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 001788085
007 cr mnu|||uuuuu
008 070516s2006 flu sbm 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0001436
Gulledge, Laura M.
Gender differences in age of onset for delinquency :
b risk factors and consequences
h [electronic resource] /
by Laura M. Gulledge.
[Tampa, Fla] :
University of South Florida,
Thesis (M.A.)--University of South Florida, 2006.
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 75 pages.
ABSTRACT: The age of onset of delinquency has long been viewed as a primary indicator for further delinquency and criminality. However, studies on the risk factors for onset, and future delinquency have focused predominantly on males. The purpose of this study was to explore gender differences and similarities in risk factors for onset and frequency of arrest. The data used in these analyses were from a longitudinal study, Pathways to Adulthood: A Three Generational Urban Study, 1960-1994. Sixty-six percent (N=1,758) of the eligible children completed the final survey. Of these children, only 515 were used in this particular study because they had documented ages of first arrest. It is hypothesized that 1) female "early" onset occurs at a later age from that of male "early" onset, 2) risk factors predictive of early onset will differ across gender, and 3) "early" onset in females will be predictive of frequency of subsequent arrests.With these data, the author uses OLS regression, logistic regression, and negative binomial regression to evaluate these hypotheses regarding age of onset, risk factors for onset, and frequency of arrest. Insufficient evidence was found to support the hypotheses of the current study. A discussion of the findings, as well as implications and calls for future research are discussed.
Adviser: John K. Cochran, Ph.D.
t USF Electronic Theses and Dissertations.
GENDER DIFFERENCES IN AGE OF ONSET FOR DELINQUENCY: RISK FACTORS AND CONSEQUENCES by LAURA M. GULLEDGE A thesis submitted in partial fulfillment of the requirements for the degree of Masters of Arts Department of Criminology College of Arts and Sciences University of South Florida Major Professor: John K. Cochran, Ph.D. Christine S. Sellers, Ph.D. Shayne Jones, Ph.D. Date of Approval: April 7, 2006 Keywords: life-course, developmental, predictors, f emales, early Copyright 2006, Laura M. Gulledge
Dedication This work is dedicated to four very dear people. Fi rst and foremost, this manuscript is dedicated to my wonderful husband, Ri chard Gulledge. You have shown me an abundance of love and support throughout this entire process. Your belief in me kept me going even when I did not believe in myself Secondly, I owe tremendous thanks to my incredible parents, Ed and Sandy Thomason. I am where I am today because of your guidance, patience, and unconditional love. Yo u have both made so many sacrifices for me and I will be forever grateful. Finally, thi s thesis is dedicated to the loving memory of my grandfather, John F. Messiner, who ins tilled in me a love of learning that will always be a part of me.
Acknowledgments There are many thanks in order to those who have a ssisted in making this manuscript possible. First, I owe my deepest gratit ude to my Major Professor, Dr. John Cochran. You provided me with countless hours of co unsel, direction, knowledge, and support. Thank you for always pushing me to reach t he highest of standards. I would also like to thank Dr. Christine Sellers and Dr. Shayne Jones, who both gave of their time and expertise to make this work the best it could be. T hank you for all of your encouragement and suggestions. Additionally, I would like to than k Dr. Chris Sullivan for his endless, patient assistance. I would also like to recognize Dr. Chris Gibson for enabling me to pursue this topic. Thank you to all of the others t hat have offered much appreciated feedback and moral support, including: Dr. Denise P aquette-Boots, Lisa Murphy, Kristina Childs, and Sandy Thomason.
