 .73
4. Ever been injured by another weapon? .54 .72
5. Ever been beaten up? .62 .73
6. Ever been robbed? .44 .73
7. In your neighborhood, lots of guns on the street? .71 .69
8. In your neighborhood, important to have gun for .52 .70
protection?

Reliability Coeﬂicient (Alpha) = .73

 

84

APPENDD( B

BIVARIATE CONTINGENCY TABLES
Separate chi-square contingency tables are provided for all of the control and independent
variables. Tables 16 through 19 present the results Of the control variables, Tables 20
through 27 present the results of socialization variables, and Tables 28 and 29 present the
results of onset variables on the level of involvement in ﬁrearm offending. Summary
statistics for the bivariate relationships are provided in the results section.
Control Variables

Table 16: Bivariate Relationship of Respondent Age and Involvement in Firearm
Offending.

 

Involvement in Firearm Offending

 

Respondent No Moderate Serious Row Totals
Age Involvement Involvement Involvement
17-20 56 76 146 278
(20%) (27%) (53%)
21-25 56 69 87 212
(26%) (33%) (41%)
Column Totals 112 145 233 490

 

)8 (2, n = 490) = 6.506, a = .04

85

Table 17: Bivariate Relationship of Respondent Race and Involvement In F irearrn

 

 

 

Offending.
Involvement in Firearm Offending
Respondent No Moderate Serious Row Totals
Race Involvement Involvement Involvement

Black 40 70 121 231
(17%) (30%) (52%)

White 56 50 71 177
(32%) (28%) (40%)

Hispanic 4 8 20 32
(13%) (25%) (63%)

Other 9 1 1 10 30
(30%) (3 7%) (33%)

Colmnn Totals 109 139 222 470

 

)8 (6, n = 470) = 17.946, a= .01

Table 18: Bivariate Relationship of Respondent’s Population of Previous Residence and

Involvement in Firearm Offending.

 

Involvement in Firearm Offending

 

Pop. of No Moderate Serious Row Totals
Residence Involvement Involvement Involvement

<50,000 59 56 68 183
(32%) (31%) (37%)

50,001 -1 00,000 19 24 42 85
(22%) (28%) (49%)

100,001 -250,000 14 3O 44 88
(16%) (34%) (50%)

250,001+ 17 32 80 129
(13%) (25%) (62%)

Column Totals 109 142 234 485

 

)8 (6, n = 485) = 25.945, a: .00

86

Table 19: Bivariate Relationship of Respondent’s Survey Facility and Involvement in
Firearm Offending.

 

Involvement in Firearm Offending

 

Facility No Moderate Serious Row Totals
Involvement Involvement Involvement

Institution 1 15 32 64 l 11
(14%) (29%) (5 8%)

Institution 2 74 80 131 285
(26%) (28%) (46%)

Institution 3 23 34 40 97
(24%) (35%) (41%)

Column Totals 112 146 235 493

 

)8 (4, n = 493) = 10.064, a= .04

Socialization Independent Variables

Table 20: Bivariate Relationship of Parental Ownership or Possession of Firearms and
Involvement in Firearm Offending.

 

Involvement in F irearrn Offending

 

Parents O/P No Moderate Serious Row Totals
Involvement Involvement Involvement
No 72 85 137 294
(25%) (29%) (47%)
Yes 40 61 98 199
(20%) (31%) (49%)
Column Totals 1 12 146 235 493

 

)8 (2, n = 493) = 1.302, a: .52

87

Table 21: Bivariate Relationship of Parental Arrest or Incarceration for Firearm
Possession or Use and Involvement in Firearm Offending.

 

Involvement in Firearm Offending

 

Parents All No Moderate Serious Row Totals
Involvement Involvement Involvement
No 1 10 139 21 1 460
(24%) (30%) (46%)
Yes 2 7 23 32
(6%) (22%) (72%)
Column Totals 1 12 146 234 492

 

)7 (2, n = 492) = 9.056, a = .01

Table 22: Bivariate Relationship of Peer Ownership or Possession and Involvement in
Firearm Offending.

 

Involvement in Firearm Offending

 

Peers O/P No Moderate Serious Row Totals
Involvement Involvement Involvement
No 37 21 5 63
(59%) (33%) (8%)
Yes 72 1 17 228 417
(17%) (28%) (55%)
Column Totals 109 138 233 480

 

)7 (2, n = 480) = 66.599, 01 = .00

88

Table 23: Bivariate Relationship of Peer Arrest or Incarceration for Firearm Possession
or Use and Involvement in Firearm Offending.

