|USFDC Home | USF Electronic Theses and Dissertations||| RSS|
This item is only available as the following downloads:
xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam Ka
controlfield tag 001 001709498
007 cr mnu|||uuuuu
008 060509s2005 flu sbm s000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0001303
Self-control, gang membership, and victimization
h [electronic resource] :
b an integrated approach to the risk factors of violent victimization /
by Kristina Childs.
[Tampa, Fla.] :
University of South Florida,
Thesis (M.A.)--University of South Florida, 2005.
Includes bibliographical references.
Text (Electronic thesis) in PDF format.
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
Title from PDF of title page.
Document formatted into pages; contains 94 pages.
ABSTRACT: This study integrates one of the most empirically tested individual characteristics (self-control) with situational factors (risky lifestyle) in an attempt to explain the risks involved in violent victimization. Data came from a diverse sample of 3,907 middle school students who participated in the G.R.E.A.T. program during the 1993-1994 school year. Gang membership is used as a proxy variable to measure risky lifestyle. It is hypothesized that 1) gang membership will mediate the effect of self-control on violent victimization and 2) self-control and gang membership will interact to amplify the risk of violent victimization. Logistic regression was used to analyze the prevalence of violent victimization and negative binomial regression was used to analyze frequency of violent victimization. When all other extraneous factors were controlled, insufficient evidence was found to support the hypotheses of this study.A discussion of the findings, as well as theoretical implications and suggestions for future research are discussed.
Adviser: John Cochran, Ph.D.
General theory of crime.
t USF Electronic Theses and Dissertations.
Self-Control, Gang Membership, and Victimization: An Integrated Approach to the Risk Factors of Violent Victimization by Kristina Childs A thesis submitted in partial fulfillment of the requirements for the degree of Master of Arts Department of Criminology College of Arts and Sciences University of South Florida Major Professor: John K. Cochran, Ph.D. Christine Sellers, Ph.D. Kim Lersch, Ph.D. Date of Approval: September 13, 2005 Keywords: general theory of crime, lifestyle/exposure, G.R.E.A.T., victimization, gangs Copyright 2005, Kristina Childs
Dedication This thesis would not have been possible wi thout the love and support of my friends and family. Thank you all for believing in me, es pecially when I didnt believe in myself. Your love and encouragement is wh at made this project possible.
Acknowledgements The successful completion of this thesis was dependent on the knowledge and cooperation of several individuals. The foundation of this paper would not have been established without the bril liance of Dr. Gibson. Thank you Dr. Gibson for your patience and guidance throughout this project. In addition, I wish to thank my committee members, Dr. Christine Sellers and Dr. Kim Lersch, for their continued support. I am very fortunate to have had the opportunity to work with you. Particularly, I would like to express my sincere appreciation to Dr. J ohn Cochran, whose wisdom and patience made this thesis possible. Finally, I would like to thank all the faculty and graduate students within the Department of Criminology.
i Table of Contents List of Tables List of Figures Abstract v Chapter 1 Introduction v Chapter 2 Theories and Correlates of Vi ctimization 2 Routine Activities/Opportunity Theory 2 Risky Lifestyle/Exposure to Crime 7 Integrating Routine Activities a nd Lifestyle/Exposure 10 Individual Level Correlates of Victimization 13 Chapter 3 Self-Control 16 General Theory of Crime 16 Literature Review 21 Self-Control and Victimization 25 Chapter 4 Gang Membership and Risky Lifestyle 31 Hypotheses 33 Chapter 5 Methods 35 Victimization 37 Self-Control 39 Gang Membership 41 Control Variables 41 Analysis 44 Chapter 6 Results 46 Prevalence of Violent Victimization 46 Frequency of Violent Victimization 55 Chapter 7 Conclusions and Discussion 65
ii List of Tables Table 1 Demographics of the Sample 36 Table 2 Logistic Regression Predicting Prevalence of Hit by Someone Trying to Hurt You 47 Table 3 Logistic Regression Predicting Pr evalence of Robbery 48 Table 4 Logistic Regression Predicting Prevalence of Attacked with a Weapon 48 Table 5 Logistic Regression Pred icting Prevalence of Violent Victimization 48 Table 6 Logistic Regression Predicting Prevalence of Hit by Someone Trying to Hurt You Including the Control Variables 49 Table 7 Logistic Regression Predicting Prevalence of Robbery Including the Control Variables 51 Table 8 Logistic Regression Predicti ng Prevalence of Attacked with a Weapon Including the Control Variables 52 Table 9 Logistic Regression Pred icting Prevalence of Violent Victimization Including the Control Variables 53 Table 10 Logistic Regression Predicting Prevalence of Hit by Someone Trying to Hurt You by Gang Membership 54 Table 11 Logistic Regression Predic ting Prevalence of Attacked with a Weapon by Gang Membership 55 Table 12 Negative Binomial Regr ession Predicting Frequency of Hit by Someone Trying to Hurt You 57 Table 13 Negative Binomial Regression Predic ting Frequency of Robbery 57
iii Table 14 Negative Binomial Regression Predicting Frequency of Attacked with a Weapon 57 Table 15 Negative Binomial Regression Predicting Frequency of Violent Victimization 58 Table 16 Negative Binomial Regre ssion Predicting Frequency of Hit by Someone Trying to Hurt You Includi ng the Control Variables 59 Table 17 Negative Binomial Regr ession Predicting Frequency of Robbery Including the Control Variables 60 Table 18 Negative Binomial Regr ession Predicting Frequency of Attacked with a Weapon Including th e Control Variables 61 Table 19 Negative Binomial Regression Predicting Frequency of Violent Victimization Including the Control Variables 62 Table 20 Negative Binomial Regr ession Predicting Frequency of Hit by Someone Trying to Hurt You by Gang Membership 64
iv List of Figures Figure 1 Low Self-Control Predictive Model 33
v Self-Control, Gang Membership, and Victimization: An Integrated Approach to the Risk Factors of Violent Victimization Kristina Childs ABSTRACT This study integrates one of the most em pirically tested indi vidual characteristics (self-control) with situationa l factors (risky lifestyle) in an attempt to explain the risks involved in violent victimi zation. Data came from a diverse sample of 3,907 middle school students who participated in the G.R.E.A.T. program during the 1993-1994 school year. Gang membership is used as a proxy va riable to measure risky lifestyle. It is hypothesized that 1) gang membership will me diate the effect of self-control on violent victimization and 2) self-control and gang member ship will interact to amplify the risk of violent victimization. Logistic regression was used to analy ze the prevalence of violent victimization and negative binomial regression was used to analyze frequency of violent victimization. When all other extraneous factors were co ntrolled, insufficient evidence was found to support the hypotheses of this study. A discussion of the findi ngs, as well as theoretical implications and suggestions for future research are discussed.
1 Chapter 1 Introduction Several attempts have been made to identi fy the risk factors involved in criminal victimization. Based on these attempts, the l eading theories that have emerged (routine activities, criminal opportunities, and lifestyle/exposure) explain victimization using situational factors such as the presence of motivated offenders, suitable targets, and a lack of capable guardianship (Hindelang et al. 1978, Cohen & Felson 1979, Miethe & Meier 1990). These theories have been empirical ly supported (Cohen & Felson 1979, Meithe et al. 1987) and widely accepted as plausible expl anations of the risk factors involved in victimization. However, they do not involve a complete assessment of all the possible factors that could influence th e risk of victimization. More precisely, these theories are only concerned with the situational factors involved, and do not take into account any individual characteristics that may put an individual at risk for victimization. For example, research has consistently shown that victims of criminal behavior tend to be young, male, and members of a racial or ethnic minority (Lauritsen, Sampson, & Laub 1991, Lauritsen & Quinet 1995). Othe r individual characte ristics such as marriage, peer groups, and number of siblings have also been shown to be associated with criminal victimization (Sampson & Lauritsen 1990, Schreck et al. 2002). In addition to the demographic characte ristics, the bulk of empirical literature concludes that offending and victimization are highly corr elated (Sampson & Lauritsen
2 1990, Esbensen & Huizinga 1991, Lauritsen et al. 1991, Lauritsen et al. 1992, Lauritsen & Quinet 1995). Offenders have also been found to be victims of crime at a substantially higher rate than nonoffenders (Sampson & La uritsen 1990, Lauritsen et al. 1991). Based on these findings, the similarities between offe nders and victims raise the possibility that a common underlying factor may be influenci ng the likelihood of b ecoming both a victim and an offender. To explore this possibil ity, a consideration of one of the most empirically supported (Pratt & Cullen 2000) i ndividual traits associated with deviant behavior will be used in an attempt to explain the risk of violent victimization. Specifically, this paper will address the effect that low self-control has on the risk of being victimized. Chapter 2 begins with a brief explanation of various situationa l factors that have been found to be related to victimization. Specifically, the routine activities, risky lifestyle/exposure, and structur al-choice perspectives are re viewed. Next, individual-level correlates of victimization are discussed. Based on these correlations and the offendervictim link, it is argued that self-control has the potential to influence victimization. Chapter 3 presents a description and liter ature review of Gottfredson and Hirschi's (1990) general theory of crime. Based on the work of Schreck (1999), a connection between the elements of self -control and victimization follo ws. Finally, previous studies that test this self-control-victimization theory are summarized. Chapter 4 provides a justif ication of using gang membership as a proxy variable to measure risky lifestyle. Evidence of an association between gang membership and various risky behaviors, ga ng membership and self-contro l, and gang membership and victimization is discussed. The hypotheses for this study conclude this chapter.
3 Chapter 5 consists of an overview of the methods used in this study. Sample characteristics, description of the variab les, and analytic techniques are provided. Chapter 6 summarizes the results of the study. Analyses of the influence of selfcontrol and gang membership on the prevalence and frequency of violent victimization are discussed. Chapter 7 concludes this paper with a discussion of the findings. Theoretical implications, as well as limitations and sugges tions for future research are provided.
4 Chapter 2 Theories and Correlates of Victimization Routine Activities/Opportunity Theory For the past few decades, the risk of vict imization has been attributed to various situational factors that are either present or absent in an individuals environment. A handful of theories have emerged regarding these factors and are now the most widely cited causes of victimization (Miethe et al. 1987, Miethe & Meier 1990, Schreck et al. 2002). These theories are based on the notion that in order for crime (and in turn, victimization) to occur, a pe rpetrator, victim, and/or prope rty must converge in time and space. Further, the occurrence can be f acilitated if there ar e other persons or circumstances in the situation that encourage it, or it can be prevented if the potential victim or another person is present w ho can deter it (Akers & Sellers 2004:33). Cohen and Felson (1979) developed a routin e activities theory to explain these elements. They provided three categories of variables that increas e the likelihood of a direct-contact predatory (per sonal or property) violation. The first requirement is the presence of a motivated offender. This means that the perpetrator possesses both the inclination and ability to carry out the act. These offenders commit crime if and when the situation allows. Second, a suitable target fo r the act must also be present. A suitable target is one that is availa ble and attractive to the offender. The more affluence or monetary possessions one has, the more attractive a target he or she is to an offender. The
5 last variable is the absence of capable guardians to prev ent violations. Guardianship refers to the extent to which an individual is able to protect on eself, as well as his or her belongings. Guardianship can be in the form of police officers, as well as ordinary citizens. Cohen and Felson (1979) emphasized that the lack of any one of these variables is sufficient to deter crime. It is the converge nce in time and space of these three situations that leads to direct predat ory crimes. All three elements vary from person to person, depending on the circumstances and location of their daily environment. In other words, the spatial and temporal stru cture of routine legal activi ties should play an important role in determining the location, type, and qua ntity of illegal acts occurring in a given community or society (Cohen & Felson 1979:590) Thus, criminal behavior, and in turn, victimization are most likely to occur when there is at least one person motivated to commit a criminal act, a target that is attractiv e to the offender, and the absence of formal or informal deterrents at the same place and time. To test their theory, Cohen and Felson (1979) analyzed various forms of government data from 1947-1975. They looked at family activities, human behavior patterns such as age, marriage and employment rates, trends in the market for consumer products, business establishments and investme nts, and compared them with trends in stolen property, robbery, burglar y, and murders. At the microlevel, victimization rates were inversely related to age, lower for married individuals, and higher for those who were unemployed or lived in single-adult hous eholds. At the macro-level, trends in human activities were found to occur during the same period that sales of consumer goods and business establishments were signi ficantly increasing. In turn, these trends
6 were associated with higher rates of all f our types of crime. These comparisons yielded significant support for th eir theory. Based on these findings, Cohen et al. (1981a, 1981b) tested an extension of routine activities, which they labeled opportunity theory. They included five factors that were used to explain the risk of victimization. These va riables are exposure, guardianship, proximity to potential offenders, target attractiveness, and definitional properties of assault, burglary, personal larceny, and robbery. Nine propositions regarding the mediating role of these variab les were tested using data from the 1974-1977 National Crime Survey. Findings revealed s upport for a routine act ivities/opportunity explanation of victimization and lead to the conclusion th at the structure of routine activities provides the opportunity for these crimes to occur. Since then, several studies have b een conducted to test macro (Sampson & Wooldredge 1987, Stahura & Sloan 1988, L ee 2000) and micro-levels of routine activities/opportunity theory (Nelson & Huff 1998, Mustaine & Tewksbury 1998). For example, using a sample of the largest SMSAs in the United States, Messner and Blau (1987) analyzed crime rates in the 1980s. Patterns of leisure (television watching, commercial cinemas, sports activities, and en tertainment producers) and crime rates were consistent with predictions made in r outine activities theor y. Nonhousehold leisure activities were found to be significantly related to higher crime rates, whereas concentration of household activities tended to reduce risk of crime. In regard to micro-level findings, Th ompson & Fisher (1996) used data from 14,258 households in the 1983 National Crime Su rvey: Victim Risk Supplement. They found that all three dimensions of routine activities were important predictors of burglary
7 and larceny events. In additi on, they found that these routin e variables provided a unique opportunity structure for these crimes to occu r. Thus, the routine activities perspective appears to be able to account for crime at both macro and micro levels of analysis. Several other studies have been conduc ted to test the ability of routine activities/opportunity th eory to explain the risk of vi ctimization. These studies have tested both the risk of victimization and rates or trends of victimiza tion. In general, these studies have also yielded findings consis tent with a routine activities/opportunities perspective. Being a victim of a property offense, personal offense, motor vehicle accident, homicide, assault, robbery, burglary, personal larceny, and crime in the workplace have all been shown to be related to measures of routine activities/opportunity (Hough 1987, Sampson & Wooldredge 1987, Lasley & Rosenbaum 1988, Sherman et al. 1989, Forde and Kennedy 1990a, Forde and Kennedy 1990b, Wooldredge et al. 1992, Keane & Arnold 1996, Nelson & Huff 1998, Lee 2000) It should be noted, however, that the majority of these findings have also been criticized (Lauritsen et al. 1992) for various limitations, most notably reliance on indirect measures (age, gender, marriage, employment) of the theoretical concepts and fa iling to include all three elements of the routine activities theory. Risky Lifestyle/Exposure to Crime Hindelang et al. (1978) also developed a th eory that attribut es variations in victimization to differences in personal ch aracteristics. Their lifestyle/exposure model proposed that the risk of vict imization is a function of an individuals lifestyle. Briefly, lifestyle is considered to be contingent upon demographi c characteristics, vocational and recreational activities. Specifi cally, they maintain that variations in lifestyle are
8 differentially related to proba bilities of being in certain places at certain times and coming into contact with persons who have certain characteristics. Because victimization is not randomly distributed and offenders are not representative of the general population, lifestyle differences are proposed to lead to differences in expos ure to situations that have a high rate of victimization. Eight propositions are used to explain the ways in which risky lifestyles increase the probability of being victimized. 1) The probability of suffering a personal victimization is directly rela ted to the amount of time a person spends in public places, most notably at night. This is based on the ev idence that criminal events are not randomly distributed across time and place. Rather, robb ery, rape, and assault have been shown to occur disproportionately at night and on the street (Hi ndelang et al. 1978). 2) The probability of being in public places, particular ly at night, varies by lifestyle. Demands of lifestyles influence where an individual spends his or her time. For example, individuals who work at night will have a higher risk of victimization because their lifestyle requires them to be out of the home at times when vic timization is most likely to occur. 3) Contact and social interactions occur among indivi duals with the same lifestyles. Because Hindelang et al. (1978) used demo graphics to predict lifestyles, they also maintained that these characteristics are indicative of soci al interactions. 4) The chances of being victimized depend on the extent to which an individual shares the same demographic characteristics as offenders. This pro position was based on the extant evidence (Hindelang et al. 1978) that victims and o ffenders disproportionately share the same demographic characteristics (male, young, urba n residents, low SE S). 5) The proportion of time an individual spends among nonfamily members is a function of lifestyle. 6) The
9 probability of personal victimization increa ses with the amount of time spent with nonfamily members. To support proposition five and six, Hindelang et al. (1978) pointed to the statistics that show the disproportionate amount of rape, robbery, and assault that is carried out by strangers. 7) Variations in lifes tyle are associated with variations in the amount of time spent around persons with offender characteristics. Th e ability to isolate oneself from persons with these character istics is dependent u pon the vocational and leisure activity one is involved i n. 8) Variations in lifestyle ar e associated with variations in the convenience, desirabilit y, and vincibility of the person as a target for victimization. Lifestyles that place individuals in dangerous areas, out of the home at night, or around nonfamily members will provide a more convenient, desirable target an d will be less able to resist the offender successfully. To test these eight propositions, Hindel ang et al. (1978) analyzed data from victimization surveys conducted in eight c ities in 1972. Their analyses yielded strong support for a risky lifestyle/e xposure explanation of victim ization. Age, marital status, employment status, and gender we re found to be closely associat ed with risk of personal victimization. These demographic characterist ics were assumed to carry different role expectations, structural constr aints, and individual adaptations that determine lifestyle and exposure to crime. As a result, Hindela ng et al. (1978) concluded the theoretical model of the likelihood that an individual wi ll suffer a personal victimization depends heavily on the concept of lifestyle (daily activities, both vocational and leisure). Building upon these findings, many cr iminologists have tested the risky lifestyle/exposure perspective. Using data from the 1982 Ca nadian Urban Victimization Survey, Forde and Kennedy (1990b) looked at several nighttime activities such as times