i Table of Contents List of Tables ii Abstract iii Chapter One Introduction 1 Chapter Two Developmental / Life-Course Criminology 4 Theoretical Concepts 8 Theoretical Importance 9 Theoretical Evolution 10 Early Antisocial Behavior 12 Theoretical Perspective of Early Onset 13 Male Studies 14 Female Studies 17 Hypotheses 19 Chapter Three Research Design and Methodo logy 21 Dependent Variables 23 Independent Variables 24 Analytic Plan 30 Chapter Four Results 33 Modeling Age of First Arrest 35 Comparing Across Means 36 Modeling Â“EarlyÂ” Onset of Arrest in Males and Femal es 37 Modeling Â“EarlyÂ” Onset as a Predictor of Frequency of Arrest by Gender 41 Chapter Five Conclusion and Discussion 44 References 55 Appendices 65 Appendix A 66 Appendix B 67 Appendix C 69 Appendix D 70 Appendix E 71 Appendix F 72 Appendix G 73 Appendix H 74
ii List of Tables Table 1. Description of the Data 30 Table 2. Age of First Arrest 34 Table 3. Ordinary Least Squares Regression Pred icting Age of First Arrest 36 Table 4a. Mean Comparisons by Age and Gender 38 Table 4b. Significant Differences in Individual R isk Factors by Gender 39 Table 5. Logistic Regression Predicting Early O nset of Arrest by Gender 41 Table 6. Negative Binomial Regression Predictin g Frequency of Arrest by Gender 43
iii GENDER DIFFERENCES IN AGE OF ONSET FOR DELINQUENCY: RISK FACTORS AND CONSEQUENCES Laura M. Gulledge ABSTRACT The age of onset of delinquency has long been viewe d as a primary indicator for further delinquency and criminality. However, studi es on the risk factors for onset, age of onset, and future delinquency have focused predomin antly on males. The purpose of this study was to explore gender differences and similar ities in risk factors for onset and frequency of arrest. The data used in these analyse s were from a longitudinal study, Pathways to Adulthood: A Three Generational Urban S tudy, 1960-1994. Sixty-six percent (N = 1,758) of the eligible children comple ted the final survey. Of these children, only 515 were used in this particular study because they had documented ages of first arrest. It is hypothesized that 1) female Â“earlyÂ” o nset occurs at a later age from that of male Â“earlyÂ” onset, 2) risk factors predictive of e arly onset will differ across gender, and 3) Â“earlyÂ” onset in females will be predictive of f requency of subsequent arrests. With these data, the author uses OLS regression, lo gistic regression, and negative binomial regression to evaluate these hypotheses re garding age of onset, risk factors for onset, and frequency of arrest. Insufficient eviden ce was found to support the hypotheses of the current study. A discussion of the findings, as well as implications and calls for future research are discussed.
1 Introduction In the fields of both criminology and psychology, t he age of onset of delinquency and antisocial behavior has been viewed as a primar y indicator for further delinquency and criminality. However, the relationship between earlier onset and future delinquency and crime has been primarily limited to studies of males (Piquero and Chung, 2001). Traditionally, female delinquency has been disregar ded and underestimated (ChesneyLind and Okamoto, 2001). Most theories of delinquen cy are based solely on male children and adolescents and very rarely is any con sideration given to whether deviance in females differs (Storvoll and Wichstrom, 2002). Females make up a smaller grouping of detained and adjudicated juvenile delinquents co mpared to males, yet female rates of delinquency have steadily increased over the past s everal years (McCabe, Lansing, Garland, & Hough, 2002). Increases in female arrest s markedly surpassed those of males for most of the last decade (Bureau of Justice Stat istics, 1999). Despite these increases, research on the age of ons et, the risk factors for onset, and later criminality, as stated above, have focuse d almost entirely on males. Female populations have been excluded from delinquency stu dies, as most believe that delinquency in females is only a small variation fr om delinquency in males (Hoyt and Scherer, 1998). Because of this, it is very difficu lt to generalize any empirical findings discovered about males to females (McCabe et al, 20 02). However, males and females are not similar populations and findings about male s can not be assumed to also hold true for females. In addition, some studies have found that early-ons et in females is non-