 

Involvement in Firearm Offending

 

Peers NI NO Moderate Serious Row Totals
Involvement Involvement Involvement
No 69 47 18 134
(52%) (35%) (13%)
Yes 39 92 216 347
(l 1%) (27%) (62%)
Column Totals 108 139 234 481

 

{(2, n=481)= 119.564, a= .00

Table 24: Bivariate Relationship of Gang Membership and Involvement in Firearm
Offending.

 

Involvement in Firearm Offending

 

Gang No Moderate Serious Row Totals
Membership Involvement Involvement Involvement
No 99 107 118 324
(31%) (33%) (36%)
Yes 8 29 104 141
(6%) (21%) (74%)
Column Totals 107 136 222 465

 

Y (2, n = 465) = 60.336, 01 = .00

89

Table 25: Bivariate Relationship of Hard Drug Use and Involvement in Firearm

 

 

 

Offending.
Involvcmant in Firearm Offending
Hard Drug Use No Moderate Serious Row Totals
Scale Involvement Involvement Involvement

0 — Never Used 69 82 1 12 263
(26%) (31%) (43%)

1 13 14 37 64
(20%) (22%) (5 8%)

2 5 10 20 35
(14%) (29%) (57%)

3 2 9 18 29
(7%) (31%) (62%)

4 4 7 7 18
(22%) (39%) (39%)

5 — Used All 2 8 8 18
(11%) (44%) (44%)

Column Totals 95 130 202 427

 

{(10, n=427) = 15.359, 0:: .12

90

Table 26: Bivariate Relationship of Drug Sales Frequency and Involvement in Firearm

 

 

 

Offending.
Involvement in Firearm Offending
Drug Sales No Moderate Serious Row Totals
Frequency Involvement Involvement Involvement
l-Never Sold 60 44 27 131
(46%) (34%) (21%)
2 13 12 7 32
(41%) (38%) (22%)
3 8 9 5 22
(36%) (41%) (23%)
4 7 10 14 31
(23%) (32%) (45%)
5 2 18 27 47
(4%) (38%) (57%)
6-Every Day 19 43 151 213
(9%) (20%) (71%)
Column Totals 109 136 23 1 476

 

)7 (10, n = 476) = 123.148, a= .00

91

Table 27: Bivariate Relationship of Neighborhood Exposure to Violence and
Involvement in Firearm Offending.

 

Involvement in Firearm Offending

 

Exposure to No Moderate Serious Row Totals
Violence Involvement Involvement Involvement

O-No Exposure l9 8 2 29
(66%) (28%) (7%)

1 13 4 O 17
(77%) (24%) (0%)

2 22 15 5 42
(52%) (36%) (12%)

3 7 18 14 39
(18%) (46%) (36%)

4 15 24 16 55
(27%) (44%) (29%)

5 12 21 49 82
(15%) (26%) (60%)

6 9 20 57 86
(11%) (23%) (66%)

7 4 13 51 68
(6%) (19%) (75%)

8 — Max. Exposure 1 5 25 31
(3%) (16%) (81%)

Column Totals 102 128 219 449

 

)7 (16, n = 449) = 157.912, a= .00

92

Onset Independent Variables

Table 28: Bivariate Relationship of Age at First Firearm Ownership and Involvement in
Firearm Offending.

 

Involvement in Firearm Offending

 

First Firearm No Moderate ' Serious Row Totals
Involvement Involvement Involvement
0-13 14 44 117 175
(8%) (25%) (67%)
14+ 28 73 108 209
(13%) (35%) (52%)
Column Totals 42 1 17 225 384

 

)7 (2, n = 384) = 9.277, a= .01

Table 29: Bivariate Relationship of Age at First Arrest and Involvement in Firearm
Offending

 

Involvement in F Hem Offending

 

First Arrest - No Moderate Serious Row Totals
Involvement Involvement Involvement
0-14 3 1 60 93 1 84
(17%) (33%) (51%)
15+ 73 77 130 280
(26%) (28%) (46%)
Column Totals 104 137 223 464 '

 

)7 (2, n = 464) = 5.587, a: .06

93

APPENDD( C
VARIABLE TRANSFORMATION ILLUSTRATION
Table 30 provides an illustrative table regarding the transformation process of the
gang membership variable. I

Table 30: Gang Membership Variable Transformation

 

 

Gangrnemb Gangnew MissGang
Non-Gang (O) 295 (65%) 319 (70%) 429 (95%)
Gang Member (1) 134 (30%) 134 (30%) 24 (5%)
Missing 24 (5%) 0 0
Total 453 453 453

 

With the removal of missing cases from the variables missing 4 or fewer percent
of their univariate distribution, the original distribution of the dichotomous variable gang
membership (variable “gangmemb”) was reduced to 295 non-gang members coded 0, 134
gang members coded 1, and an additional 24 missing responses. A new variable,
“gangnew,” was created that included the 24 missing cases from the original variable
“gangrnemb.” Speciﬁcally, the 24 missing cases were coded O and added to the original
“gangrnem ” distribution of 295 resulting in a total “gangnew” frequency of319 persons
coded 0. In order to control for the effect of adding the missing cases into the variable
“gangnew,” a missing data indicator variable (“missgang”) was created and coded 0 for
non-missing cases and 1 for missing cases. The two new variables created, “gangnew”
and “missgang,” are then entered into the ordered logit regression model as independent

variables or predictors.