10 per month the respondent went to a bar, movie, restaurant, played sports, took a walk, or stayed at work late and compared them to the number of times they reported victimization. The amount of time spent on these activities was named time spent in risky lifestyles, which was inversely relate d to criminal victimization, or what the authors called dangerous results. These findings yielded significant support for the lifestyle perspective, and Forde and Kennedy (1990b) concl uded that victimization is contingent upon the exposure that results fr om certain lifestyles. Furthermore, Kennedy and Baron (1993) conducted a case study to ex amine a subculture of violence, and found that subcultural lifestyles not only influence exposure to violence, but also shape the behavioral choices in re sponse to victimization. Additionally, many other lifestyle characte ristics such as frequency of alcohol use, activities on college campuses, sports, fre quency at bars, time spent out of the house, age, gender, street activity, unemployment, fr iendship networks, marital status, nighttime activity, deviance, violent lifesty les, drug use, association an d time spent with delinquent peers, money spent on amusement arcades a nd sports events, ne ighborhood density, and activity on different street segments have all been shown to be associated with higher levels of victimization (Riley 1987, Samps on & Wooldredge 1987, Lasley & Rosenbaum 1988, Lasley 1989, Forde and Kennedy 1990a, Horney et al. 1995, Nelson & Huff 1998, Robinson 1999, Bjarnason et al. 1999, Rapp-Pa glicci & Wodarski 2000, Zhang et al. 2001, Schreck et al. 2002, Schreck & Fisher 2004). Integrating Routine Activit ies and Lifestyle/Exposure Although routine activities and lifestyle/ exposure theories were developed separately, their theoretical foundations are very similar. Both theories emphasize the
11 causal significance of time and place in daily life and point to differences in situational and/or demographic factors as the sources of differential risk for victimization. As Miethe et al. (1987: 184) explai n, both theories presume that di fferences in routine activities or lifestyles mediate the demographic correlates of victimization. As a result, the literature on victimizati on tends to integrate these two theories by using variables that account for both concepts. For the most part, routine activities are measured using indirect variables such as age, gender, marriage, and nighttime activities, which in turn, tend to be the same variables that are characteristic of a risky lifestyle. Consequently, criminologists have a tendency to integrate the theories and develop a set of routine activities/risky lif estyle measures (Garofalo et al. 1987, Miethe et al. 1987, Lasley 1989, Kennedy & Forde 1990b, Miethe and Meier 1990, Horney et al. 1995, Robinson 1999, Bjarnason et al. 1999, Cochran et al. 2000, Rapp-Pagl icci & Wodarski 2000, Schreck & Fisher 2004). Moreover, Miethe and Meier (1990) propos ed a structural-choice theory of victimization, which integrates aspects of rout ine activities and risky lifestyle. Under this model, exposure and proximity are consid ered the structural components and attractiveness and guardianship are the choi ce components. Exposure and proximity are the structural components because they repres ent the nature of social interactions and predispose individuals to higher risk of vic timization. Attractiveness and guardianship, on the other hand, are choice components because th ey are presumed to determine selection of targets. It is the inte raction in time and space of these structural and choice components that lead to variation in the risk of victimization. According to Miethe and Meier (1990:246), it does seem reasonable to predict that differences in risks of
12 victimization by target-selection factors (i.e. attractiveness, guardianship) should be most pronounced among persons with high proximity and exposure to crime. Using data from the 1982 British Crime Su rvey, Miethe & Meier (1990) analyzed victimization events of burglary, assault, a nd petty theft. Proximity to high-crime areas was measured by place of residence, perceive d safety of the neighborhood at night, and average rate of offending in the area, while exposure to crime was measured using questions regarding household activities. Measures of target attractiveness included the social-class rank of the respondent and whether or not he or she owned a VCR. Guardianship was measured with two items, whether or not the respondent lives alone and uses some sort of personal protection such as carries a weapon or has a burglar alarm. Analyses revealed that proximity and expos ure were in the expected direction and statistically significant for all three types of crime. This fi nding supports the structural component of the theory. Living alone revealed a statistically signifi cant effect in the expected direction. However, all of the ot her choice component va riables were either non-significant or associated in the opposite direction. The handful of studies that have tested th is structural-choice theory also provide mixed results. Wooldredge et al. (1992) examined workplace victimization on college campuses and found that exposure was associated with personal and property victimization while target attractiveness was not. Hoyt et al. (1999) looked at victimization among homeless and runaway youth. They found significant support for exposure to risk and only modest support for target attractiveness and guardianship. Finally, Cochran et al. (2000) examined a cr ime hot spot and f ound that victimization was associated with proximity and guardia nship, but not with exposure or target
13 attractiveness. These findings, along with Meithe and Meiers (1990) findings, provide mixed support for an integrated structural-choice perspective. Regardless, routine activities, risky lif estyle/exposure, and structural-choice theories all propose that the si tuational factors present in one s life account for variation in the risk of victimization. These persp ectives have been em pirically supported (Hindelang et al. 1978, Cohen & Felson 1979, Meith e et al. 1987) and widely accepted as plausible explanations of the risk factors involved in victimization. However, they have also been critici zed for not providing a complete assessment of all the possible factors that could influence the risk of victimization (Lauritsen 2001, Schreck et al. 2002). More precisely, thes e theories are only concerned with the situational factors involved, and do not take into account any individual characteristics that may put an individual at risk for criminal victimizati on. As a result, these theories assume that all other things are equal, impl ying that all individual traits are invariant and unimportant. Individual Level Correlates of Victimization The extant literature on victimization provides evidence of several individuallevel characteristics that have been found to be consistent predictors of victimization. For example, time and again, research has shown that victims of criminal behavior tend to be young and male (Lauritsen et al. 1991, Laur itsen & Quinet 1995). In addition, other individual characteristics such as race, marriage, college education, SES, delinquent peers, and number of siblings have also revealed an association with criminal victimization (Sampson & Lauritsen 1990, Esbe nsen & Huizinga 1991, Lauritsen et al.
14 1991, Lauritsen & Quinet 1995, Dahlberg 1998, Lauritsen 2001, Schreck 2002, Schreck et al. 2002, Schreck & Fisher 2004). Moreover, although lifestyle th eories attribute victimiz ation to the situational characteristics present in ones life, the most common criticism of lifestyle research is the lack of direct measures of lifestyle, and in turn, the use of dem ographic, or individuallevel, variables as indirect measures (Lauritsen et al. 1991). Thus, the findings of these studies provide evidence for an associat ion between individual level factors and victimization, and then are used to imply that th ese factors lead to variations in lifestyles, which in turn, lead to variati ons in the risk of victimiza tion. In addition, when separate measures of individual (offending, age, race, gender, peers, SES) and lifestyle variables (time spent away from home, pr oximity to crime, exercise, dr iving a car, time spent with family, sports involvement, and time spe nd at school) are included in a model of victimization, the individual characteristics rema in statistically significant (Lauritsen et al. 1992, Lauritsen 2001, Schreck & Fisher 2004). Thus, these i ndividual-level variables seem to have an effect inde pendent of lifestyle choices. Furthermore, one of the most empirica lly supported characteristics related to victimization is criminal offending. The extant literature provides evidence that selfreported offending is a significant predictor of victimization (Lauritsen et al. 1992, Lauritsen & Quinet 1995). Esbensen and Huiz inga (1991) analyzed data from 877 middle school youth and found that the mean rate of both property and violen t victimization was two to three times higher for those who re ported self-reported delinquency. Using two waves of the British Crime Survey, Sampson and Lauritsen (1990) also found that offense activity, serious or petty, directly incr eased the risk of violent victimization. In
15 addition, via data from the National Youth Surv ey, Lauritsen et al. (1991) concluded that self-reported offenders are approximately three times more likely than nonoffenders to be victims of assault, robbery, or vandalism. The similarities between offending and victim ization in their covariates as well as the robust association between offending and victimization raise the possibility that a common underlying factor may be influencing the likelihood of becoming both a victim and an offender. To explore this possibil ity, a consideration of one of the most empirically supported (Pratt & Cullen 2000) i ndividual traits associated with deviant behavior will be used in an attempt to explain the risk of cr iminal victimization. Specifically, this paper will address the effect that low self-control has on the risk of being victimized.
16 Chapter 3 Self-Control General Theory of Crime Although there are several different forms of deviance, Gottfredson and Hirschi (1990) argue that certain features of behavior are characteris tic of all deviant acts, across all populations and age groups. Based on these common characteristics, they developed a General Theory of Crime (GTC), which attemp ts to explain criminal behavior in terms of one versatile phenomenon. GTC was derived from the works of earlier classical theories, which argue that crimin al behavior is the result of the pursuit of self-interest. Particularly, Gottfredson and Hirschi (1990) base their general theory on the assumption that all human behaviors, criminal or noncri minal, are motivated by their perceived costs and rewards. Based on this premise, Gottfredson and Hirschi (1990) make a distinction between the nature of crime and the nature of criminality. Crimes are defined as acts of force or fraud undertaken in th e pursuit of self-i nterest. They are characterized by time and space, require little effort or skill, and provide immediate, short-term rewards. In sum, Gottfredson and Hirschi (1990:16) explai n crime as events whose temporal and spatial distributions are highly predictable, that require little preparation, leave few lasting consequences, and often do not produce the result that was intended. Criminal acts are seen as the result of the pursuit of immediate, certain, easy benefits.
17 Criminality, on the other hand, is defined as a characteristic of people. This definition is based on the extant literature th at provides evidence of stable individual differences in the propensity to commit crimin al acts. Gottfredson and Hirschi (1990) use self-control to account for these differences, and therefore, to define the nature of criminality. More specifically, they conceptua lize criminality as the extent to which an individual is compelled to crime. Howeve r, this depends on the perceived costs and rewards of the act, which in turn, are depe ndent upon the individuals level of selfcontrol. Furthermore, Gottfredson and Hirschi ( 1990) use six elements to describe an individuals level of self-cont rol. These elements are level of impulsivity, laziness, risk seeking, preference for physical activity, empathy, and tolerance for frustration. First, level of impulsivity is related to ones ability to defer gratification of desires. A person with low self-control is more likely to possess a here and now orientation and lack the ability to consid er long-term consequences. Criminal acts provide immediate gratificati on of desires and delayed cons equences, and therefore, are more likely to be committed by people who respond to their instantaneous desires. In contrast, people with high self-control tend to consider long-term consequences and defer gratification in order to avoid the costs. At the same time, individuals with low se lf-control are lazy. They prefer simple and easy gratification of desires. Criminal acts provide a ccumulation of these desires without much effort or planning, such as m oney or property without work and hardship. People with low self-control also tend to be adventuresome and active, whereas individuals with high self-cont rol are described as cautious and cognitive. Due to the
18 exciting, dangerous, and risky nature of crimin al acts, people with low self-control will seek out these risky behavior s at a much higher rate. Gottfredson and Hirschi (1990) maintain that crime does not require much thought or skill. People lacking self-control do not value cognitive abilities and prefer to engage in physical, rather than mental act ivities. Because crime does not require any training or skill, individuals low in self-contro l are attracted to these acts, based on their physical nature. Empathy is referred to as an individuals level of self-centeredness and sensitivity to others. Individuals who lack self-control tend to be self-centered and indifferent to the needs of others. Criminal acts result in pa in and loss. Thus, people who do not possess empathy will be indifferent to these feelings. Finally, tolerance for frustration is char acterized by an indi viduals ability to tolerate momentary frustration. Individuals with low self-control do not have this ability and as a result, become frustrated very easil y. Consequently, the poten tial for a violent or aggressive situation is increased. According to Gottfredson and Hirschi (1990), all six compone nts are essential dimensions of self-control and must be pr esent for crime to occur. Further, these dimensions do not operate separately, but rather come together to represent one underlying trait: self-control. The theorists assert that thes e traits are not alternative ways of having low self-control, nor are some causes of the others. Rather, the six traits are constitutive of low self-c ontrol and tend to come together to form a unidimensional latent trait (Arnek lev et al. 1993: 229).
19 Moreover, not only is self-control used to explain criminal behavior, it is also used to explain the tendency to engage in noncriminal acts such as accidents, smoking, drug use, drinking, and gambling. For example, gambling is risky and provides fast, easy money without much thought or planning. Th ese analogous behaviors share the same traits as crime, immediate gratificati on and long-term consequences. The pleasure obtained by these criminal and analogous acts are immediate, direct, and obvious, whereas the consequences are delayed, indirect and sometimes not as obvious. In sum, an individual who is impulsive, lazy, prefers risk and physical activity, lacks empathy and becomes easily frustrated pos sesses a low level of self-control, and in turn, will engage in criminal and analogous behaviors at a much higher rate than individuals who do not possess these traits. Furthermore, Gottfredson and Hirschi (1990) assert that an in dividuals level of self-control is developed by age twelve and remains stab le throughout the life course. They believe that individuals are not predispos ed to or taught low se lf-control, but rather that the cause of low self-control is ineff ective parenting. According to the GTC, the minimum conditions a parent must set up in orde r for a child to devel op an adequate level of self-control are 1) monitoring of the child s behavior; 2) recogni zing deviant behavior when it occurs; and 3) punish ing the deviant behavior. Although these conditions seem obvious and easy to carry out, parents do not always succeed in meeting all three require ments. Gottfredson and Hirschi (1990) rule out the possibility that parents are actively providing ineffective parenting, but instead, point to four situations in which parenting systems have a tendency to go wrong. First, some parents simply do not care for their ch ild. In this situation, none of the three
20 requirements are met. Second, parents may not have the time or energy to monitor the childs behavior. This may be the result of l ong hours at work, stress, or illness. Third, parents may care for and monitor their child s behavior, but do not realize that the behavior is deviant. On the other hand, parents may recognize deviant behavior, but engage in deviance themselves, and therefor e do not feel that the behavior deserves punishment. Last, when everything else is in place, parents may not have the ability to punish their child effectively. Not only does a lack of punishment influence a childs development of self-control but punishment that is too lenient or too harsh can also have a damaging effect. There are numerous possibl e circumstances that could lead to the presence of one or more of these situations Sibling size, parental criminality, singleparent families, and parental employment are a few examples. Gottfredson and Hirschi (1990) include one final situation that is necessary for criminal and analogous behavior to occur: opportunity. This theory does not imply that crime is an automatic consequence of low se lf-control. Instead, low self-control will lead to criminal and analogous behaviors when the opportunities are available and the circumstances allow for it. Opportunities and circumstances vary from person to person, depending on the individuals immediate e nvironment. As stated by Longshore and Turner (1998:82) in their theory, the link be tween crime and self-control is conditional on criminal opportunity, which is a function of structural or situational circumstances encountered by the person. The general theory of crime provides an explanation of self-control that is applicable across all forms of criminal and analogous behavior, as well as all populations. This theory assumes that all human behavior is contingent upon th e perceived costs and
21 rewards of the behavior. An individual who is impulsive, lazy, risky, physical, selfcentered, and unable to tolera te frustration possesses a lo w level of self-control. Depending on the opportunity present, these indi viduals will be more likely to engage in criminal and analogous acts, due to the perceived benefits of the behavior. Literature Review For more than a decade, research has cons istently revealed support for the general theory of crime. The bulk of empirical lite rature provides weak to moderate support for the generality, dimensionality, stability, a nd cause of low self-control. However, the ability of self-control to pr edict criminal and analogous behaviors is overwhelmingly supportive. In regard to Gottfredson and Hirschis (1990) notion of a g eneral theory of crime, Vazsonyi & Crosswhite (2004) perfor med a comparative analysis to examine the differences in the effect of self-control on Caucasian and African American youth and found a strong similarity in the ability of self-contro l to predict behavi or. Vazsonyi et al. (2001) looked at a cross-national sample of youth in four different countries and found that low self-control accounted for 17% to 28% of the vari ance in self-reported delinquency. In addition, low self-control has been found to be a sign ificant predictor of criminal behavior across several age groups including children (Brannigan et al. 2002) middle school students (Unnever & Corne ll 2003), high school students (Wood et al.1993, Sorenson & Brownfield 1995, Hay 2001, Pe rrone et al. 2004), college students (Cochran et al. 1998, Gibbs et al. 1998, Selle rs 1999, Arneklev et al. 1999) and adult samples (Burton et al. 1998, Burton et al. 1999). All of these findings support the ability of self-control to predict beha vior across diverse populations.