94

APPENDD( D
MULTINOMIAL LOGIT REGRESSION RESULTS
Table 31 provides the multinomial logit regression results for onset socialization
Models 9 and 10 that examine the effect of the onset age at ﬁrst ﬁrearm ownership.
Again, Models 9 and 10 were found to be in violation of the proportional odds
assumption. Therefore, alternate models were produced that did not include the
constraint of the proportional odds assumption (Long, 1997).

Table 31: Multinomial Logit Regression Odds Ratios for Models 9 and 10 in Violation
of the Proportional Odds Assumptionf (n=45 3).

 

Model 9 Model 10
Moderate Serious Moderate Serious
Involvement (2) Involvement (3) Involvement (2) Involvement (3)
Odds Ratios Odds Ratios Odds Ratios Odds Ratios

 

ML Estimates

Constant (SE) -.62 (1.17) -1.67 (1.33) -.52 (1.17) -1.75 (1.35)
Odds Ratios

F. FirearmNew 1.07 .93 1.06 .93
FFMiss .20 .01*** .18 .01***
BlackNew 3.07** 2.74* 3.16" 2.86*
RaceMiss 4.88 4.34 4.62 4.02
Detroit .84 1 .90 .81 1.97
Institution 2 .59 .54 .57 .55
Parents O/P .72 .52 .73 .51
Peers O/P 1.32 6.94" 1.26 6.66”
V. ExposeNew 1.18* 1.81*** 1.15 1.70**"‘
ETVMiss 2.49 21.43“” 2.21 15.89***
GangNew 2.77* 5.73***
MissGang 1.19 1.15
-2LL 670.48 653.65

LR )7 272.18*** 289.01***

(20 df) (24 df)

 

TNO involvement in ﬁrearm offending (1) is held as the reference category.
‘p < 0.05, "p < 0.01, ”*p < 0.001

For the multinomial logit Model 9, black respondents and increased exposure to

neighborhood violence increased the likelihood of moderate and serious involvement in

95

ﬁrearm offending. Association with peers who owned or possessed ﬁrearms and missing
responses on the variable of age at ﬁrst ﬁrearm ownership and the variable of
neighborhood exposure to violence also increased the likelihood for serious involvement
in ﬁrearm offending. The onset measure of age at ﬁrst ﬁrearm ownership and the
socialization measure of parental ownership or possession of ﬁrearms did not inﬂuence
moderate and serious involvement. Although insigniﬁcant, the likelihood for moderate
involvement in ﬁrearm offending was decreased for respondents from Detroit, while the
likelihood for serious involvement was increased. Since the odds ratios differ in
directionality, the dummy variable “Detroit” is speculated to be the variable producing
the violation of the proportional odds assumption.

For multinomial logit Model 10, black respondents and gang membership
increase the likelihood of moderate and serious involvement. Association with peers who
own or possess ﬁrearms, neighborhood exposure to violence, and missing responses on
the variable of age at ﬁrst ﬁrearm ownership and the variable of neighborhood exposure
to violence increased the likelihood of serious involvement. Once again, the onset
measure of ﬁrst ﬁrearm ownership and measure of parental socialization were found to
be insigniﬁcant, while the directionality for the insigniﬁcant dummy variable “Detroit”

differed in directionality between moderate and serious involvement”.

 

'5 Multinorrrial logit Model 9 and 10 also produced logit coefﬁcients for the variable of age of ﬁrst ﬁrearm
ownership that differed in directionality with the likelihood for moderate involvement increased for those
respondents who have obtained their ﬁrst ﬁrearm at older ages, while the likelihood for serious
involvement decreased with older ages. However, the close approximation to 1 for each of the odds ratios
in Models 9 and 10 suggest a miniscule difference between moderate or serious involvement and the
reference category of no involvement.

96

APPENDD( E
AD HOC BIVARIATE ANOVA RESULTS
Tables 32 through 34 provide the ad hoc bivariate AN OVA analyses for missing
responses on the variable of neighborhood exposure to violence (Table 32), the variable
of age at ﬁrst ﬁrearm ownership (Table 33), and the variable of age at ﬁrst arrest for any
crime (Table 34). For all of the analyses, the missing response variable was entered as an
independent variable in order to gauge the effect the missing variable has on the
remaining variables used to predict involvement in ﬁrearm offending. The ANOVA
results are not presented for the new variable upon which the missing response indicator
was created since the two measures are highly correlated. That is, the results of the
missing responses on the variable of neighborhood exposure to violence will not include
the predictor of neighborhood exposure to violence, missing responses on the variable of
age at ﬁrst ﬁrearm ownership will not include the predictor of age at ﬁrst ownership, and
missing responses on the variable of age at ﬁrst arrest will not include the predictor of
age at ﬁrst arrest for any crime. An interpretation of the results is provided in the

discussion section.

97

Table 32: Bivariate ANOVA of Missing Responses on the Variable of Neighborhood
Exposure to Violence on Predictors (n=453).

 

Non-missing Values Missing Values

 

(n = 422) (n=31)

Dependent Variable F Mean (s.d.) Mean (s.d.)
R’s Age .82 20.41 (1.88) 20.10 (1.64)
BlackNew 4.44* .45 (.50) .65 (.49)
MissBlack 5.18* .04 (.20) .13 (.34)
Detroit 3.25 .24 (.43) .39 (.50)
Institution 2 .21 .59 (.49) .55 (.51)
Parents WP .06 .41 (.49) .39 (.50)
Parents M 2.12 .06 (.24) .13 (.34)
Peers WP .22 .87 (.33) .90 (.30)
GangNew .23 .30 (.46) .26 (.44)
MissGang 7.88“ .05 (.21) .16 (.37)
Drug Sales Freq. .47 4.02 (2.16) 4.29 (2.07)
H. Drug UseNew 1.66 .81 (1.37) .48 (1.21)
MissHDU 1.18 .13 (.33) .19 (.40)
First FirearmNew 4.02* 11.13 (5.97) 8.87 (6.95)
MissFF 4.59* .19 (.40) .35 (.49)
First ArrestNew .47 14.43 (4.49) 15.00 (3.45)
MissFA .39 .06 (.24) .03 (.18)

 

‘p < 0.05, ”p < 0.01

98

Table 33: Bivariate AN OVA of Missing Responses on the Variable of Age at First
Firearm Ownership on Predictors (n=453).

 

Non-missing Values Missing Values

 

(n=360) (n=93)

Dependent Variable F Mean (s.d.) Mean (s.d.)
R’s Age 2.77 20.32 (1.80) 20.68 (2.09)
BlackNew 432* .44 (.50) .56 (.50)
MissBlack .52 .05 (.22) .03 (.18)
Detroit .18 .25 (.43) .27 (.45)
Institution 2 .01 .59 (.49) .59 (.49)
Parents O/P 28.65*** .47 (.50) .17 (.38)
Parents All 380* .08 (.27) .02 (.15)
Peers O/P 59.12*** .93 (.25) .66 (.48)
V. ExposeNew 5981*“ 4.77 (2.25) 2.73 (2.35)
MissVE 4.59* .06 (.23) .12 (.32)
GangNew 2074*" .34 (.48) .11 (.31)
MissGang .31 .05 (.22) .06 (.25)
Drug Sales Freq. 52.64 4.39 (2.02) 2.73 (2.35)
H. Drug UseNew 13.35 .91 (1.42) .33 (.98)
MissHDU 5.71 .11 (.31) .20 (.41)
First ArrestNew .16 14.43 (4.26) 14.63 (5.01)
MissFA .69 .05 (.22) .08 (.27)

 

*p < 0.05, "p < 0.01, "*p < 0.001

99

Table 34: Bivariate ANOVA of Missing Responses on the Variable of Age at First Arrest

 

 

on Predictors (n=453L
Non-missing Values Missing Values
(n=427) (n=26)

Dependent Variable F Mean (s.d.) Mean (s.d.)
R’s Age 20.41 (1.86) 20.15 (2.07)
BlackNew .69 .47 (.50) .38 (.50)
MissBlack 1.34 .05 (.22) 0
Detroit 1.31 .25 (.43) .35 (.49)
Institution 2 .27 .59 (.49) .31 (.47)
Parents O/P 1.16 .41 (.49) .31 (.47)
Parents All .05 .07 (.25) .08 (.27)
Peers O/P .02 .88 (.33) .88 (.33)
V. ExposeNew .10 4.34 (2.41) 4.50 (2.47)
MissVE .39 .07 (.26) .04 (.20)
GangNew .09 .30 (.46) .27 (.45)
MissGang .12 .05 (.23) .04 (.20)
Drug Sales Freq. 8.55“ 4.11 (2.14) 2.85 (2.11)
H. Drug UseNew .05 .79 (1.37) .73 (1.22)
MissI-IDU .69 .13 (.34) .08 (.27)
First FirearmNew 2.08 11.07 (6.05) 9.31 (6.14)
MissFF .69 .20 (.40) .27 (.45)

 

‘p < 0.05, "p < 0.01

100

REFERENCES

Agresti, A. (1989). Tutorial on modeling ordered categorical response data.
Psychological Bulletin, 105(2), 290-301.