22 The literature regarding the dimensionality of self-control is quite complex. While several researchers (Grasmick et al. 1993, Burton et al. 1998, Cochran et al. 1998, Piquero et al. 1998, Arneklev et al. 1999) argue that self -control is a unidimensional construct, others (Wood et al. 1993, L ongshore et al. 1996, Longshore et al. 1998, Vazsonyi & Crosswhite 2004) claim that it is multidimensional. Wood et al. (1993) found that their self-control scale accounted for greater variance in a general delinquency scale than any one of the six dimension subscales alone. However, when the self-control and delinquency scales were disaggregated, diffe rent dimensions accounted for different types of behavior. Moreover, several studies have found that risk-seeking (Wood et al. 1993, Winfree & Bernat 1998, LaGrange & S ilverman 1999, Nakhaie et al. 1999) and impulsivity (Arneklev et a. 1999) are the strongest correlat es of criminal and analogous behaviors. Therefore, although most tests of the self-control scal es yield loadings consistent with a unidimensional construct, wh en the dimensions are analyzed separately, a different picture tends to emerge. Furthermore, only a handful of studies within the self-control literature have examined the stability of self-control. Polakowski (1994) Arneklev et al. (1998), Arneklev et al. (1999) and Burton et al. ( 1999) compared levels of self-control across age-groups and found evidence in support of Gottfredson and Hirschis (1990) stability hypothesis. Using seven waves of data from the NLSY, Turner & Piquero (2002) found stable group differences in levels of se lf-control for offenders versus nonoffenders. However, the results also revealed variati on in individual levels of self-control across time. This finding is important because it is based on longitudinal data and contradicts the notion of self-control as a stable trait. In sum, the scarce literature assessing the stability
23 of self-control provides weak to moderate support and indicates a need for further examination. The empirical literature dire ctly testing the interacti on of criminal opportunities and self-control has also provided moderate support. The interacti on of self-control and opportunity has been shown to provide greate r explanatory power than self-control or opportunity alone. Significant interactive effects have been found when testing criminal behavior, academic dishonesty, force, fraud, in timate partner violence, drug use, property crime, and violent offenses (Grasmick et al. 1993, Cochran et al. 1998, Longshore 1998, Longshore & Turner 1998, Sellers 1999, La Grange & Silverman 1999, Smith 2004). In addition, Nagin and Paternoster (1993) and Piquero and Tibbetts (1996) incorporated measures of routine activities into their models of self-control and offending and found that the interaction of self-control and r outine activities accounted for more of the variance in self-reported delinquency. Further, the few studies that have directly examined the causes of self-control (Hay 2001, Perrone et al. 2004, Pratt et al. 2004) provide moderate support for Gottfredson and Hirschis (1990) supposition th at ineffective parental monitoring and discipline are the sour ces of low self-control. Polakow ski (1994) found that parental monitoring at ages 8 to 10 signi ficantly predicted level of se lf-control at ages 12 to 14, which in turn, predicted official delinquency la ter in life. In additi on, two separate studies have been conducted to test the relationship between parental management and academic dishonesty. Gibbs et al. (1998) concluded that the effects of parental management on behavior are indirect through self-control, which is consis tent with the general theory of
24 crime. Cochran et al. (1998), on the othe r hand, found parental attachment to be significantly related to self-control, but not parental management. Regardless, the bulk of empi rical literature provides substantial evidence that, on average, low self-control is a significant pr edictor of criminal offending (Grasmick et al.1993, Wood et al. 1993, Gibbs & Giever 1995, Burton et al. 1998, Deng & Zhang 1998, Longshore 1998, Burton et al. 1999, Wright et al. 1999). Th e predictive ability of self-control has been supporte d with tests of specific offenses such as academic dishonesty, skipping school, under age smoki ng, traffic viola tions, drunk driving, vandalism, software piracy, property offens es, and predicted self-reported delinquency (Keane et al. 1993, Cochran et al. 1998, Gi bbs et al. 1998, LaGrange & Silverman 1999, Stylianou 2000, Vazsonyi et al. 2001, Higgins & Makin 2004, Smith 2004, Cauffman et al. 2005). Violent behaviors such as self -reported intimate partner violence, gang violence, sexual assault, bullying, and number of arrests have also been positively correlated with low self-c ontrol (Nagin & Paternoste r 1993, Sellers 1999, Chapple & Hope 2003, Unnever & Cornell 200 3, Piquero et al. 2005). Further, when other demographic variables such as race, gender, and religious participation are added into a model of crim inal behavior, the e ffect of self-control remains statistically significant (Wood et al. 1993, Sorenson & Brownfield 1995, Burton et al. 1998, Lynskey et al. 2000). Low self-c ontrol also remained significant after controlling for competing theoretical vari ables such as soci al bond, differential associations, social control, and strain (Sorenson & Brownfield 1995, Burton et al. 1998, Nagin and Paternoster 1993, Polakowski 1994, Piquero & Tibbetts 1996, Nakhaie et al. 1999, Schreck et al. 2002).
25 The literature on self-control and analogous behaviors is just as conclusive. In fact, Paternoster and Brame (1998) conduc ted a study of 369 adolescent males and concluded that the effect of self-control on analogous behaviors is approximately equal to its effect on criminal behavior. Accidental injuries, motor vehicle accidents, alcohol and drug use, gambling, and gang memb ership have been shown to be significantly associated with self-control (Arnekle v et al. 1993, Junger et al 1995, Sorenson & Brownfield 1995, Paternoster & Brame 2000, Burton et al 1998, Junger & Tremblay 1999). Moreover, negative social circumstances such as unemployment, interpersonal problems, homelessness, association with delinquent peers, poor school performance, failure in the criminal justice system, and low attachment to parents have also been found to be consequences of low self-control (Kra uss et al 2000, Delisi 2001, Baron 2003). To add to this extant body of literature, Pratt and Cullen (2000) conducted a metaanalysis of 21 studies that directly tested the relationship between self-control and criminal or analogous behavior. The analysis yielded strong empirical support for the general theory of crime. Not only did the re sults indicate self-controls strong predictive ability, they also provided evidence of the ge nerality of self-control across measures and behaviors. Based on this meta-analysis, Pra tt and Cullen (2000:953) concluded that selfcontrol must be considered an important pred ictor of criminal behavior and the general theory warrants a measure of acceptance. Self-Control and Victimization The extant literature has shown that the general theory of crime is predictive of a wide variety of deviant beha viors, as well as a broad ra nge of negative consequences. One negative consequence that merits closer examination within the context of self-
26 control is victimization. Given the strong em pirical correlation betw een victimization and offending, it is not unreasonable to explore the extent to whic h low self-control serves as a common thread linking both offending and victimization. Gottfredson and Hirschi (1990) maintain th at low self-control behavior brings immediate, easy and short-term satisfaction of desires; however, they also maintain that this type of behavior involves secondar y consequences for an individuals life. Vulnerability to crime is one such consequen ce. Drug users, for example, are less likely to be aware of their surrounding, defend themse lves, or guard their belongings. Offenders tend to associate with other offenders, whic h may involve untrustw orthy acquaintances. Persons with high self-control are more likely to be able to recogni ze these consequences and refrain from engaging in behaviors that increase vulnerability to crime. Based on the potential associa tion of self-control and vulnerability to crime, Forde and Kennedy (1997) tested a model that integr ated the general theory of crime with aspects of routine activities and risky lifestyles, in an attempt to provide a better explanation of both offending and victimization. Data came from telephone surveys of 2,052 persons living in Canada. Measures in cluded a self-control scale, routine activities/risky lifestyle measure, routine c onflict scale, several imprudent behaviors, whether or not the respondent has been vi ctimized, and number of arrests. Results indicated that respecifying the general theory of crime to include aspects of proximate causes, such as a risky lifes tyle or routine activities, provides a stronger explanatory model of both offending and victimization. Elaborating on these findings, Schreck (1999) proposed an extension of the general theory of crime that included vi ctimization. Using the six elements that
27 Gottfredson and Hirschi (1990) use to define self-control, and in turn criminal and analogous behaviors, Schreck (1999) explained of how these same characteristics could predispose an individual to victimization. The first element of self-control is relate d to an individuals level of impulsivity or tolerance for deferred gratification. Schr eck (1999) argued that individuals with low levels of deferred gratifica tion are less likely to perceive long-term consequences, and therefore will engage in behaviors that put themselves or th eir possessions in danger, due to the immediate rewards these behaviors are expected to provide. For example, an impulsive person is more likely to accept dr ugs from a stranger due to the immediate rewards of the high, rather than consider where the drug came from or whether or not it is safe. The second element, empathy, refers to the extent that an individual possesses genuine concern for and sensitivity to ot hers. A person low in empathy does not necessarily treat people poorl y, but rather their acts ar e not motivated by genuine kindness. These characteristics may lead an individual to possess few friends or close relationships. Schreck (1999) reasons, theref ore, that people low in empathy may not form good relationships with the people around them, for example their neighbors. According to routine activities, this would decrease guardia nship, and make an individual more vulnerable to victimization. The third element of self-control is tole rance for frustration. Pe ople with low selfcontrol do not a have a high tolerance for frustration. Frustration cause s an individual to become easily angered, upset, or quarrelsome (S chreck 1999). Potentially, this trait could lead to a hostile situation, and in turn, may result in a personal attack.
28 The fourth element of self-control is diligence. Individuals who have low selfcontrol tend to lack persistence and are char acterized by laziness. Lack of diligence may lead to inconsistency in taking safety precau tions, hence providing an easy target for a motivated offender. Preference for physical activity versus me ntal activity is the fifth component. Persons with low self-control tend to prefer physical activities, and accordingly, are less likely to use their mental capacity to asse ss the risks and possible consequences of a situation. The final component of se lf-control is risk seeking. Individual s with low selfcontrol are more inclined to seek out risky activities, such as nighttime activities or entering dangerous areas, which may put them in a more vulnerable position for victimization (Schreck 1999). To test his theory, Schreck (1999) surveyed a large sample of college students and found strong evidence to support his claim. Th e results showed that self-control had a significant direct effect on the chances of vi ctimization, and that th e effect of low selfcontrol also substantia lly reduced the effects of other demographic variables, such as gender, income, and criminal behavior. Ho wever, Schreck (1999, p 637) points out that his theory does not account for all victims of crime, but that those who engage in low self-control behavior risk greater vulnerability to crime. Building on these findings, Stewart et al (2004) tested Schrecks theory using data from a sample of 466 female offenders These authors created their own self-control scale using thirteen items and also included a measure of risky lifestyle into the model. Results indicate further support for Schreck s (1999) hypothesis. Women who reported
29 low levels of self-control reported higher levels of victimization. Ev en after controlling for demographic and lifestyle correlates of victimization, low self-control remained statistically significant. Finally, the empirical literature provides evidence that, on average, offenders seek out easy or vulnerable targets (Cohen & Fe lson 1979). Based on a rational choice model of offending, vulnerable targets minimize cost s by making the crime easier to commit. Regarding Gottfredson and Hirschi's (1990) elements of self-control, attacking a vulnerable target requires less mental planni ng and frustration, while at the same time facilitates the act, in turn, provi ding gratification at a faster rate. Taken together, it seems evident that low self-control influences an individuals tendency to become an easy target, and in turn, their vulne rability to victimization. This study elaborates on the findings of Forde and Kennedy (1997), Schreck (1999) and Stewart et al. (2004) by assessing the independent e ffect that low self-control has on violent victimization, the mediating re lationship of low self-control and risky lifestyle, and their in teractive effect on vi olent victimization. Although each of these studies (For de and Kennedy 1997, Schreck 1999, Stewart et al. 2004) provided evidence of an associa tion between self-control and victimization, they all have their s hortcomings. For instance, the meas ures used in Forde and Kennedys (1997) study are a limitation. They relie d on nighttime activities to measure lifestyle/routine activities and number of arre sts to measure crime risk. These measures do not take into account the fu ll range of the lifestyle con cept (Hindelang et al. 1978) or offending history. Therefore, the measures of Forde and Kennedys study lack content validity. Self-reported offending provides a much more accurate estimation of offending
30 history and risky lifestyle is conceptualized with various as pects of an in dividuals life (e.g. employment, living situa tion, location of residence). Stewart et al.s (2004) sample is a major limitation to their study. Using a nonrandom sample of adult female offenders lim its the ability to ge neralize the findings to any population other than the small subset of women who are involv ed in high rates of drug use and offending. Finally, Schreck (1999) di d not include a control for peer delinquency. The literature provides evidence that delinquent peer association is a signi ficant pred ictor of victimization (Lauritsen et al. 1992, Schreck et al. 2002, Schreck et al. 2004). Therefore, the significant relationships that were f ound in his study may be spurious, due to the uncontrolled effect of prior delinquency. Th e current study extends on Schrecks (1999) work by including a measure of peer delinquenc y, as well as a ris ky lifestyle component. The current study intends to improve th ese shortcomings in order to provide adequate evidence of an association between self-control and viol ent victimization. In addition, integrating risky lifestyle into the model will provide greater insight into the true relationship of self-cont rol and violent victimization. Put another way, incorporating risky lifestyle and gang membership into one multivariate model will help determine whether self-control has a direct effect on victimization, or an indirect effect, through its effect on lifestyle choices.
31 Chapter 4 Gang Membership and Risky Lifestyle The concept of lifestyle is based on an individuals daily activities, both vocational and leisure. Variations in daily ac tivities lead to variations in exposure to situations that have a high risk of victim ization. Individuals who engage in dangerous daily activities, such as drug use, staying out late, or association with criminal offenders, have a higher risk of victimization. Thus, r isky lifestyle is de fined as frequently engaging in activities that have a high risk of criminal victimization. For the purposes of this study, gang membership will be used to measure risky lifestyle. Although there is no universal definition for gang membership, gang involvement clearly leads to a risky lifestyl e. The empirical literature provides strong evidence that gang members participate in ris ky behaviors at a subs tantially higher rate than nongang members (Curry & Spergel 1992, Battin et al. 1998, Hill et al. 1999, Curry 2000). These risky behaviors include self-reported deli nquency, property offending, substance use, drug sales, gambling, more time spent with delinquent friends, and gun ownership (Curry & Spergel 1992, Esbensen & Huizinga 1993, Bjerregaard & Lizotte 1995, Battin et al. 1998, Esbensen & Winfr ee 1998, Hill et al. 1999, Curry 2000, Hope & Damphousse 2002). Gang members have also been found to report higher levels of violent offenses such as aggravated assault, sexual assault, intimate violence, and robbery (Battin et al. 1998, Hope & Dampho usse 2002, Thornberry et al. 2003).
32 In addition, individuals who engage in these types of beha viors tend to report victimization at a substantia lly higher rate (Esbensen & Huizinga 1991, Lauritsen et al. 1991, Sampson & Lauritsen 1990). Based on th ese findings, Curry et al. (2002) and Peterson et al. (2004) examined the effect of gang membership on victimization. Both studies revealed a significant association between gang membership and victimization. For instance, using a large sample of middl e school students, Pe terson et al. (2004:807) concluded that gang members were victims of violence at a higher rate than were youths who were never gang affiliated in each of th e years preceding and the years(s) following gang membership, and they tended to experience the most victimization in the year of membership. Furthermore, Hope and Damphousse (2002) ap plied the general th eory of crime to gang membership on a sample of 1,139 middl e and high school st udents. Not only did low self-control emerge as a significant predic tor of membership in a gang, but the effect also remained significant after controlling for delinquent behavior. Thus, it is expected that using gang membership as a form of ris ky lifestyle will provide evidence in support of Forde and Kennedys (1997) integrated theory of low self-control, risky lifestyle, and victimization. The current study elaborates on previous wo rk by testing a conceptual model that brings together situational and individua l factors and relates them to violent victimization. Specifically, this study is based on the findi ngs that low self-control is directly related to victim ization (Schreck 1999, Stewart et al. 2004) and Forde and Kennedys (1997) conclusion that integrati ng risky lifestyle/exposure and self-control into one theoretical model provides a stronger explanation for victimization.
As Figure 1 illustrates, self-control is the central concept in the model. Gang Membership Low Self-control Figure 1. Low Self-contr ol Predictive Model Victimization + + + + First, I propose that both self-control and ga ng membership will be directly related to violent victimization. Second, I evaluate the mediating effect of self-control and gang membership on violent victimization. Third, I explore the interacti on of these individual and situational factors, which I propose will reveal the greatest risk of violence. Hypotheses 1. Based on Schrecks (1999) findings, self-control has a direct effect on victimization. Individuals who report low levels of self-control will report higher prevalence and frequency of violent victimization. 2. Gang membership will have a direct eff ect on violent victimization. Research has shown that gang members are victimized at a significantly higher rate than nongang members (Curry et al. 2002, Peterson et al. 2004). Hence, individuals who report higher levels of gang involvement will report higher prevalence and frequency of violent victimization. 3a. Integrating the findings that self-cont rol is a significan t predictor of gang membership (Hope & Damphousse 2002) and that gang members report higher levels of victimization than nongang me mbers (Peterson et al. 2004), it seems 33
34 plausible that the effect of self-control on victimization will be mediated by gang membership. That is, the effects of self -control on prevalen ce and frequency of violent victimization will be signific antly reduced once gang membership is added into the model. 3b. On the other hand, according the Gottfred son and Hirschi (1990), the effects of gang membership on violent victimization will be significantly reduced once the effect of self-control is intr oduced as a control variable. 4. Elaborating on Forde and Kennedys (1997) extension of the ge neral theory of crime to include risky lifestyle, self-c ontrol and gang membership will have an interactive effect on victimization. Specifi cally, the effect of low self-control on violent victimization will be stronger for gang members than nongang members.