Agresti, A. (2002). Categorical data analysis (2"d ed.). Hoboken, NJ: John Wiley and
Sons.

Akers, R. L. (1985). Deviant behavior: A social learning approach (3rd ed.). Belmont,
CA: Wadsworth Publishing Company.

Akers, R. L., Krohn, M. D., Lanza-Kaduce, L., & Radosevich, M. (1979). Social learning
and deviant behavior: A speciﬁc test of a general theory. American Sociological
Review, 44(4), 636-655.

Akers, R. L., & Sellers, C. S. (2004). Criminological theories: Introduction, evaluation,
and application (4th ed.). Los Angeles, CA: Roxbury Publishing Company.

Allen, M. P. (1997). Understanding regression analysis. New York, NY: Plenum Press.

Anderson, E. (1998). The social ecology of youth violence. In M. Tonry & M. H. Moore
(Eds), Youth violence: Crime and justice, an annual review of research (V 01. 24,
pp. 65-104). Chicago, IL: University of Chicago Press.

Bachman, R. & Paternoster, R. (2004). Statistics for criminology and criminal justice
(2nd ed.). New York, NY: McGraw-Hill Companies.

Berk, R. A. (2004). Regression analysis: A constructive critique. Thousand Oaks, CA:
Sage Publications.

Bjerregaard, B. & Lizotte, A. J. (1995). Gun ownership and gang membership. Journal of
Criminal Law and Criminology, 86(1), 37-58

Blumstein, A. (1995). Youth violence, guns, and the illicit-drug industry. Journal of
Criminal Law and Criminology, 86(1), 10-36.

Blumstein, A., Cohen, J., & Farrington, D. P. (1988). Criminal career research: Its value
for criminology. Criminology, 26(1), 1-35.

Blumstein, A., Cohen, J ., Roth, J. A., & Visher, C. A. (1986). Criminal careers and
“career criminals ” (V 01. 1). Washington, DC: National Academy Press.

Braga, A. A., Kennedy, D. M., Waring, E. J ., & Piehl, A. M. (2001). Problem-orientated

policing, deterrence, and youth violence: An evaluation of Boston’s Operation
Ceaseﬁre. Journal of Research in Crime and Delinquency, 38(3), 195-225.

101

Braga, A. A., Weisburd., D. L., Waring, E. J., Mazerolle, L. G., Spelrnan, W., &
Gaj ewski, F. (1999). Problem-orientated policing in violent crime places: A
randomized controlled experiment. Criminology, 37(3), 541-580.

Brame, R. & Paternoster, R. (2003). Missing data problems in criminological research:
Two case studies. Journal of Quantitative Criminology, 19(1), 55-78.

Bureau of Alcohol, Tobacco, and Firearms. (2000, February). Commerce in ﬁrearms in
the United States. Washington, DC: Department of the Treasury, Bureau of
Alcohol, Tobacco, and Firearms.

Bureau of Justice Statistics. (2004, August). Survey of state procedures related to ﬁrearm
sales, midyear 2003 (NCJ 203701). Washington, DC: US. Department of
Justice, Ofﬁce of Justice Programs, Bureau of Justice Statistics.

Bureau of Justice Statistics. (n.d.) Key facts at a glance: Crimes committed with ﬁrearms,
1973-2003 table. Retrieved January 13, 2005, from
http://www.ojp.usdoj.gov/bjs/glance/tables/guncrimetab.htm.

Caetano, D. F. (1979). The domestic arms race. Journal of Communication, 29(2), 39-46.

Cao, L., Adams, A., & Jensen, V. J. (1997). A test of the black subculture of violence
thesis: A research note. Criminology, 35(2), 367-379.

Cao, L., Cullen, F . T., & Link, B. G. (1997). The social determinants of gun ownership:
Self-protection in an urban environment. Criminology, 35(4), 629-650.

Catalano, S. M. (2004, September). Criminal victimization, 2003. Bureau of
Justice Statistics National Crime Victimization Survey (N CJ 205455).
Washington, DC: US. Department of Justice, Ofﬁce of Justice Programs,
Bureau of Justice Statistics.