35 Chapter 5 Methods Data came from Esbensen and Osgoods (1999) evaluation of the Gang Resistance Education and Trai ning Program (G.R.E.A.T.), a gang prevention program for middle school students. This program was funded by the Bureau of Alcohol, Tobacco and Firearms, taught by uniformed law enforcemen t officers, and consisted of a nine-week curriculum that focused on goal setting, cultural differences, resi stance to peer pressure, drug education, conflict reso lution, and the consequences of criminal behavior. Sites in which the program was delive red during the 1993-1994 school year were chosen for inclusion into the initial evaluation. This resulted in particip ants from fortytwo schools at eleven sites, including Las Cruces, NM; Omaha, NE; Phoenix, AZ; Philadelphia, PA; Kansas City, MO; M ilwaukee, WI; Orlando, FL; Will County, IL; Providence, RI; Pocatello, ID; and Torrance, CA. These sites provide a very diverse sample ranging from large urban areas to small rural cities, racially homogenous to racially heterogeneous areas, and worki ng to upper class families (Esbensen and Deschenes 1998). Attendance rates on the day of the surv ey ranged from 75 to 93 percent and participation ranged from 98 to 100 percent of the students who were present the day the surveys were administered. The original sample consisted of 5,935 students (Esbensen and Deschenes 1998). Cases that included unanswered questions pertaining to the
36 measures used in this analysis were exclude d from the original sample. As a result, the final sample used in this study is 3,907 middle school students. Demographic characteristics of the sample are presented in Table 1. Table 1: Demographics of the Sample Variable N = 3907 Percentage Mean Sex Male 1848 47.3 Female 2059 52.7 Race/Ethnicity White 1820 46.6 Black 839 21.5 Hispanic 665 17.0 American Indian 88 2.3 Asian 264 6.8 Other 61 1.6 Mixed 170 4.4 Age 13.78 12 or younger 9 .3 13 1201 30.7 14 2369 60.6 15 313 8.0 16 11 .3 17 or older 4 .1 Gang Membership Yes 311 8.0 No 3596 92.0 Self-control 23.40 Due to the large sample size and diversity of the sites, the G.R.E.A.T. data have been analyzed for several types of research questions including th e relationship between gender and gang membership, attitudes to ward police officers, race and gender differences among gang and nongang members, definitional issues pertaining to gang membership, and the role of single-parent families on juvenile delinquency (Esbensen & Deschenes 1998, Esbensen & Winfree 1998, Deschenes & Esbensen 1999, Esbensen et
37 al. 1999, Esbensen et al. 2001a, Taylor et al. 2001, Anderson 2002). In addition, the G.R.E.A.T. data have also been used to evaluate the theoretical significance of the general theory of crime, social learning, and social bonding theory (Lynskey et al. 2000). However, a few limitations regarding the da ta should be noted. First, although this sample is very diverse, it is not truly random because the sites were chosen on the basis of whether or not the G.R.E.A.T. program was implemented in a certain school year. Furthermore, the sample consists of th e students who attended school that day, and therefore does not include students who were absent, su spended, expelled, or had dropped out, which in turn, could lead to an under representation of high-risk youth. Secondly, these data are cro ss-sectional and based on self -reported measures. Not only does this limit our ability to make causal infe rences, but it also warrants some skepticism when relying on teenagers memory and honest y. Despite these limitations, the diversity of the sample allows for exploration of a large sample of mi ddle school youth. Victimization The primary dependent variable is victim ization. Victimization is defined as the exploitation or harm suffered by an individual, resulting from the actions or behaviors of other individuals. For this pa rticular study, prevalence and fre quency of victimization are measured using eight different variables. Prevalence was measured by whether or not, in the past twelve months, the respondent experienced one or more of the following: hit by someone purposely trying to hurt them, robbed, or attacked with a w eapon. All three items were recoded into dichotomous variables and responses were coded 0 = never been victimized and 1 = victimized one or more times. Then, these three variables were summed and
38 dichotomized into an overall violent victim ization scale. Approximately half (51.1%) of the sample reported that they have been victim ized one or more times in the past twelve months. Although prevalence of victimization serv es as a sound measure of victimization, including frequency of victimi zation into the analysis is necessary for several reasons. First, there is a great likeli hood that anyone will be victimized once in his or her lifetime. However, as the frequency of victimization in creases, so does the possibility that certain factors are influencing the occurrence of these experiences. In addition, since this data set does not include information on the serious ness of the victimization experience, frequency of victimization could also be used as an indicator of seriousness because the more times an individual is victimized, th e more serious these experiences become. Frequency of victimization was measured using an open-ended question regarding how many times, in the past twelve months, th e respondent experienced each of the three victimization items. Responses ranged from zero to more than 989, included dont know and positive-unreadable answers. We chose to adopt Peterson et al.s (2004) decision to right-censor the response s at 12 or higher. The premise is that more than 12 victimization experiences in one year consti tutes high-frequency victimizat ion (Peterson et al. 2004). First, dont know and unreadable responses were coded to missing. Then, after all of the items were right censored at 12 or higher, the dont know and positive unreadable responses were recoded from mi ssing to the positive integer (whole number greater than zero) nearest to the mean of each item (robbery: mean = .22, recoded to 1.0; hit on purpose: mean = 1.77, recoded to 2.0; attacked with a weapon: mean = .32, recoded to 1.0). By interpolati ng the closest positive integer, the intervals of the variable
39 remain the same and the case can still be included in the an alysis (Kalton 1983, Studenmund 2001). Finally, responses to these three items were summed and recoded into one overall count of victimization. This ove rall count of victimization is also right-censored at 12 or more. Of the respondents, 295 (6.7%) experi enced twelve or more victimization experiences. Self-control The central explanatory variable is self -control, which is referred to as the propensity of the actor to seek short-term, immediate pleasure without consideration of the long-term consequences. Gottfredson and Hi rschi (1990) explain that an individual who is impulsive, lazy, prefers risk and physical activity, lacks empathy and becomes easily frustrated possesses a low level of self-c ontrol and in turn, will engage in criminal and analogous behavior at a much higher ra te than individuals who do not possess these traits. The current standard measure of self-cont rol is the Grasmick et al. (1993) scale, which is comprised of twenty-four items th at measure all six components described by Gottfredson and Hirschi (1990). However, the G.R.E.A.T. data were not intended to measure self-control, specifically as Go ttfredson and Hirschi did, and therefore are limited to only two components, impulsivity and risk seeking. Although this will be a major limitation to the measure of self-control and its ability to co rroborate Schrecks (1999) research, risk seeking and impulsivity have been shown to be the dimensions of self-control that carry the most explanatory power (Arn eklev et al. 1993, Wood et al.
40 1993, Winfree & Bernat 1998, Arneklev et al 1999, Nakhaie et al. 1999, LaGrange & Silverman 1999). Moreover, the items used to measure these two components were taken from the Grasmick et al. scale (1993). However, th ey include a neither agree nor disagree answer choice, whereas the Grasmick scale do es not. Both impulsivi ty and risk seeking are measured on a Likert scale that ranges from one for strongly disagree to five for strongly agree. Impulsivity is reflected by the respondent s level of delaye d gratification and here and now orientation. Four items are us ed to measure this component. I often act on the spur of the moment; I dont devote mu ch time to preparing for my future, I often do whatever brings me pleasure here and now, and Im more concerned with what happens to me in the short run. The scale produced from these items revealed a Cronbachs alpha of .64 and formed a singlefactor solution (eig envalue = 1.93) with loadings ranging from .62 to .74. Risk seeking corresponds to the respondents preference for risk, excitement, and adventure. It is also operati onalized with four items; I like to test myself every now and then by doing something risky, sometimes I will take a risk for the fun of it, I sometimes find it exciting to do things for wh ich I might get in trouble, and excitement and adventure are more important to me than security. These items yielded a Cronbachs alpha of .82 and formed a single-factor solu tion (eigenvalue = 2.60) with loadings that ranged from .73 to .87.
41 Finally, both summated scales are used to create one eight-item self-control scale. Analysis revealed a single-f actor solution, with an eigenva lue of 1.48, loadings of .86 for both dimensions and a Cronbachs alpha of .80. Gang Membership Gang membership is defined as whether or not the respondent identifies oneself as part of a gang. It is included as a proxy vari able to measure risky lifestyle. This variable will be used as an independent variable to explain violent victimization and to examine the relationship between self -control, gang membership, a nd violent victimization. Gang membership is measured dichotomously, using one self-definitional question; Are you now in a gang? Although this method of identifying gang members has some definitional limitations, Winfree et al. (1992) conducted a study specifically designed to test various definitions of gang st atus and their ability to predict behavior. They found that the self-reported definition of gang membership was a better pred ictor of gang related activity compared to the more restrictive definition, which was based on a series of questions regarding initiation rite s and gang symbols. Thus, self-definitional gang membership has been used in previous research and is an accepted measure in the gang literature (Klein 1995, Hope and Damphousse 2002, Peterson et al. 2004). Responses are 0 = no and 1 = yes. Of the res pondents, 311 (8.0%) stat ed that they were current gang members at the time of data collection. Control Variables Measures of the respondents age, ra ce, gender, self-reported delinquency, delinquent friends, parental attachment, parent al supervision, and participation in the G.R.E.A.T. program are used as control variab les for this analysis. All of these variables
42 have been found to be significantly related to the risk of victimization (Esbensen & Huizinga 1991, Lauritsen et al. 1991, Lauritsen et al. 1992, Lauritsen & Quinet 1995, Schreck et al. 2002, Sampson & Lauritsen 1990), and therefore have the potential to influence the results of the study. Gender is a dichotomous variable coded 1 = male and 0 = female. Age is a continuous variable. The or iginal survey provides seven categories of race/ethnicity: White, Black, Hispanic, Asian, American Indian, other, and mixed. These categories were then recoded into five dummy variables, using white as the reference group. Due to the small number of responses, other and mixed were combined into one dummy variable named other. Association with delinquent peers is meas ured using sixteen items regarding how many of the respondents friends participated in various behaviors including skipping school, stealing, robbery, and selling and us ing drugs. Responses were based on an ordinal scale ranging from 1= none of them to 5 = all of them. Three components revealed eigenvalues greater than one; however the first component yielded an eigenvalue of 8.77 and explained 55% of the variance, whereas the second and third components yielded eigenvalues of 1.24 and 1.06, respectively. Based on the scree discontinuity test (Cattell 1966) the first item explained the most variance and therefore, was the only component extracted. Factor loadin gs for this component ranged from .62 to .83. Cronbachs alpha equaled .94. Prior delinquency is operationalized usi ng twenty-three items regarding whether nor not, in the past year, the respondent enga ged in various activitie s including; skipping school, purposely damaging property, drug use, carrying or using a weapon, stealing, and selling drugs. These items were measured as dichotomous variables, 0 = no and 1 = yes.
43 Factor analysis of these seventeen items pr oduced a single-factor solution (eigenvalue = 7.23) with loadings from .40 to .70. Cronbachs alpha yielded an internal consistency of .88. Parental attachment is m easured with twelve Likert -type items. Six questions were asked regarding the respondents rela tionship with a mother-figure and fatherfigure, separately. These items measure the extent to which the respondent felt each parent could talk about anyt hing with them, knew his or her friends, understood, trusted, praised, and gave advice. Factor analysis of these twelve it ems revealed a singlefactor solution with an eigenvalue of 5.33 and loadings from .55 to .76. Cronbachs alpha was .89. Parental supervision is operationalized with four Likert-type items regarding whether or not the respondent agrees or disa grees with the following statements; when I go someplace, I leave a note or call my parents, my parents know where I am when I am not at home, my parents know who I am w ith when I am not at home, and I know how to get in touch with my parents if they are not home. Factor analysis formed a single-factor solution (eigenvalue = 2.22) and loadings that ranged from .33 -.68. Cronbachs alpha indicated an intern al consistency of .73. In an analysis of longitudinal data, participation in the G.R.E.A.T. program produced a significant reduction in victimi zation (Esbensen et al. 2001b). However, a similar evaluation of the cross-sectional data (data used in this study) did not reveal a significant difference in victimization (E sbensen & Osgood 1999). While these findings are equivocal, we include a measure of particip ation as a control variable. This variable
44 makes a dichotomous distinction between part icipation (= 1) and nonparticipation (= 0). Of the sample, 45.4% (1994) reported part icipation in the G.R.E.A.T. program. Analysis The prevalence of violent victimization is measured as a dichotomous dependent variable, indicating that logistic regression is the appropriate technique to use. Initially, separate models for each of the four measures of victimization are examined to determine the independent effects of self-control on the prevalence of violent victimization (hypothesis 1) and gang membership on th e prevalence of violent victimization (hypothesis 2). Next, a multivariate model is em ployed to test the effects of self-control on the prevalence of violent victimization, controlling for gang membership. Hypothesis 3 (mediating effect) will be s upported if the effects of self -control on victimization are significantly reduced in the multivariate model. To test the fourth hypothesis, separate models splitting the gang members fro m the nongang members are examined. These models allow a comparison of the effects that self-control has on violent victimization for gang members versus nongang members. A ztest of the maximum likelihood estimator (Brame et al. 1998) is employed to estimate the statistical significance of the difference in coefficients for the two groups. This will determ ine whether or not an interactive effect is present (hypothesis 4). Next, negative binomial regression is used to examine the effect of self-control and gang membership on the frequency of vi olent victimization. Ordinary Least Squares (OLS) regression is inappropriate for this particular analysis because the data are discrete counts and reveal a skewed di stribution with many of th e observations at zero. As a result, OLS would yield smaller standard erro rs, inflated t-values, and a statistically
45 significant effect, when in fact there is not an effect present (Studenmund 2001). Poisson regression would seem to be the better choice because this technique assumes discrete counts, has a skewed distribu tion, and restricts predicted values to positive numbers (Kleinbaum et al. 1988). However, Poiss on also assumes that the variance of the dependent variable equals the mean; the data in this study do not satisfy this assumption (Haight 1967). Significant vari ation in frequency of victim ization is revealed (robbed: mean = .22, SD = 1.13; hit: mean = 1.77, SD = 3.13; attacked: mean = .32, SD = 1.36, overall: mean = 2.07, SD = 3.38). Thus, negati ve binomial regression is the preferred technique to use in th is situation because it includes a random component reflecting the uncertainty about the true rates at which events occur for individual cases and ensures an accurate estimate of the probability distribut ion of each variable (Gardner et al. 1995, Long & Freese 2003). The first model in the series of negati ve binomial regression analyses includes separate analyses for the effect of self-cont rol on each of the four victimization counts (hypothesis 1). The second model includes analyses testing the effect of gang membership on each of the victimization counts (hypothesis 2). Next, a multivariate model tests the effect of se lf-control on each victimizati on count, controlling for gang membership. If hypothesis 3 is supported, the effect of self-control will be significantly reduced in the multivariate model. The final model includes separate analyses for gang members and nongang members. If necessary, the coefficients and standard errors for each model will then be used to calculate a z-score, which will determine whether or not self-control and gang membership interact to enhance the frequency of violent victimization (hypothesis 4).
46 Chapter 6 Results The purpose of this study was to examine several multivariate models that predict prevalence and frequency of violent vict imization. The question addressed by these models is as follows: does low self-control and gang membership influence prevalence and frequency of violent victimization? Logi stic regression was utilized to examine the prevalence of violent victimization and negative binomial regression was used to examine frequency of violent victimization. Preliminary analyses were run to assess the bivariate relationship between selfcontrol and each item, gang membership and each item, and the multivariate relationship between self-control, gang membership and each victimization item. Next, age, gender, race, participation in the G.R.E.A.T. program parental attachment, parental supervision, peer delinquency and prior delinquency were incorporated into th e model as control variables. Three more models of each item were analyzed. Model 1 includes the control variables and self-cont rol, Model 2 includes the control variables and gang membership, and Model 3 includes the control variables a nd both self-control and gang membership. Finally, separate models for ga ng and nongang members are provided. Prevalence of Violent Victimization Separate tables are displayed for each prevalence measure (ever been hit by someone trying to hurt you, ever been robbed, ever been attacked with a weapon, ever
47 been victimized). Descriptive statistics reve aled that 48.3% of the sample reported being hit by someone trying to hurt them, 8.3% had been robbed one or more times, and 10.7% had been attacked with a weapon. Of the respondents, 51.1% reported violent victimization one or more times in the past year. Tables 2-5 display the logistic regr ession results, without controlling for extraneous factors. All of the model chisquares were significant at the .05 level, indicating that each model was able to predic t victimization better than chance. Model 1 in each table revealed a statistically significan t effect of self-contro l. It appears that individuals with lower levels of self-control are more likely to report victimization one or more times. Model 2 of each table also reve aled a significant bi variate relationship between gang membership and victimization. Gang members appeared to be more likely to report being hit, robbed, a ttacked, or one or more violen t victimization experiences. However, in Model 3 in all of the tables, th e effects of gang membership and self-control were significantly reduced. That is, incl uding the self-control and gang membership variables into one model significantly reduced the independent effect that each variable had on victimization. Table 2: Logistic Regression Predicting Prevalence of Hit by Someone Trying to Hurt You (N = 3907) Variables Model 1 Model 2 Model 3 B SE Exp (B) B SE Exp (B) B SE Exp (B) Self-Control .06* .01 1.06 ---.06* .01 1.06 Gang Membership ---.68* .12 1.98 .41* .13 1.51 Constant -1.48* .14 .23 -.12* .03 .89 -1.41* .14 .24 116.45* 32.33* 127.29* Nagelkerke R .04 .01 .04 p .05
48 Table 3: Logistic Regression Predicting Prevalence of Robbery (N = 3907) Variables Model 1 Model 2 Model 3 B SE Exp (B) B SE Exp (B) B SE Exp (B) Self-Control .08* .01 1.08 ---.06* .01 1.06 Gang Membership ---1.31* .15 3.70 1.03* .16 2.81 Constant -4.22* .27 .02 -2.56* .07 .08 -3.90* .27 .02 54.35* 62.77* 91.57* Nagelkerke R .03 .04 .05 p .05 Table 4: Logistic Regression Predicting Prevalence of Attacked with a Weapon (N = 3907) Variables Model 1 Model 2 Model 3 B SE Exp (B) B SE Exp (B) B SE Exp (B) Self-Control .13* .01 1.13 ---.09* .01 1.10 Gang Membership ---2.09* .13 8.11 1.70* .14 5.48 Constant -5.23* .27 .01 -2.44* .06 .09 -4.71* .27 .01 176.17* 224.68* 313.81* Nagelkerke R .09 .11 .16 p .05 Table 5: Logistic Regression Predicting Prevalence of Violent Victimization (N = 3907) Variables Model 1 Model 2 Model 3 B SE Exp (B) B SE Exp (B) B SE Exp (B) Self-Control .07* .01 1.07 ---.06* .01 1.06 Gang Membership ---1.02* .13 2.79 .75* .14 2.11 Constant -1.51* .14 .22 -.03 .03 .97 -1.40* .14 .25 140.81* 67.07* 173.25* Nagelkerke R .05 .02 .06 p .05 However, as seen in Tables 6-9, once th e control variables were entered into the models, a very different picture emerged. Table 6 provides logistic regression models for ever been hit by someone trying to hurt you, including the controls. All three models
49 yield significant chi-squares. Th e independent effect of self-control is presented in Model 1. The coefficient for self-c ontrol (B= .004) is not signifi cant at the .05 level. This indicates that, once all other ex traneous effects were controlled, level of self-control did not have a statistically signi ficant effect on the likelihood of being hit by someone. Model 2 reveals a significant effect of gang member ship. However, it is not in the hypothesized direction. The B-coefficient equals -.543, which indicates that nongang members were significantly more likely to re port being hit by someone tryi ng to hurt them. This finding is contrary to the results in Table 2. When the control vari ables were not included in the analyses gang membership yielded a positive effect. As seen in Model 3, the effect of gang membership remained statistically signi ficant once self-contro l was added into the model. Throughout all three models in Table 6, gender, Hispanic, parental attachment, and prior delinquency revealed statistically si gnificant coefficients. Males, offenders, and respondents who reported lower leve ls of parental attachment were more likely to report being hit by someone; Hispanics and respondent s with fewer delinquent peers were less likely to report ever being hit by someone trying to hurt them. Table 6: Logistic Regression Predicting Prevalence of Hit by Someone Trying to Hurt You Including the Control Variables (N = 3907) Variables Model 1 Model 2 Model 3 B SE Exp (B) B SE Exp (B) B SE Exp (B) Age -.06 .06 .94 -.05 .06 .96 -.05 .06 .95 Gender .68* .07 1.97 .68* .07 1.97 .68* .07 1.97 Black -.10 .09 .91 -.09 .09 .91 -.09 .09 .91 Hispanic -.23* .10 .79 -.21* .10 .81 -.21* .10 .81 Indian .23 .24 1.26 .23 .24 1.26 .28 .24 1.26 Asian -.13 .14 .88 -.11 .14 .90 -.10 .14 .90 Other .08 .15 1.08 .10 .15 1.10 .10 .15 1.10 G.R.E.A.T. -.04 .07 .96 -.04 .07 .96 -.04 .07 .96
50 Table 6 (Continued) Variables Model 1 Model 2 Model 3 B SE Exp (B) B SE Exp (B) B SE Exp (B) Parental Attachment -.01* .00 .99 -.01* .00 .99 -.01* .00 .99 Parental Supervision .02 .01 1.02 .02 .01 1.02 .02 .01 1.02 Peer Delinquency -.01* .00 .99 -.01* .00 .99 -.01* .00 .99 Prior Delinquency .16* .01 1.17 .17* .01 1.19 .17* .01 1.18 Self-Control .00 .01 1.00 ---.00 .01 1.00 Gang Membership ----.54* .15 .58 -.54* .15 .58 Constant .35 .85 1.42 .19 .84 .58 .20 .85 1.13 485.28* 497.41* 497.58* Nagelkerke R .16 .16 .16 p .05 Table 7 summarizes prevalence of robbe ry. Model chi-squares for each of the three models were statistically significant at the .05 level. However, self-control (B = -.008) and gang membership (B = .0 11) did not reveal significant (p .05) coefficients. In fact, the association of se lf-control was opposite the directi on that would be expected. Regardless, the likelihood of being robbed was not related to level of self-control or gang membership. Peer delinquency, prior delinquency, bl ack, Hispanic, and gender remained significant predictors of robbery in all three models. Males, blacks, and offenders were more likely to report being robbed one or more times and being Hispanic and higher levels of delinquent peers decr eased the likelihood of robbery.
51 Table 7: Logistic Regression Predicting Prevalence of Robbery Including the Control Variables (N = 3907) Variables Model 1 Model 2 Model 3 B SE Exp (B) B SE Exp (B) B SE Exp (B) Age .03 .10 1.03 .03 .10 1.03 .03 .10 1.03 Gender .73* .13 2.08 .73* .13 2.07 .73* .13 2.08 Black .55* .15 1.73 .56* .15 1.75 .55* .15 1.73 Hispanic -.56* .20 .57 -.56* .21 .57 -.56* .21 .57 Indian .63 .34 1.87 .62 .34 1.86 .63 .38 1.87 Asian -.19 .31 .83 -.18 .31 .84 -.19 .31 .83 Other .29 .24 1.34 .30 .24 1.34 .29 .24 1.34 G.R.E.A.T. .05 .12 1.05 .05 .12 1.06 .05 .12 1.05 Parental Attachment .00 .01 1.00 .00 .01 1.00 .01 .01 1.00 Parental Supervision -.03* .02 .97 -.03 .02 .97 -.03 .02 .97 Peer Delinquency -.01* .01 .99 -.01* .01 .99 -.01* .01 .99 Prior Delinquency .17* .02 1.18 .17* .02 1.18 .17* .02 1.18 Self-Control -.01 .01 .99 ----.01 .01 .99 Gang Membership ---.01 .20 1.01 .01 .20 1.01 Constant -3.53* 1.43 .03 -3.74* 1.39 .02 -3.53 1.43 .03 293.14* 292.78* 293.14* Nagelkerke R .17 .17 .17 p .05 Results for prevalence of attacked w ith a weapon are displayed in Table 8. Overall, each model was significant. Model 1 shows that self-control did not yield a significant effect (B = .015) at the .05 level. Model 2 reveals a gang coefficient equal to .507, which was significant and in the hypothesized direction. Gang members were more likely to report being attacked with a wea pon, one or more times. When self-control was added into Model 3, the gang membership coefficient remained statistically significant. Thus, a mediating effect was not found. All three models found gender, prior deli nquency, and parental supervision to be significantly related to prevalence of attacked with a weapon. Males, offenders, and
52 respondents who reported less pare ntal supervision were more likely to experience being attacked with a weapon. Regarding race: black, Indian, and respondents in the other/mixed category were also more likely to report being attacked with a weapon. Table 8: Logistic Regression Predicting Prevalence of Attacked with a Weapon Including the Control Variables (N = 3907) Variables Model 1 Model 2 Model 3 B SE Exp (B) B SE Exp (B) B SE Exp (B) Age .20 .09 1.22 .18 .10 .06 .18 .10 1.19 Gender .53* .12 1.70 .55* .12 1.73 .54* .13 1.72 Black .71* .15 2.03 .67* .15 1.96 .70* .15 2.01 Hispanic .23 .17 1.26 .19 .17 1.21 .20 .17 1.22 Indian .81* .32 2.24 .85* .32 2.34 .84* .32 2.31 Asian -.43 .35 .65 -.51 .35 .60 -.49 .35 .61 Other .68* .22 1.97 .63* .22 1.88 .64* .22 1.90 G.R.E.A.T. .12 .12 1.13 .12 .12 1.13 .12 .12 1.13 Parental Attachment .00 .00 1.00 -.00 .00 .99 -.00 .01 .99 Parental Supervision -.05* .02 .96 -.05* .02 .95 -.04* .02 .96 Peer Delinquency -.03 .01 .99 -.00 .01 .99 -.01 .01 .99 Prior Delinquency .19* .02 1.21 .18* .02 1.20 .18* .02 1.19 Self-Control .02 .01 1.02 ---.02 .01 1.02 Gang Membership ---.51* .17 1.66 .51* .17 1.67 Constant -6.44* 1.39 .00 -5.62* 1.35 .00 -6.08* 1.40 .00 577.37* 584.55* 586.17* Nagelkerke R .28 .28 .28 p .05 Table 9 presents the resu lts for prevalence of the ove rall violent victimization measure. The chi-square for each model was stat istically significant at the .05 level. Still, the coefficients for self-control (B = .005) and gang member ship (B = -.267) are not statistically significant. Neither level of se lf-control nor gang membership was able to predict prevalence of vi olent victimization.
53 In all three models of Table 9, males and offenders were more likely to report one or more violent victimization experiences. Parental attachment was also found to have a significant inverse effect. That is, lower levels of parental attachment were found to be associated with one or more violent victim ization experiences. In addition, contrary to previous research (Schreck et al. 2002, Lauritsen et al. 1992, Schreck & Fisher 2004), delinquent peer association was inversely re lated to this item. Respondents with higher levels of delinquent peers were less likely to report a violent vi ctimization experience. Table 9: Logistic Regression Predicting Prevalence of Violent Victimization Including the Control Variables (N = 3907) Variables Model 1 Model 2 Model 3 B SE Exp (B) B SE Exp (B) B SE Exp (B) Age -.04 .06 .96 -.03 .06 .97 -.03 .06 .97 Gender .74* .07 2.10 .74* .07 2.10 .74* .07 2.10 Black .05 .09 1.05 .05 .09 1.05 .05 .09 1.06 Hispanic -.20 .10 .82 -.19 .10 .83 -.19 .10 .83 Indian .30 .24 1.34 .30 .24 1.35 .29 .24 1.34 Asian -.12 .14 .89 -.11 .14 .89 -.11 .14 .90 Other .19 .16 1.21 .20 .16 1.22 .20 .16 1.22 G.R.E.A.T. -.06 .07 .94 -.06 .07 .94 -.06 .07 .94 Parental Attachment -.01* .00 .99 -.01* .00 .99 -.01* .00 .99 Parental Supervision .01 .01 1.01 .01 .01 1.01 .01 .01 1.01 Peer Delinquency -.02* .00 .99 -.02* .00 .99 -.02* .00 .99 Prior Delinquency .18* .01 1.20 .19* .01 1.21 .19* .01 1.21 Self-Control .01 .01 1.01 ---.00 .01 1.00 Gang Membership ----.27 .16 .77 -.26 .16 .77 Constant -.05 .87 .95 -.05 .85 .95 -.16 .87 .86 594.36* 596.63* 596.98* Nagelkerke R .19 .19 .19 p .05 These logistic regression results do not support hypothesis 1. Self-control was not found to be a significant predictor of any of the four prevalence items. Hypothesis 2 was
54 only slightly supported. Gang members were more likely to report being attacked with a weapon. However, they were significantly less li kely to be hit by someone trying to hurt them and not associated with robbery and ove rall violent victimization. Given these null findings, Hypothesis 3 (mediating effect) was also unsupported. Next, the interactive effect proposed in hypothesis 4 was tested by running separate logistic regression models for gang members and nongang members for the two prevalence items that revealed an associa tion with gang membership (hit by someone trying to hurt you and attacked with a weapon). As seen in Tables 10 and 11, both the gang (hit: B = .026, attack: B = .007) and nongang (hit: B = -.001, attack: B = .017) models yielded self-control coefficients that were not significant at the .05 level. Therefor e, calculation of the z-score of the maximum likelihood estimator (Brame et al. 1998) wa s not necessary. Based on the insignificant findings of self-control, it can be conclude d that gang membership and self-control did not interact to amplify the risk of violent victimization. Regardless of gang membership, self-control did not have an effect on preval ence of violent victimi zation, once the effects of all the other extraneous influences were controlled. Table 10: Logistic Regression Predicting Prevalence of Hit by Someone Trying to Hurt You by Gang Membership (N = 3907) Variables Gang Members ( N = 311) Nongang Members (N =3596) B SE Exp (b) B SE Exp (b) Z-Score Age .12 .19 1.13 -.07 .06 .94 -Gender -.09 .27 .91 .74* .08 2.10 -Black -.36 .36 .70 -.08 .10 .92 -Hispanic -.20 .34 .82 -.22* .11 .80 -Indian .61 .88 1.85 .16 .25 1.17 -Asian 1.76* .81 5.83 -.20 .15 .82 -Other .25 .45 1.28 .07 .16 1.08 -G.R.E.A.T. -.09 .26 .92 -.04 .07 .96 -Parental Attachment .00 .01 1.00 -.01* .00 .99 -
55 Table 10 (Continued) Variables Gang Members ( N = 311) Nongang Members (N =3596) B SE Exp (b) B SE Exp (b) Z-Score Parental Supervision -.01 .04 .99 .03 .01 1.03 -Peer Delinquency -.02 .01 .98 -.01* .00 .99 -Prior Delinquency .10* .04 1.11 .19* .01 1.20 2.87* Self-Control .03 .03 1.03 -.00 .01 .99 -Constant -1.91 2.86 .15 .32 .90 1.38 -27.37* 474.02* Nagelkerke R .12 .17 p .05 Table 11: Logistic Regression Predicting Prevalence of Attacked with a Weapon by Gang Membership (N = 3907) Variables Gang Members ( N = 311) Nongang Members (N=3596) B SE Exp (b) B SE Exp (b) Z-Score Age .37 .21 1.45 .12 .11 1.13 -Gender .79* .29 2.20 .49* .14 1.63 .65 Black .57 .40 1.76 .70* .16 2.01 -Hispanic .24 .36 1.27 .17 .19 1.19 -Indian -.68 .99 .51 1.04* .33 2.83 -Asian -.19 .63 .83 -.70 .44 .50 -Other 1.01* .46 2.76 .51* .26 1.67 .68 G.R.E.A.T. .09 .27 1.09 .14 .13 1.15 -Parental Attachment -.00 .01 .99 -.00 .01 .99 -Parental Supervision -.01 .04 .99 -.06* .02 .95 -Peer Delinquency .01 .01 1.01 -.01 .01 .99 -Prior Delinquency .14* .04 1.15 .19* .02 1.20 1.12 Self-Control .01 .03 1.01 .02 .01 1.02 -Constant -8.73* 3.11 .00 -5.02* 1.59 .01 -73.40* 298.12* Nagelkerke R .28 .19 p .05 Frequency of Violent Victimization Descriptive statistics were used to de termine the distribution of each frequency item. Frequency of hit by someone trying hurt you revealed a mean of 1.78 and standard deviation of 3.13. The mean and st andard deviation for robbery was .22 and 1.12, respectively. Frequency of attacked w ith a weapon produced a mean of .30 and
56 standard deviation of 1.37. Finally, the overa ll measure of violent victimization scale yielded a mean of 2.07 and standard deviation of 3.38. Tables 12-20 summarize negative binomial regression analyses for each of the four frequency items. Preliminary analyses without the control variables are provided first. Multivariate models including all of the control variables follow. The alpha ( ) parameter at the bottom of the table determines the degree of dispersion. An alpha that is greater than zer o indicates overdispersion in the dependent variable and, in turn, the need to use the negative binomial regression model instead of the Poisson regression model (Long & Freese 2003). The model chi-square ( ) represents the signifi cance of the model (p .05), which indicates if the model is able to explain more than chance alone. The coefficient, standard error, and percent change of each variable are provided. Tables 12-15 summarize the independent relationship of self-control, gang membership, and frequency of violent victimization. Overall, all of the models yielded significant chi-squares and alpha parameters greater than zero. Further, self-control (Models 1) was statistically significant in a ll four tables. For each additional increase in self-control (indicating lower le vels of self-control), it a ppeared that the expected frequency of being hit, attacked, robbe d, or violently victimized increased. Gang membership (Models 2) was also significantl y associated with all four items. Being a gang member appeared to increase the expected frequency of being hit, robbed, attacked with a weapon, or violently victimized. However, when self-control and gang membership were incorporated into one multivariate model (Models 3), the effect of gang
57 membership on each of the four items was signi ficantly reduced, while the effect of selfcontrol was only slightly reduced. Table 12: Negative Binomial Regression Predictin g Frequency of Hit by Someone Trying to Hurt You (N = 3907) Variables Model 1 Model 2 Model 3 B SE % B SE % B SE % Self-Control .05* .01 5.6 ---.05* .01 5.1 Gang Membership ---.60* .11 83.0 .35* .11 41.4 Constant -.75* .12 -.51* .03 --.68* .12 -123.71* 37.30* 135.04* Pseudo R .01 .00 .01 2.68 2.81 1.67 p .05 Table 13: Negative Binomial Regression Predic ting Frequency of Robbery (N = 3907) Variables Model 1 Model 2 Model 3 B SE % B SE % B SE % Self-Control .08* .01 8.5 ---.06* .01 6.5 Gang Membership ---1.53* .26 363.3 1.15* .26 216.5 Constant -3.59* .29 --1.77* .08 --3.30* .30 -50.32* 46.69* 73.81* Pseudo R .02 .01 .02 17.19 17.37 16.10 p .05 Table 14: Negative Binomial Regression Predicting Frequency of Attacked with a Weapon (N = 3907) Variables Model 1 Model 2 Model 3 B SE % B SE % B SE % Self-Control .11* .01 11.9 ---.09* .01 9.8 Gang Membership ---2.06* .21 686.8 1.77* .20 486.6 Constant -4.10* .28 --1.64 .07 --3.96* .27 -110.57* 128.71* 208.79* Pseudo R .03 .03 .05 12.55 11.99 10.10 p .05
58 Table 15: Negative Binomial Regression Predictin g Frequency of Violent Victimization (N = 3907) Variables Model 1 Model 2 Model 3 B SE % B SE % B SE % Self-Control .06* .01 6.1 ---.05* .01 5.4 Gang Membership ---.79* .10 119.2 .53* .10 69.5 Constant -.72* .11 .64* .03 --.62* .12 -159.13* 71.51* 189.23 Pseudo R .01 .00 .01 2.51 2.62 2.47 p .05 However, in line with the prevalence items, including the control variables substantially altered the effects of self -control and gang membership on frequency of violent victimization. Table 16 presents the results for frequency of hit by someone trying to hurt you, holding all other variables constant. All three models yielded alphas greater than zero and significant model chi-s quares. Thus, negative binomial regression is the appropriate technique to use and the models are able to explain more than chance. Model 1 reveals a coefficient for self-contro l (B = .007) that was not significant (p .05). Accordingly, holding all other va riables constant, self-control did not have a statistically significant effect on the frequency of hit by someone trying to hurt you. Model 2 tested the independent effect of gang membership. The coefficient for gang membership (B = -.326) was statistically significant, but not in the hypothesized di rection. These results indicate that being a gang member decrea sed the expected frequency of being hit by someone by 28 percent, holding all other variables constant. Furthermore, when selfcontrol and gang membership we re included in the same mode l (Model 3), the effect of gang membership was not significantly reduced and therefore, a mediating effect was not supported.