Chaiken, J. M. & Chaiken, M. R. (1990). Drugs and predatory crime. In M. Tonry & J.
Q. Wilson (Eds), Drugs and crime: Crime and justice, an annual review of
research (V 01. 13, pp. 203-239). Chicago, IL: University of Chicago Press.

Cloward, R. A. & Ohlin, L. E. (1963). Delinquency and opportunity: A theory of
delinquent gangs. New York, NY: The Free Press of Glencoe.

Cook, P. J. & Ludwig, J. (2000). Gun violence: The real costs. New York, NY: Oxford
University Press.

Dean, C. W., Brame, R., & Piquero, A. R. (1996). Criminal propensities, discrete groups
of offenders, and persistence in crime. Criminology, 34(4), 547-574.

102

Decker, S. H., Pennell, S., & Caldwell, A. (1997, January). Illegal ﬁrearms: Access and
use by arrestees. National Institute of Justice, Research in Brief (N CJ 163496).
Washington, DC: US. Department of Justice, Ofﬁce of Justice Programs,
National Institute of Justice.

Diener, E. & Kerber, K. W. (1979). Personality characteristics of American gun-owners.
Journal of Social Psychology, 107(2), 227-238.

Dixon, J. & Lizotte, A. J. (1987). Gun ownership and the “Southern subculture of
violence.” American Journal of Sociology, 93(2), 383-405.

Elder, G. H. (1994). Time, human agency, and social change: Perspectives on the life
course. Social Psychology Quarterly, 5 7(1), 4-15.

Ellison, C. G. (1991). An eye for an eye? A note on the southern subculture of violence
thesis. Social Forces, 69(4), 1223-1239.

Fagan, J. & Wilkinson, D. L. (1998). Guns, youth violence, and social identity in inner
cities. In M. Tonry & M. H. Moore (Eds), Youth violence: Crime and justice, an
annual review of research (V 01. 24, pp. 105-188). Chicago, IL: University of
Chicago Press.

Federal Bureau of Investigation. (2004). Crime in the United States 2003. Washington,
DC: US. Department of Justice, Federal Bureau of Investigation.

Federal Bureau of Investigation (n.d.). Crime reporting in the age of technology: The
national incident-based reporting system. Criminal Justice Information Services
Newsletter (V 01. 4, No. 1). Washington, DC: US. Department of Justice, Federal
Bureau of Investigation, Criminal Justice Information Services Division.

Felson, R. B., Liska, A. E., South, S. J., & NcNulty, T. J. (1994). The subculture of
violence and delinquency: Individual vs. school context effects. Social Forces,
73(1), 155-173.

Fisher, C. S. (1995). The subcultural theory of urbanism: A twentieth-year assessment.
The American Journal of Sociology, 101(3), 543-577.

Fox, J. (1991). Regression diagnostics: An introduction. Sage University Paper Series on
Quantitative Applications in the Social Sciences (No. 07-079). Newberry Park,
CA: Sage Publications.

Gottfredson, S. D. & Gottfredson, D. M. (1994). Behavioral prediction and the problem
of incapacitation. Criminology, 32(3), 441-474.

Gove, W. R., Hughes, M., & Geerken, M. (1985). Are Uniform Crime Reports a valid
indicator of the index crimes? An afﬁrrnative answer with minor qualiﬁcations.

103

Criminology, 23(3), 451-501.

Hagan, J., Heﬂer, G., Classen, G., Boehnke, K., & Merkens, H. (1998). Subterranean
sources of subcultural delinquency beyond the American dream. Criminology,
36(2), 309-341.

Hawley, F. F. & Messner, S. F. (1989). The southern violence construct: A review of
arguments, evidence, and the normative context. Justice Quarterly, 6(4), 481-511.

Heirner, K. (1997). Socioeconomic status, subcultural deﬁnitions, and violent
delinquency. Social Forces, 75(3), 799-833.

Hirschi, T. & Gottfredson, M. (1983). Age and the explanation of crime. American
Journal of Sociology, 89(3), 552-584.

Hofstadter, R. (1970). America as a gun culture. In L. Nisbet (Ed) The gun control
debate (2“d ed. ,.pp 29- 37). Amherst, NY: Prometheus Books.

Howell, J. C. & Decker, S. H. (1999, January). The youth gangs, drugs, and violence
connection. Oﬂice of Juvenile Justice and Delinquency Prevention Juvenile
Justice Bulletin (NCJ 171152). Washington, DC: US. Department of Justice,
Ofﬁce of Justice Programs, Ofﬁce of Juvenile Justice and Delinquency
Prevention.