59 All three models in Table 16 reveal significant coefficients for gender, Hispanic, parental attachment, and prior delinquency. Re sults indicate that being a male increased the expected frequency of being hit by some one by 46 percent; being Hispanic decreased the expected frequency by 20 percent. Each increase in parental at tachment decreased the expected frequency by 1 percent and each in crease of self-reported offending increased the expected frequency by approximately 10 percent. Table 16: Negative Binomial Regression Predictin g Frequency of Hit by Someone Trying to Hurt You Including the Control Variables (N = 3907) Variables Model 1 Model 2 Model 3 B SE % B SE % B SE % Age -.02 .05 -2.4 -.02 .05 -1.4 -.02 .05 -1.7 Gender .38* .06 45.6 .38* .06 45.7 .38* .06 45.6 Black -.06 .08 -5.7 -.06 .08 -5.8 -.05 .08 -5.1 Hispanic -.22* .08 -19.5 -.19* .08 -17.7 -.20* .08 -17.8 Indian .28 .19 32.7 .26 .19 30.2 .27 .19 31.3 Asian -.10 .12 -9.4 -.08 .12 -7.7 -.07 .12 -7.2 Other -.11 .12 -10.4 -.10 .12 -9.8 -.10 .12 -9.0 G.R.E.A.T. -.04 .06 -3.7 -.04 .06 -4.3 -.04 .06 -3.9 Parental Attachment -.01* .00 -1.3 -.01* .00 -1.3 -.01* .00 -1.3 Parental Supervision .01 .01 1.2 .01 .01 1.2 .01 .01 1.3 Peer Delinquency -.00 .00 -.2 -.00 .00 -0.1 -.00 .00 -.1 Prior Delinquency .09* .01 9.4 .10* .01 10.6 .10* .01 10.3 Self-Control .01 .01 .7 ---.01 .01 .7 Gang Membership ----.33* .12 -27.9 -.32* .12 -8.4 Constant .64 .70 -.62 .69 -.47 .70 -386.10* 392.15* 393.53* Pseudo R .03 .03 .03 2.34 2.33 2.33 p .05 Analyses predicting frequency of robbe ry are summarized in Table 17. All three models yielded significant chi-squares and alph a statistics greater than zero. Neither selfcontrol (B = -.024) nor gang membership (B= -.451) were significant predictors of
60 frequency of robbery. In fact, self-control and gang membership both yielded a negative coefficient. This finding is in the opposite di rection of what is exp ected and consistently found in the literature (Pratt & Cullen 2000, Peterson et al. 2004). Gender, Black, Hispanic, Indian, and pr ior delinquency remained significant predictors of frequency of robbery. Being a male, black, Indian, or associating with delinquent peers increased the expected fr equency of robbery and being Hispanic decreased the expected frequency. Table 17: Negative Binomial Regression Predic ting Frequency of Robbery Including the Control Variables (N = 3907) Variables Model 1 Model 2 Model 3 B SE % B SE % B SE % Age -.02 .12 -2.2 -.01 .12 -.8 -.01 .12 -.7 Gender .95* .16 158.4 .95* .16 158.2 .96* .16 161.1 Black .42* .18 52.3 .46* .18 59.0 .43* .18 53.7 Hispanic -.33 .22 -28.4 -.35 .22 -29.8 -.35 .22 -29.3 Indian .92* .44 150.7 .94* .44 154.7 .92* .44 151.0 Asian -.08 .32 -8.0 -.05 .32 -4.6 -.06 .32 -5.7 Other .22 .30 24.5 .31 .30 36.1 .26 .30 30.2 G.R.E.A.T. -.15 .15 -14.2 -.13 .15 -1.7 -.15 .15 -13.5 Parental Attachment -.01 .01 -.6 -.00 .01 -.3 -.01 .01 -.5 Parental Supervision -.04 .03 -3.5 -.03 .03 -2.8 -.04 .03 -3.6 Peer Delinquency .01 .01 1.0 .01 .01 1.0 .01 .01 1.3 Prior Delinquency .14* .02 14.6 .14* .02 14.8 .15* .02 16.0 Self-Control -.02 .02 -2.4 ----.03 .02 -2.6 Gang Membership ----.45 .28 -36.3 -.49 .28 -38.4 Constant -1.86 1.72 -----2.18 1.73 -236.22* 235.99* 239.15 Pseudo R .07 .07 .07 10.70* 10.70* 10.66* p .05 Table 18 provides results for frequency of attacked with a weapon. Each model yielded a significant chi-square and alpha pa rameter. Self-control (B = .020) and gang
61 membership (B = .279) failed to obtain statis tical significance. Henc e, self-control and gang membership were not associated with frequency of attacked with a weapon. Race/ethnicity was found to be a signifi cant predictor of frequency of robbery. Being in the Black, Indian, or other cat egory increased the expected frequency of robbery, whereas being in the Asian categ ory decreased the expected frequency. Consistent with the bulk of empirical literature (Lauritsen et al. 1991, Schreck & Fisher 2004), higher levels of prior and peer delinquency were f ound to be associated with increased frequency of robbery. Being a male also increased the e xpected frequency of robbery by 84 percent. Table 18: Negative Binomial Regression Predicting Frequency of Attacked with a Weapon Including the Control Variables (N = 3907) Variables Model 1 Model 2 Model 3 B SE % B SE % B SE % Age -.03 .10 -3.2 -.05 .10 -5.1 -.05 .10 -5.0 Gender .61* .13 84.0 .62* .13 84.9 .61* .13 84.0 Black .62* .16 85.2 .58* .16 79.2 .61* .16 84.7 Hispanic .26 .17 30.1 .24 .17 27.3 .24 .17 26.8 Indian 1.99* .34 629.3 1.90* .33 566.1 1.99* .34 634.6 Asian -.86* .35 -57.6 -.94* .36 -61.1 -.90* .36 -59.4 Other .71* .24 102.8 .64* .24 88.9 .69* .24 98.4 G.R.E.A.T. .10 .13 9.9 .09 .12 9.2 .11 .13 11.1 Parental Attachment -.00 .01 -0.0 -.00 .01 -.30 -.00 .01 -.1 Parental Supervision .01 .02 1.3 .01 .02 .7 .01 .02 1.1 Peer Delinquency .03* .01 3.3 .03* .01 3.4 .03* .01 3.1 Prior Delinquency .14* .02 15.2 .138* .02 14.8 .13* .02 14.2 Self-Control .02 .01 2.0 ---.02 .01 2.2 Gang Membership ---.278 .21 32.0 .31 .21 36.0 Constant -4.59* 1.51 --3.65* 1.49 --4.22* 1.53 -486.64* 485.87* 488.88* Pseudo R .12 .12 .12 6.22 6.21 6.19 p .05
62 Results for the overall frequency of violen t victimization measure are presented in Table 19. Each model yielded an alpha gr eater than zero and significant model chisquares. Consistent with the findings in Tables 16-18, self-control did not yield a statistically significant eff ect (B = .007) at the .05 level. In addition, although gang membership revealed an insignificant associa tion (B = -.197), it is not in the expected direction. Peterson et al. (2004) found that gang members report higher levels of violent victimization; however these data show that being a gang member decreased the expected frequency of violent victimization. Further, neither coefficient was reduced in Model 3. Overall violent victimization was signi ficantly related to gender, Hispanic, parental attachment, and prior delinquency. Holding other variables constant, being a male increased the expected frequency of violent victimizatio n by 51 percent; being Hispanic decreased the expected frequency by approximately 15 percent; lower levels of parental attachment were asso ciated with higher frequency of victimization and; each additional increase on the prio r delinquency scale, increas ed expected frequency of violent victimization by nearly 11 percent. Table 19: Negative Binomial Regression Predic ting Frequency of Violent Victimization Including the Control Variables (N = 3907) Variables Model 1 Model 2 Model 3 B SE % B SE % B SE % Age .00 .04 0.0 .01 .05 .7 .01 .05 .5 Gender .41* .01 50.6 .41* .06 50.9 .41* .06 50.6 Black .05 .07 5.4 .05 .07 5.0 .06 .07 5.8 Hispanic -.18* .08 -16.0 -.16* .08 -14.8 -.16* .08 -15.0 Indian .31 .18 35.9 .29 .18 34.2 .30 .18 35.0 Asian -.10 .11 -9.3 -.09 .11 -8.6 -.08 .11 -8.1 Other -.05 .12 -4.9 -.05 .12 -4.6 -.04 .12 -3.9 G.R.E.A.T. -.06 .05 -5.4 -.06 .05 -5.8 -.06 .05 -5.5 Parental Attachment -.01* .00 -1.2 -.01* .00 -1.2 -.01* .00 -1.1
63 Table 19 (Continued) Variables Model 1 Model 2 Model 3 B SE % B SE % B SE % Parental Supervision .01 .01 .6 .01 .01 .50 .01 .01 .6 Peer Delinquency -.00 .00 -.1 .00 .00 .00 -.00 .00 -.00 Prior Delinquency .10* .01 10.2 .11* .01 11.1 .10* .01 10.8 Self-Control .01 .01 .70 ---.01 .01 .7 Gang Membership ----.20 .11 -17.9 -.19 .11 -17.5 Constant .311 .665 -.36 .66 -.21 .67 -504.20* 505.56* 507.26* Pseudo R .04 .04 .04 2.11 2.11 2.11 p .05 Negative binomial regression analyses of the frequency items yielded conclusions consistent with the logistic regression resu lts presented above. Once the control variables were entered into the models, self-control was not found to be a significant predictor of frequency of violent victimization. Gang membership revealed a significant association to frequency of hit by someone trying to hurt you; however, it was not in the hypothesized direction. The other three frequency items were not significantly associated with gang membership. As a result of these findings, a mediating effect of self-control and gang membership was also not found. Given the null findings in Tables 17-19 for the effects of self-control and gang membership, it was not necessary to test thei r interactive effects. Thus, frequency of hit by someone trying to hurt you was the only item tested because of the significant effect of gang membership. Table 20 displays the resu lts. Both groups yi elded significant model chi-squares and alphas greater than zero. Yet, self-control was not si gnificant for either group. Consequently, it was not ne cessary to calculate a z-score of the coefficients and it
64 can be concluded that self-c ontrol and gang membership did not interact to amplify the frequency of violent victimization. Table 20: Negative Binomial Regression Predicting Fr equency of Hit by Someone Trying to Hurt You by Gang Membership (N = 3907) Variables Gang Members ( N = 311) Nongang Members (N=3596) B SE B SE Z-Score Age -.04 .11 -.02 .05 -Gender .18 .18 .40* .06 -Black .02 .25 -.07 .08 -Hispanic .16 .23 -.25* .09 -Indian -.19 .54 .31 .20 -Asian .51 .37 -.11 .12 -Other -.01 .28 -.10 .13 -G.R.E.A.T. -.13 .17 -.04 .06 -Parental Attachment -.00 .01 -.01* .00 -Parental Supervision .02 .03 .01 .01 -Peer Delinquency -.00 .01 -.00 .00 -Prior Delinquency .09* .03 .10* .01 .63 Self-Control .02 .02 .01 .01 -Constant -.38 1.69 .56 .75 -38.01* 329.80* Pseudo R .03 .03 1.57 2.42 p .05 Unfortunately, the G.R.E.A.T. data do not support the hypotheses in this study. Once the control variables were added into th e analyses, self-control was not found to be associated with violent victimization in any of the models. Out of the eight items analyzed, only prevalence of attacked with a weapon revealed a signi ficant association in the hypothesized direction, once the control variables were introduced. Prevalence and frequency of hit by someone trying to hurt you revealed a significant association with gang membership opposite the hypothesized direc tion, and all of the other items yielded an insignificant effect. Cons equently, neither mediating nor moderating relationships were observed between self-control and gang membership.
65 Chapter 7 Conclusion and Discussion The purpose of this study was to test an integrated model of violent victimization. Victimization has been found to be associated wi th a variety of situat ional factors that are either present or absent in an individuals life (Hindelang et al. 1978, Cohen & Felson 1979, Miethe & Meier 1990). More recently, evid ence of an association between certain individual level factors, such as low self -control and victimizati on, has emerged (Forde and Kennedy 1997, Schreck 1999, Stewart et al. 2004). Based on these findings, the core of this study focused on investigating the e ffects of low self-control and risky lifestyle (gang membership) on prevalence and fr equency of violent victimization. It was predicted that lower levels of se lf-control would lead to higher prevalence and frequency of violent victimization. The bivariate analyses s upported this hypothesis; however, the multivariate models did not. Once the control variables were added into the analyses, neither logistic regression nor negative binom ial regression revealed a significant effect of self-contro l on violent victimization. It was also assumed that gang membership would have an effect on both the prevalence and frequency of violent victimization. The prevalence of attacked w ith a weapon was the only victimization item to provide support for this hypothesis Conversely, nongang members were found to report greater prevalence and frequency of being hit by someone trying to hurt you.
66 This finding may be the result of the tendenc y of some youths to join gangs for protection (Molidor 1996, Mark, 1997, Reibol dt 2001, Walker & Mason 2001). A self-control and gang membership medi ating effect was also predicted. Once the control variables were added into the analyses, the G.R.E.A.T. data failed to support this relationship. When both variables were added into the multivariate model, neither self-control nor gang membership was signi ficantly reduced. Last, it was hypothesized that gang membership would condition th e effects of self-control on violent victimization. To assess this possibility, the data were pa rsed into gang and nongang subsamples and the effect of self-control on vi olent victimization was assessed for each group. In each of these models self-control failed to atta in statistical significance. Therefore, it can be concluded that self-cont rol and gang membership did not interact to increase the risk of violent victimization. These findings do not support the previous wo rk that this study was intended to replicate. Not only did Forde and Kennedy ( 1997) find an association between selfcontrol and victimization, they also concluded that integr ating self-control and risky lifestyle provides a stronger explanatory m odel of victimization. Furthermore, Schreck (1999) and Stewart et al. (2004) also concl uded that low self-cont rol is a significant predictor of victimization. The present study, however, failed to find a significant relationship between self-contro l and violent victimization when other correlates of victimization were controlled. As a result, these findings provide evidence against Schrecks (1999) self-control-v ictimization theory and warrant further investigation into this relationship.
67 The insignificant findings for self-control also provide evidence against Gottfredson and Hirschis (1990) general theory of crime. Gottfredson and Hirschi (1990) maintain that self-control is a versatile characteristic that accounts for all individual differences in behavior. The results of Schr eck (1999) and Stewart et al. (2004) supported the versatility of self-control by providing ev idence that low self-control is able to explain criminal victimization, in additi on to criminal and analogous behaviors. However, the current study failed to find an association between self -control and violent victimization and in turn, provides eviden ce against the notion of self-control as a versatile predictor of behavior. The results of this study also provide weak support fo r the risky lif estyle/exposure theory (Hindelang et al. 1978). Only one item, the prevalence of being attacked with a weapon, revealed an association with gang me mbership that is consistent with these theories. All other items were either insignificant or significant in the opposite direction. Although gang membership is not a ve ry comprehensive measure of risky lifestyle, these results are surprising. Using th e longitudinal G.R.E.A.T. data, Peterson et al. (2004) found that gang membership wa s significantly associ ated with violent victimization. These contradictory findings may be the result of the cross-sectional nature of the data used in this study. Peterson and associates (2004) were able to assess victimization over a five-year period and comp are time of gang membership to time of victimization. The current study, on the other hand, was only able to look at the year prior to the survey and, therefore, unable to dete rmine if gang membership leads to future victimization.
68 Further, gender and prior delinquency we re the only two control variables that remained significant predictors of victimi zation throughout all of the models. Not only were males more likely to report one or more victimization experiences, being a male increased the expected frequency of vict imization by 46 to 161 percent. This strong association between gender and victimization supports the bulk of empirical literature that concludes that males are victimized at a substantially higher rate than females (Lauritsen et al. 1991, Lauritsen 2001, Schreck & Fisher 2004). A possible explanation for this consistent finding is that males tend to lead riskier lifestyles. According to the principle of homogamy (Hindelang et al. 1978, Lauritsen et al. 1991) victimization is more likely to occur when indivi duals associate with members of demographic groups that contain a disproportionate share of offenders. Based on this principle, Hindelang and associat es (1978) argue that adolesce nt males are more likely to become victims of crime because they are more likely to associate with other adolescent males who are involved in delinquent activities themselves. Previous research has supported this assumption by showing that ma les are more likely to associate with delinquent peers (Simons et al. 1980, Mears et al. 1998) and delinquent peer association is related to criminal victimization (Laurits en et al. 1992, Schreck et al. 2002, Schreck et al. 2004). At the same time, prior criminal offending is one of the strongest risk factors of victimization (Esbensen and Huizinga 1991, Sampson & Lauritsen 1990, Lauritsen et al. 1992, Lauritsen & Quinet 1995). The findings of this study also contribute to this offender-victim link. For each additional incr ease in prior delinquency, the expected frequency for violent victimizati on increased 10 to 16 percent.
69 Several implications can be drawn fr om these results. First, Hirschi and Gottfredson (1993) argue that us ing a behavior scale of low self-control provides a better measure of self-control than cognitive, self-report measures because low self-control can affect survey response. For example, impul sive people may not think through the answers clearly, lazy individuals may feel disinclined to partic ipate, and individuals who do not have the capacity to understand the survey questions will become easily frustrated. Accordingly, Hirschi and Gottfredson believe that an accurate measurement of selfcontrol involves objectiv e observation of the acti ons that are used to measure self-control, by someone other than the individual whose self -control is being meas ured (Tittle et al. 2003). To add to this, Marcus (2004:42) explains that relying on att itudinal measures of self-control provides several drawbacks. Attitudinal questio ns force the respondent to remember, evaluate, and summarize target attitudes. The more cognitive operations these items require, the greater the likelihood of a distorted response. Furthermore, these cognitive questions rely on an individuals capacity fo r self-assessment, including feelings regarding past behavi or. These are qualities one woul d not expect to find in low self-control individuals. We would expect an attitudinal meas ure or self-reflective scale for self-control to yield more reliable-and thus valid-scores at the high pole than at the low pole of the trait it measures (Marcus 2004: 42). This is an una ttractive quality of attitudinal measures because th e low self-control individuals ar e, most often, the interest of the research question. Based on these arguments, the prior delinque ncy measure used in this study could be considered an alternative latent indicat or of self-control. Mo reover, according to
70 Hirschi and Gottfredson (1993; 1995) and Marcus (2004), the prior delinquency scale may also be a more comprehensive, and thus stronger, indicator of self-control. This may be especially true because th e self-control measure was comp rised of only two of the six components that Gottfredson and Hirschi (1990) use to define self-c ontrol and therefore, was a relatively weak indicator of the concept of low self-control. If true, this argument would explain the nonsignificant effect of se lf-control once the cont rol variables (which included prior delinquency) were added into the analyses. On the other hand, prior delinquency could also be a measure of risky lifestyle. The extant literature provides strong evidence that criminal offenders participate in risky behaviors such as more time spent away from home, more time spent away from home past 10 pm, less time spent with parents, substance use, and more time spent with delinquent peers, at a disproportionately higher rate than nonoffenders (Riley 1987, Esbensen & Huizinga 1993). These same behavior s have also been associated with higher levels of victimization (Hindelang et al. 1978, Riley 1987, Kennedy and Forde 1990b) and are characteristic of a risky lifestyle. It is also possible that self-control had an indirect effect on victimization, through its effect on prior delinquency. The literature shows that low self-control is associated with delinquent behavior (Pratt & Cullen 2000 ) and that prior delinquency is related to victimization (Esbensen & Huizinga 1991, Laurit sen et al. 1991). Therefore, it is likely that low self-control l eads to prior delinquency (a risky li festyle), which in turn, leads to violent victimization. In other words, it is likely that prior delinquency mediates the effect of self-control on violen t victimization. Unfortunately, identifying the causal nature
71 of this relationship is beyond the scope of this study, due to the crosssectional nature of the data. Finally, a few shortcomings of this study n eed to be addressed. First, the measures used were a major limitation to this study. Th e G.R.E.A.T. survey was not designed to measure self-control as explicated by Gottfredson and Hirschi (1990). Consequently, only two of the six dimensions of self-control (risk-seeking and impulsivity) were measured in the survey. Although risk-seeking and impulsivity have been shown to carry the most explanatory power (Nakhaie et al. 19 99, Arneklev et al. 1993, Wood et al. 1993, LaGrange & Silverman 1999, Winfree & Bernat 1998, Arneklev et al. 1999), Gottfredson and Hirschi (1990) define self -control as a unidimensional construct comprised of six components. Including only two of the six components means measuring less than half of the concept of self-control, wh ich is a serious problem of c ontent validity. Accordingly, the self-control scale used in this study is a very weak indicator of low self-control. Moreover, Forde and Kennedy (1997), Sc hreck (1999), and Stew art et al. (2004) used self-control measures that included all six components of self-control. They all found a significant association between low self-control and victimization. Thus, the other four components of self-control must c ontribute to the effect that low self-control has on victimization. For that reason, it could also be argued that th e self-control measure in this study was not really measuring self-c ontrol at all. Instead, it may actually be an impulsive/risk-seeking scale, since it was co mprised of only these two characteristics. Relying on gang membership as a measur e of risky lifestyle is also a limitation. The decision to use gang membership as a proxy for risky lifestyle was based on the bulk of empirical literature that shows a correlation between gang membership and several of
72 the same behaviors (age, offending, substance use, drug sales, gambling, more time spent with delinquent friends, and gun ownership [B attin et al. 1998, Bjerregaard & Lizotte 1995, Curry 2000, Curry & Spergel 1992, Esbensen & Huizinga 1993, Esbensen & Winfree 1998, Hill et al. 1999, Hope & Da mphousse 2002]) that are used to conceptualize risky lifestyle. However, gang membership does not measure the full range of the risky lifestyle concept. Further, us ing only one self-definitional question (Are you in a gang now?) does not provide insight into the behaviors or circumstances that are characteristic of each gang. Classifying gang membership by a set of cr iteria that includes risky behavior (offending, initiation rites) ma y have provided a more accurate measure of risky lifestyle, and in turn, a stronger association with violent vict imization. Also, neglecting to includ e property victimization in the analyses limited the measurement of victimization. Based on the theoretical linkage of self-control and victimization provided by Schreck (1999), low self-control has the pot ential to decrease guardianship. According to the routine activities perspective, this leads to unprotected belongings, which make attractive targets to motivated offenders. Taken together, it seems possible that low self-control may influe nce an individuals te ndency to guard their possessions, which would increase the risk of property victimization. The use of cross-sectional data limits the results of this study in two ways. First, because data were collected at one point in time, information is only available for the year prior to the survey. Victimization is a relatively rare experi ence, and therefore, collecting information once, over a one-year period severely limits the chances for victimization to be reported. Collecting in formation multiple times, over a longer time span, would increase the chances of a re spondent reporting victimization and provide
73 more information on the dependent variable. Additionally, causal relationships cannot be determined with cross-sectional data. Theref ore, it is impossible to establish whether gang membership leads to future victimiza tion, or if prior delinquency mediated the effects of self-control and/or risky lifestyle on violent vict imization. Last, the use of secondary data is also a limitation to this study, not only in regard to the measure of self-control, but also in the ability to generalize the results. Even though the G.R.E.A.T. data we re collected on a large sample of youth in different geographical areas, it was not a truly random sample. In addition, the sample consisted of middle school students, which is a relative ly young population to collect information on criminal behavior and victimization. Conse quently, these results cannot be generalized beyond the sample of interest. Future research into the risk factor s of victimization should not ignore the possible relationship between self-control, ri sky lifestyle, and vi ctimization. However, more comprehensive measures of the variable s of interest are needed. For example, had the self-control measure include d all six components, it is ve ry possible that a different picture would have emerged. At the same time an accurate measure of self-control would provide stronger insight into the true relationship between prior delinquency, self-control, and victimization. Furthermore, a measure of risky lifestyle that is comprised of the characteristics described by Hindelang et al. (1978), such as staying out late, frequenting dangerous areas, less time spent with family, and a ssociating with criminal offenders, would increase the validity of the risky lifestyle measure, in turn, pote ntially increasing the chances of finding a significant association be tween risky lifestyle and victimization that
74 is in line with the bulk of empirical literature (Hindelang et al. 1978, Forde and Kennedy 1990b, Miethe & Meier 1990). Victimization is a traumatizing experience that can leave lasting effects. Therefore, the need to recognize which factors lead to this type of experience is critical. Until recently, only situational f actors have been used to explain victimization. However, considerations of the individual-level infl uences on victimization are beginning to emerge. Further research into both the indi vidual and situationa l risk factors of victimization is needed to identify the fu ll range of direct and indirect causes of victimization, so that appr opriate action can be taken to minimize these risks.
REFERENCES Akers, R. & Sellers, C. (2004). Criminological Theories: In troduction, Evaluations, and Application. Los Angeles, CA: Roxbury Publishing Company. Anderson, A. (2002). Individual and contextual influences on delinquency: The role of the single-parent family. Journal of Criminal Justice, 30, (6), p. 575-87. Arneklev, B., Cochran, J. & Gainey, R. ( 1998). Testing Gottfredson and Hirschis low self-control stability hypothe sis: An exploratory study. American Journal of Criminal Justice, 23, (1), p.107-27. Arneklev, B., Grasmick, H. & Bursik, R. (1999). Evaluating the dimensionality and invariance of low self-control. Journal of Quantitative Criminology, 15, (3), p. 307-31. Arneklev, B., Grasmick, H., Tittle, C. & Bursik, R. (1993). Low self-control and imprudent behavior. Journal of Quantitative Criminology, 9, (3), p. 225-47. Baron, S. (2003). Self-control, social cons equences, and criminal behavior: Street youth and the general theory of crime. Journal of Research in Crime and Delinquency, 40, (4), p. 403-25. Battin, S., Hill, K., Abbott, R., Catalano, R. & Hawkins, J. (1998). The contribution of gang membership to delinquenc y beyond delinquent friends. Criminology, 36, (1), p. 93-115. Bjarnason, T., Sigurdardottir, T. & Thorli ndsson, T. (1999). Human agency, capable guardians, and structural constraints: A lifestyle appro ach to the study of violent victimization. Journal of Youth and Adolescence, 28, (1), p. 105. Bjerregaard, B. & Lizotte, A. (1995) Gun ownership and gang membership. Journal of Criminal Law and Criminology, 86, (1), p. 37-58. Brame, R., Paternoster, R., Mazerolle, P. & Piquero, A. (1998). Testing for the equality of maximum-likelihood coefficients between two independent equations. Journal of Quantitative Criminology, 14, (3), p. 245-61. 75
Brannigan, W., Pevalin, D. & Wade, T. (2002) Self-control and social control in childhood misconduct and aggression: The role of family structure, hyperactivity, and hostile parenting. Canadian Journal of Criminology, 44, (2), p. 119-43. Brownfield, D. & Sorenson, A. (1993). S elf-control and juve nile delinquency: Theoretical issues and an empirical assessm ent of selected elements of a general theory of crime. Deviant Behavior, 14, (3), p. 243-64. Burton, V., Cullen, F., Evans, T., Alarid, L. & Dunaway, R. (1998). Gender, selfcontrol, and crime. Journal of Research in Crime and Delinquency, 30, (2), p. 123-47. Burton, V., Evans, T., Cullen, F., Olivares, K. & Dunaway, R. (1999). Age, selfcontrol, and adults offending: A resear ch note assessing a ge neral theory of crime. Journal of Criminal Justice, 27, (1), p. 45-5. Cattell. R. (1966). The scree te st for the number of factors. Multivariate Behavioral Research, 1, p. 245-76. Cauffman, E., Steinberg, L. & Piquero, A. (2005). Psychological neuropsychological and physiological correlates of serious antisocial behavior in adolescence: The role of self-control. Criminology, 43, (1), p. 133-74. Chapple, C. & Hope, T. (2003). An analys is of the self-control and criminal versatility of gang and dating violence offenders. Violence and Victims, 18, (6), p. 671-90. Cochran, J., Wood, P., Sellers, C. & Cham lin, M. (1998). Academic dishonesty and low self-control: An empirical test of a general theory of crime. Deviant Behavior: An Interdisciplinary Journal, 19, p. 227-55. Cochran, J., Bromley, M., & Branch, K. (2000). Victimization and fear of crime in an entertainment district crime hot spot: A test of structural-choice theory. American Journal of Criminal Justice, 24, (2), p. 189-201. Cohen, L. & Felson, M. (1979). Social cha nge and rate trends : A routine activity Approach. American Sociological Review, 44, (4), p. 588-608. Cohen, L., Cantor, D. & Kluegel, J. (1981b) Robbery victimization in the U.S.: An analysis of a nonrandom event. Social Science Quarterly, 62, (4), p. 644-57. Cohen, L., Kluegel, J. & Land, K. (1981) Social inequality and predatory victimization: An explosion a nd test of a formal theory. American Sociological Review-Albany New York, 46, (5), p. 505-24. 76