Huebner, B. M. ,,Bynum T. S., & Hinduja, S. (2001). Firearm use among Michigan ’s
youthful oﬂ'ender population. East Lansing, MI: Michigan Justice Statistics
Center, Michigan State University.

Heumann, M., Loftin, C., & McDowall, D. (1982). Federal ﬁrearms policy and
mandatory sentencting. Journal of Criminal Law and Criminology, 73(3), 1051-
1060.

Jacobs, J. B. (2002). Can gun control work? New York, NY: Oxford University Press.

Jensen, G. F. (1972). Parents, peers, and delinquent action: A test of differential
association perspective. American Journal of Sociology, 78(3), 562-575.

Kennedy, D. M., Piehl, A. M., & Braga, A. A. (1996). Youth violence in Boston: Gun
markets, serious youth offenders, and a use-reduction strategy. Law and
Contemporary Problems, 59(1), 147-196.

Kennett, L. B. & Anderson, J. L. (1975). The gun in America: The origins of a national
dilemma. Westport, CT: Greenwood Press.

Kleck, G. (1997). Targeting guns: Firearms and their control. Hawthorne, NY: Aldine
De Gruyter.

104

Kleck, G. (1991). Point blank: Guns and violence in America. Hawthorne, NY: Aldine
De Gruyter.

Kleck, G. & Patterson, E. B. (1993). The impact of gun control and gun ownership levels
on violence rates. Journal of Quantitative Criminology, 9(3), 249-288.

Kohn, A. A. (2004). Shooters: Myths and realities of America ’s gun cultures. New York,
NY: Oxford University Press.

Laub, J. H. & Sampson, R. J. (1993). Turning points in the life course: Why change
matters to the study of crime. Criminology, 31(3), 301-325.

Lizotte, A. J. & Bordua, D. J. (1980). Firearms ownership for sport and protection: Two
divergent models. American Sociological Review, 45(2), 229-244.

Lizotte, A. J ., Bordua, D. J., & White, C. S. (1981). Firearm ownership for sport and
protection: Two not so divergent models. American Sociological Review, 46(4),
499-503.

Lizotte, A. J., Krohn, M. D., Howell, J. C., Tobin, K., & Howard, G. J. (2000). Factors
inﬂuencing gun carrying among young urban males over the adolescent-young
adult life course. Criminology, 38(3), 811-834.

Lizotte, A. J ., Tesoriero, J. M., Thomberry, T. P., & Krohn, M. D. (1994). Patterns of
adolescent ﬁrearm ownership and use. Justice Quarterly, 11(1), 51-74.

Loeber, R. & Le Blanc (1990). Toward a developmental criminology. In M. Tonry & N.
Morris (Eds), Crime and justice: An annual review of research (V 01. 12, pp. 375-
473). Chicago, IL: University of Chicago Press.

Loeber, R. & Stouthamer-Loeber, M. (1986). Family factors as correlates and predictors
of juvenile conduct problems and delinquency. In M. Tonry & N. Morris (Eds),
Crime and justice: An annual review of research (V 01. 7, pp. 29-149). Chicago,
IL: University of Chicago Press.

Loeber, R, Stouthamer-Loeber, M., Van Kammen, W., & Farrington, D. P. (1991).
Initiation, escalation and desistance in juvenile offending and their correlates.
Journal of Criminal Law and Criminology, 82(1), 36-82.

Loftin, C. & Hill, R. H. (1974). Regional subculture and homicide: An examination of the
Gastil-Hackney thesis. American Sociological Review, 39(5), 714-724.

Loftin, C. & McDowall, D. (1981). “One with a gun gets you two”: Mandatory

sentencing and ﬁrearm violence in Detroit. The Annals of the American Academy
of Political and Social Science, 455, 150-168.

105

Long, J. S. (1997). Regression models for categorical and limited dependent variables.
Thousand Oaks, CA: Sage Publications.

Luckenbill, D. F. & Doyle, D. P. (1989). Structural position and violence: Developing a
cultural explanation. Criminology, 27(3), 419-436.

Marks, A. & Stokes, C. S. (1976). Socialization, ﬁrearms, and suicide. Social Problems,
23(5), 622-629.

Marquis, K. H. (1981). Quality of prisoner self-reports: Arrest and conviction response
errors. Santa Monica, CA: Rand.

McDowall, D. (1991). Firearm availability and homicide rates in Detroit, 1951-1986.
Social Forces, 69(4), 1085-1101.

McGarrell, E. F., Cherrnak, S., Weiss, A., & Wilson, J. (2001). Reducing ﬁrearms
violence through directed police patrol. Criminology and Public Policy, 1(1), 119-
148.

McKelvey, R. D. & Zavoina, W. (1975). A statistical model for the analysis of ordinal
level dependent variables. Journal of Mathematical Sociology, 4(1), 103-120.