Curry, D. (2002). Self-reported gang i nvolvement and officially recorded delinquency. Criminology, 38, (1), p. 1253-74. Curry, D. & Spergel, I. (1992). Gang invol vement and delinquency among Hispanic and African-American adolescent males. Journal of Research in Crime and Delinquency, 29, (3), p. 273-91. Curry, D., Decker, S. & Egley, A. (2002). Gang involvement and delinquency in a middle school population. Justice Quarterly, 19, (2), p. 275-92. Dahlberg, L. (1998). Youth violence in the United States: Major trends, risk factors, and prevention approaches. American Journal of Preventative Medicine, 14, (4), p. 259-72. Delisi, M. (2001). Designed to fail: Self-control and i nvolvement in the criminal justice system. American Journal of Criminal Justice, 26, (1), p.131-48. Deng, X. & Zhang, L. (1998). Correlates of self-control: An empirical test of selfcontrol theory. Journal of Crime and Justice, 21, (2), p. 89-110. Deschenes, E. & Esbensen, F. (1999). V iolence and gangs: Gender differences in perceptions and behavior. Journal of Quantitative Criminology, 15, (1), p. 63-85. Esbensen, F. & Deschenes, E. (1998). A multisite examination of youth gang membership: Does gender matter? Criminology, 36, (4), p. 799-827. Esbensen, F., Deschenes, E. & Winfree, L. (1999). Differences between gang girls and gang boys. Youth and Society, 31, (1), p. 27-53. Esbensen, F. & Huizinga, D. (1991). Juve nile victimization and delinquency. Youth and Society, 23, (2), p. 202-28. Esbensen, F. & Huizinga, D. (1993). Gangs, drugs, and delinquency in a survey of urban youth. Criminology, 31, (4), p. 565-89. Esbensen, F. & Osgood, D. (1999). Gang resi stance education and training (GREAT): Results from the national evaluation. Journal of Research in Crime and Delinquency, 36, (2), p. 194-225. Esbensen, F., Osgood, D., Taylor, T., Peterson, D. & Freng, A. (2001b). How great is G.R.E.A.T.? Results from a longitu dinal quasi-experimental design. Criminology & Public Policy, 1, (1), p. 87-118. Esbensen, F. & Winfree, L. (1998). Race and gender differences between gang and nongang youths: Results from a multisite survey. Justice Quarterly, 15, (3), p. 505-26. 77
Esbensen, F., Winfree, L. & He, N. (2001a). Youth gangs and definitional issues: When is a gang, and why does it matter? Crime and Delinquency, 47, (1), p. 10530. Forde, D. & Kennedy, L. (1990a). Routine activities and crime: An analysis of victimization in Canada. Criminology, 28, (1), p. 137-52. Forde, D. & Kennedy, L. (1990b). Risky lifes tyles and dangerous results: Routine activities and exposure to crime. Social Science Review, 74, (4), p. 208-11 Forde, D. & Kennedy, L. (1997). Risky lifestyles, routine activitie s, and the general theory of crime. Justice Quarterly, 14, (2), p.266-88. Gardner, W., Mulvey, E. & Shaw, E. (1995). R egression analyses of counts and rates: Poisson, overdispersed poisson, a nd negative binomial models. Psychological Bulletin, 118, (3). P. 392-404. Garofalo, J., Siegel, L. & Laub, J. ( 1987). School-related victimizations among adolescents: An analysis of Nati onal Crime Survey (NCS) narratives. Journal of Quantitative Criminology, 3, (4), p. 321-38. Gibbs, J. & Giever, D. (1995). Self-Control and its manifestations among university students: An empirical test of Gottfreds on and Hirschis general theory of crime. Justice Quarterly, 12, (2), p. 231-55 Gibbs, J., Giever, D. & Martin, J. (1998). P arental management and self-control: An empirical test of Gottfredson a nd Hirschis general theory. Journal of Research in Crime and Delinquency, 35, (1), p. 40-70. Gottfredson, M. & Hirschi, T. (1990). A General Theory of Crime. Palo Alto, CA: Stanford University Press. Grasmick, H., Tittle, C., Bursik, R. & Arneklev, B. (1993). Testing the core implications of Gottfredson and Hirs chis general theory of crime. Journal of Research in Crime and Delinquency, 30, (1), p. 5-29. Haight, F. (1967). Handbook of the Poisson Distribution. Los Angeles, CA: John Wiley & Sons. Hay, Carter. (2001). Parenting, self-control, and delinquency: A test of self-control theory. Criminology, 39, (3), p. 707-36. Higgins, G. & Makin, D. (2004). Does social learning theory conditi on the effects of low self-control on college st udents software piracy? Journal of Economic Crime Management, 2, (2), p. 1-22. 78
Hill, K., Howell, J., Hawkins, J. & Battin-Pear son, S. (1999). Childhood risk factors for adolescent gang membership: Results from the Seattle Social Development Project. Journal of Research in Crime and Delinquency, 36, (3), p. 300-22. Hindelang, M., Hirschi, T., & Weis, J. (1978). Victims of Personal Crime Cambridge, MA: Ballinger. Hirschi, T. & Gottfredson, M. (1993). Co mmentary: Testing the general theory of crime. Journal of Research in Crime and Delinquency, 30, (1), p.47-54. Hirschi, T. & Gottfredson, M. (1995). Control Theory and th e Life-Course Perspective. Studies on Crime and Crime Prevention, 4, p. 131-42. Horney, J., Osgood, W. & Haen, M. (1995). Cri minal careers in the short-term: IntraIndividual variability in crime and it s relation to local life circumstances. American Sociological Review, 60, (5), p. 6573. Hope, T. & Damphousse, K. (2002). Applying self-control theory to gang membership in a nonurban setting. Journal of Gang Research, 9, (2), p. 41-61. Hough, M. (1987). Offenders choice of targ et: Findings from victim surveys. Journal of Quantitative Criminology, 3, (4), p. 355-70. Hoyt, D., Ryan, K. & Cauce, A. (1999). Personal victimization in a high-risk environment: Homeless and runaway adolescents. Journal of Research in Crime and Delinquency, 36, p. 371-92. Junger, M., Patterson, G. & Pulkkinen, L. (1995) Accidents, self-cont rol, and crime. Criminal Behavior and Mental Health, 5, (4), p. 386-478. Junger, M. & Tremblay, R. (1999). S elf-control, accidents, and crime. Criminal Justice and Behavior, 26, (4), p. 485-501. Kalton, G. (1983). Introduction to Survey Sampling. Sage University Paper series on Quantitative Applications in the Social Sciences, series no. 07-001. Beverly Hills, CA: Sage. Keane, C., Maxim, P. & Teevan, J. (1993) Drinking and driving, self-control, and gender: Testing a genera l theory of crime. Journal of Research in Crime and Delinquency, 30, (1), p. 30-46. Keane, C. & Arnold, R. (1996). Examini ng the relationship between criminal victimization and accidents: A routine activities approach. Canadian Review of Sociology and Anthropology, 33, (4), p. 457-79. 79
Kennedy, L. & Baron, S. (1993). Routine activi ties and a subculture of violence: A study of violence on the street. Journal of Research in Crime and Delinquency, 30, (1), p. 88-112. Kennedy, L. & Forde, D. (1990). Routine activities and crime: An analysis of victimization in Canada. Criminology, 28, (1), p. 137-52. Klein, M. (1995). The American Street Gang: Its Nature, Prevalence, and Control. New York: Oxford University Press. Kleinbaum, D., Kupper, L. & Muller, K. (1988). Applied Regression Analysis and Other Multivariate Methods. Boston, MA: PWS-Kent. Krauss, D., Sales, B. & Becker, J. (2000). Beyond prediction to e xplanation in risk assessment research: A comparison of two explanatory theories of criminality and recidivism. International Journal of Law and Psychiatry, 23, (2), p. 91-112. LaGrange, T. & Silverman, R. (1999). Low self-control and opportunity: Testing the general theory of crime as an explanati on for gender differences in delinquency. Criminology, 37, (1), p. 41-72. Lasley, J. (1989). Routine/lifest yles and predatory victimizat ion: A causal analysis. Justice Quarterly, 6, (4), p. 529-42. Lasley, J. & Rosenbaum, J. (1988) Routin e activities and multiple personal victimization. Sociology and Social Research, 73, (1), p. 47-50. Lauritsen, J. (2001). The social ecology of violent victimization: Individual and contextual effect of the NCVS. Journal of Quantitative Criminology, 17, (1), p. 3-32. Lauritsen, J., Laub, J. & Sampson, R. (1992). Conventional and de linquent activities: Implications for the prevention of vi olent victimization among adolescents. Violence and Victims, 7, (2), p. 91-105. Lauritsen, J. & Quinet, K. (1995). Repeat victimization among adolescents and young adults. Journal of Quantitative Criminology, 11, (2), p. 143-66. Lauritsen, J., Sampson, R. & Laub, J. (1991). The link between offending and victimization among adolescents. Criminology, 29, (2), p. 265-92. Lee, M. (2000). Community cohesion a nd violent predatory victimization: A theoretical extension and cross-natio nal test of opportunity theory. Social Forces, 79, (2), p. 683-706. 80
Long, J. & Freese, S. (2003). Regression Models for Categorical Dependent Variables Using Stata. College Stations, TX: Stata Press. Longshore, D. (1998). Self-cont rol and criminal oppor tunity: A prospectiv e test of the general theory of crime. Social Problems, 45, (1), p. 102-13. Longshore, D., Stein, J. & Turner, S. (1996). Self-control in a criminal sample: An examination of construct validity. Criminology, 34, (2), p. 209-27. Longshore, D., Stein, J. & Turner, S. (1998). Reliability and valid ity of a self-control measure: Rejoinder. Criminology, 36, (1), p. 175-82. Longshore, D. & Turner, S. (1998). Sel f-control and criminal opportunity. Criminal Justice and Behavior, 25, (1), p. 81-98. Lynskey, D., Winfree, T. & Esbensen, F. (2000). Linking gender, minority group status, and family matters to self-control theory: A multivariate analysis of key self-control concepts in youth-gang context. Juvenile and Family Court Journal 51, (3), p. 1-19. Marcus, B. (2004). Self-control in the general theory of crime. Theoretical Criminology, 8, (1), p. 33-55. Mark, G. (1997). Gangs, violence, and drugs. Free Inquiry in Creative Sociology, 25, (1), p. 41-50. Mears, D., Ploeger, M.& Warr, M. (1998). Explaining the gender gap in delinquency: Peer influence and moral evaluations of behavior. Journal of Research in Crime and Delinquency, 35, 3, p. 251-67. Messner, S. & Blau, J. (1987). Routine leisur e activities and rates of crime: A macrolevel analysis. Social Forces, 65, (4), p. 1035-52. Miethe, T. & Meier, R. (1990). Opportunity, choice, and criminal victimization: A test of a theoretical model. Journal of Research in Crime and Delinquency, 27, (3), p. 243-66. Miethe, T., Stafford, M. & Long, J. (1987) Social differentiation in criminal victimization: A test of routine activities/lifestyle theories. American Sociological Review, 52, p. 84-94. Molidor, C. (1996). Female gang members: A pr ofile of aggression an d victimization. Social Work, 41, (3), p. 251-57. 81
Mustaine, E. & Tewksbury, R. (1998). Predict ing risks of larceny theft victimization: A routine activity analysis usi ng refined lifestyle measures. Criminology, 36, (4), 829-58. Nagin, D. & Paternoster, R. (1993). Endur ing individual differences and rational choice theories of crime. Law and Society Review, 27, (3), p. 467-96. Nakhaie, M., Silverman, R. & LaGrange, T. ( 1999). Self-control and social control: An examination of gender, ethnicity, class, and delinquency. Canadian Journal of Sociology, 25, (1), p. 35-59. Nelson, C. & Huff, C. (1998). Strangers in the night: An applica tion of the lifestyleroutine activities approach to el derly homicide victimization. Homicide Studies, 2, (2), p. 130-59. Paternoster, R. & Brame, R. (1998). The structural similar ity of processes generating criminal and analogous behaviors. Criminology, 36, (3), p. 633-69. Paternoster, R. & Brame, R. (2000). On the association among self-control, crime, and analogous behavior. Criminology, 38, (3), p. 971-82. Perrone, D., Sullivan, C., Pratt, T. & Marg ary, S. (2004). Parental efficacy, selfcontrol, and delinquency: A test of a general theory of crime on a nationally representative sample of youth. International Journal of Offender Therapy and Comparative Criminology, 8, (3), p. 298-312. Peterson, D., Taylor, T. & Esbensen, F. (2004). Gang membership and violent victimization. Justice Quarterly, 21, (4), p.793-815. Piquero, A., Macdonald, J., Dobr in, A., Daigle, L. & Cullen, F. (2005). Self-control, violent offending, and homicide victimiza tion: Assessing the ge neral theory of crime. Journal of Quantitative Criminology, 21, (1), p. 55-71. Piquero, A. & Rosay, A. (1998). The reliabilit y and validity of Gras mick et al.s selfcontrol scale: A comment of Longshore et al. Criminology, 36, (1), p. 157-73. Piquero, A. & Tibbets, S. (1996). Specifying the direct and indir ect effects of low self-control and situa tional factors in offenders decision making. Justice Quarterly, 13, (3), p. 481-510. Polakowski, M. (1994). Linking self-and soci al control with devi ance: Illuminating the structure underlying a general theory of crime and its relation to deviant activities. Journal of Quantitative Criminology, 10, (1), p. 41-78. Pratt, T., & Cullen, F. (2000). The empiri cal status of Gottfredson and Hirschis general theory of crime: A meta-analysis. Criminology, 38, (3), p. 931-64. 82
Pratt. T., Turner, M. & Piquero, A. (2004) Parental so cialization and community context: A longitudinal analysis of the structural sources of low self-control. Journal of Research in Crime and Delinquency, 41, (3), p. 219-43. Rapp-Paglicci, L. & Wodarski, J. (2000). Antecedent behaviors of male youth victimization: An exploratory study. Deviant Behavior: An Interdisciplinary Journal, 21, p. 519-36. Reiboldt, W. (2001). Adolescent interacti ons with gangs, family, and neighborhoods. Journal of Family Issues, 22, (2), p. 211-42. Riley, D. (1987). Time and crime: The link between teenager lifestyle and delinquency. Journal of Quantitative Criminology, 3, (4), p. 339-54. Robinson, M. (1999). Lifestyle, routine activities, and residential burglary victimization. Journal of Crime and Justice, 22, (1), p. 27-56. Sampson, R. & Wooldredge, J. (1987). Linking the microand macro-level dimensions of lifestyle-routine activity and opportunity models of predatory victimization. Journal of Quantitative Criminology, 3, (4), p. 371-94. Sampson, R. & Lauritsen, J. (1990). Deviant lifestyles, proximity to crime, and the offender-victim link in personal violence. Journal of Research in Crime and Delinquency, 27, (2), p. 110-39. Schreck, C. (1999). Criminal victimization a nd low self-control: An extension and test of a general theory of crime. Justice Quarterly, 16, (3), p. 633-54. Schreck, C. & Fisher, B. (2004). Specifyi ng the influence of family and peers on violent victimization. Journal of Interpersonal Violence, 19, (9), p. 1021-41. Schreck, C., Fisher, B. & Miller, J. (2004). The social contex t of violent victimization: A study of the delinquent peer effect. Justice Quarterly, 21, 1, p. 23-48. Schreck, C., Wright, R. & Miller, J. (2002). A study of individual and situational antecedents of violent victimization. Justice Quarterly, 19, (1), p. 159-78. Sellers, C. (1999). Self-control and intimate violence: An examination of the scope and specification of the general theory of crime. Criminology, 37, (2), p. 375404. Sherman, L., Gartin, P. & Buer ger, M. (1989). Hot spots of predatory crimes: Routine activities and the criminology of place. Criminology, 27, (1), p. 27-56. 83
Simons, D., Miller, M. & Aigne r, S. (1980). Contemporary theories of deviance and female delinquency: An empirical test. Journal of Research in Crime and Delinquency, 17, p. 42-53. Smith, T. (2004). Low self-control, staged opportunity, and subsequent fraudulent behavior. Criminal Justice and Behavior, 31, (5), p. 542-63. Sorenson, A. & Brownfield, D. (1995). Adoles cent drug use and a general theory of crime: An analysis of a theoretical integration. Canadian Journal of Criminology, 37, (1), p. 19-37. Stahura, J. & Sloan, J. (1988). Urban stra tification of places, r outine activities and suburban crime rates. Social Forces, 66, (4), p. 1102-18. Stewart, E., Elifson, K. & Ster k, C. (2004). Integrating the gene ral theory of crime into an explanation of violent victim ization among female offenders. Justice Quarterly, 21, (1), p. 159-81. Studenmund, A. (2001). Using Econometrics: A Practical Guide (4 th ed.). U.S.: Addison Wesley Longman. Stylianou, S. (2002). The relationship between elements and manifestations of low self-control in a genera l theory of crime: Two comments and a test. Deviant Behavior, 23, p. 531-57. Taylor, T., Turner, K., Esbensen, F. & Wi nfree, L. (2001). Coppin an attitude: Attitudinal differences among juveniles toward police. Journal of Criminal Justice, 29, (4), p. 295-305. Thompson, C. & Fisher, B. (1996). Predic ting household victimization utilizing a multi-level routine activity approach. Journal of Crime and Justice, 19, (2), p. 49-66. Tittle, C., Ward, D. & Grasmick, H. (2003). Self-control and crime/deviance: Cognitive vs. behavioral measures. Journal of Quantitative Criminology, 19, (4), p. 333-65. Turner, M. & Piquero, A. (2002). T he stability of self-control. Journal of Criminal Justice, 30, p. 457-71 Unnever, J., Cullen, F. & Pratt, T. (2 003). Parental management, ADHD, and delinquent involvement: Reassessing Gottfre dson and Hirschis general theory. Justice Quarterly, 20, (3), p. 471-500. Vasonyi, A. & Crosswhite, J. (2004). A test of Gottfredson and Hirschis general theory of crime in African adolescents. Journal of Research in Crime and Delinquency, 41, (4), p. 407-32. 84
Vasonyi, A., Pickering, L., Junger, M. & He ssing, D. (2001). An empirical test of a general theory of crime: A four-nation comparative st udy of self-control and the prediction of deviance. Journal of Crime and Delinquency, 38, (2), p. 91-131. Walker, C. & Mason, C. (2001). Perceptions of risk factors for female gang involvement among African American and Hispanic women. Youth and Society, 32, (3), p. 303-36. Winfree, L. & Bernet, F. (1998). Social lear ning, self-control, and substance abuse by eighth grade students: A tale of two cites. Journal of Drug Issues, 28, (2), p. 53958. Winfree, L., Esbensen, F. & Osgood, D. (1996). Evaluating a school-based gangprevention program. Evaluation Review, 20, (2), p. 181-203. Winfree, L., Fuller, K., Vigil, T. & Mays, L. (1992). The definition and measurement of gang status: Policy implications for juvenile justice. Juvenile and Family Court Journal, p. 29-37. Wright, B., Caspi, A., Moffitt, T. & Silva, P. (1999). Low self-control, social bonds, and crime: Social causation, so cial selection, or both? Criminology, 37, (3), p. 479514. Wood, P., Pfefferbaum, B. & Arneklev, B. (1993). Risk-taking and self-control: Social psychological correlates of delinquency. Journal of Crime and Justice, 16, (1), p. 111-30. Wooldredge, J., Cullen, F. & Latessa, E. (1992) Victimization in the workplace: A test of routine activities theory. Justice Quarterly, 9, (2), p. 325-35. Zhang, L., Weite, J. & Wieczorek, W. ( 2001). Deviant lifestyle and crime victimization. Journal of Criminal Justice, 29, (2), p. 133-43. 85