Messner, S. F. (1983). Regional and racial effects on the urban homicide rate: The
subculture of violence revisited. The American Journal of Sociology, 88(5), 997-
1007.

Mofﬁtt, T. E. (1993). Adolescence-limited and life-course-persistent antisocial behavior:
A Developmental Taxonomy. Psychological Review, 100(4), 674-701.

Nagin, D. S. & Farrington, D. P. (1992). The onset and persistence of offending.
Criminology, 30(4), 501-523.

Nurco, D. N., Hanlon, T. E., Kinlock, T. W., & Duszynski, K. R. (1988). Differential
criminal patterns of narcotic addicts over an addition career. Criminology, 26(3),
407-423.

O’Connor, J. F. & Lizotte, A. (1978). The ‘Southern subculture of violence’ thesis and
patterns of gun ownership. Social Problems, 25(4), 420-429.

Piquero, A., Paternoster, R., Mazerolle, P., Brame, R., & Dean, C. W. (1999). Onset age

and offense specialization. Journal of Research in Crime and Delinquency, 36(3),
275-299.

Perkins, C. (2003, September). Weapon use and violent crime. Bureau of Justice
Statistics Special Report (NCJ 194820). Washington, DC: US. Department of

106

Justice, Ofﬁce of Justice Programs, Bureau of Justice Statistics.

Project Safe Neighborhoods. (2004). Project safe neighborhoods. Retreieved November
28 2004, from http://www.psn.gov/.

Sampson, R. J. & Laub, J. H. (1992). Crime and deviance in the life course. Annual
Review of Sociology, 18, 63-84.

Sheley, J. F. & Wright, J. D. (1995). In the line of ﬁre: Youth, guns, and violence in
America. New York, NY: Aldine De Gruyter.

Sherman, L. W. & Rogan, D. P. (1995). Effect of gun seizures on gun violence: “Hot
spots” patrol in Kansas City. Justice Quarterly, 12(4), 673-693.

Singleton Jr., R. A. & Straits, B. C. (1999). Approaches to social research (3rd ed.). New
York, NY: Oxford University Press.

Skogan, W. G. (1976). Citizen reporting of crime: Some national panel data.
Criminology, 13(4), 535-549.

Spergel, I. A. (1990). Youth gangs: Continuity and change. In M. Tonry & N. Morris
(Eds), Crime and justice: An annual review of research (V 01. 12, pp. 171-275).
Chicago, IL: University of Chicago Press.

Sutherland, E. H. (1939). Principles of criminology (3rd ed.). Chicago, IL: J. B. Lippincott
Company.

Thomberry, T. P. (1987). Toward an interactional theory of delinquency. Criminology,
25(4), 863-891.

Treatment Alternatives for Safe Communities, Inc. (1997). Gun ownership and use by
Chicago adult male arrestees. Chicago, IL: Treatment Alternatives for Safe
Communities, Inc.

US. Census Bureau. (2000). American fact/inder data set query. Retrieved January 13,
2005, from http://factﬁnder.census.gov.

US. Census Bureau. (1990). American factﬁnder data set query. Retrieved January 13,
2005, from http://factﬁnder.census.gov.

Vold, G. B., Bernard, T. J ., & Snipes, J. B. (2002). Theoretical criminology (5th ed.). New
York, NY: Oxford University Press.

Warr, M. (1993). Age, peers, and delinquency. Criminology, 31(1), 17-40.

Weisburd, D. & Green, L. (1995). Policing drug hot spots: The Jersey City drug market

107

analysis experiment. Justice Quarterly, 12(4), 711-735.

Winship, C. & Mare, R. D. (1984). Regression models with ordinal variables. American
Sociological Review, 49(4), 512-525.

Wolfgang, M. E. & Ferracuti, F. (1967). The subculture of violence: Towards an
integrated theory in criminology. London, UK: Tavistock Publications.

Wright, J. D. & Marston, L. L. (1975). The ownership of the means of destruction:
Weapons in the United States. Social Problems, 23(1), 93-107.

Wright, J. D. & Rossi, P. H. (1994). Armed and considered dangerous: A survey of felons
and their ﬁrearms (Expanded ed.). Hawthorne, NY: Aldine De Gruyter.

Wright, J. D., Rossi, P. H., & Daly, K. (1983). Under the gun: Weapons, crime, and
violence in America. Hawthorne, NY: Aldine De Gruyter.

Zimring, F. E. (1975). Firearms and federal law: The Gun Control Act of 1968. The
Journal of Legal Studies, 4(1), 133-198.

108

 

IIIIIIIIIII III II IIIIIIII III“

23 02736 25