2 existent (Silverthorn and Frick, 1999), or very rar e compared to that of later onset delinquency (Moffitt, 2001). The purpose of this study is to explore gender diff erences in the age of onset, risk factors for onset, and the effects of early onset o n the frequency of arrest. Three interrelated questions regarding female onset of de linquency include (1) whether or not there is evidence of early onset of delinquency in females and, if so, at what age this occurs and if this age differs from males; (2) if t here are similar or different factors that predict early onset in males and females; and, (3) if an early age of onset is related to frequency of arrest in females. The data employed t o examine these questions came from Hardy and ShapiroÂ’s study, Pathways to Adulthood: A Three Generational Urban Study, 1960-1994: [Baltimore, Maryland]. Ordinary Least Sq uares regression, logistic regression, and negative binomial regression analys es are used to evaluate the research questions regarding age of onset, risk factors for onset, and frequency of arrest. Chapter Two offers a brief history of developmental and/or life-course research. Definitions, generally accepted conclusions, and ce ntral focuses of life-course study are explained. A literature review of several major lif e-course theories is also provided. The current body of research related to early onset in males is then reviewed. The lack of study in the field regarding male and female differ ences as they relate to onset, risk factors for onset, and offending is explained. The few studies relating to female early onset, as well as how their findings compare to mal e related studies, are discussed. Finally, the hypotheses for this study are derived from the reviewed literature. Chapter Three provides an overview of the methods u sed in the current study. Characteristics of the sample are described. The fi rst outcome measure in the current
3 study is age of first arrest, or the age in years t hat the respondent was first arrested, booked, or charged by an authority of the law. Inde pendent variables that are defined and described include family adversity factors, a famil y conflict tactics scale, neurologicalcognitive indicators, drug use, deviant peer associ ations, school deviance, sexual abuse, and frequency of arrest. Models are then presented that employ the analytic techniques of Ordinary Least Squares regression, logistic regress ion, and negative binomial regression. Chapter Four presents the results of multivariate a nalyses of these data. First, the age at which onset occurs and whether this age diff ers among females and males is discussed using various frequency distributions. Ne xt, the influence of risk factors for early onset across gender is explored through vario us multivariate regression techniques. Finally, negative binomial regression analysis enab les discussion of whether of not there are gender differences in frequency of arrest for f emales. Finally, Chapter Five concludes with a summary and discussion of the current study and focuses on its purpose, design, major fin dings, and theoretical implications of the results. Limitations, policy implications, and suggestions for future research are also discussed.
4 Chapter Two Developmental / Life-Course Criminology Throughout the past quarter century, research and t heory in the social and behavioral sciences have embraced two main perspect ives in order to better understand the nature and complexity of human behavior over ti me (Elder, 1994). The life-course perspective emerging from sociology focuses on the various common patterns of trajectories and turning points that people experie nce throughout their life course. The developmental perspective from psychology examines the stages and processes in human psycho-social development. Although both perspectiv es hold long-standing traditions of study within their respective disciplines, the inte rdisciplinary field of criminology has overlooked the developmental aspects of deviance an d crime until fairly recently. Traditional avenues of criminological research comp are differences across individuals that are believed to generate crime; by contrast, d evelopmental/life-course approaches in criminology focus on changes over time within indiv iduals that are also believed to generate crime. Developmental or life-course crimin ology is a broad multidisciplinary arena that weaves together ideas from many differen t perspectives (Benson, 2002). In particular, life-course criminology evolved through the years as a blend between developmental psychology and life-course sociology. Developmental psychology is defined as the scientif ic study of psychological transformations that take place as people mature (H ogan, 2000). Areas of study in this field include psycho-physiological processes such a s motor or perceptual abilities,
5 language skills, abstract reasoning and understandi ng, moral appreciation, problemsolving, and identity construction. These topics al low answers to questions involving the differences between children and adults, and the di ffering processes that lead to the attainment of knowledge. Influential developmental psychologists such as Piaget, Erikson, and Kohlberg have investigated key questio ns regarding stages of cognitive development, psychosocial development, and moral de velopment, respectively (Carpendale, 2000; Jenkins, 2005). Though developmental psychology currently encompass es the entire life span, the field originally stressed the importance of early c hildhood through late adolescence. Moreover, developmental psychology primarily deals with psychological development in the broad sense without an explanation of how this development affects manifestation of deviant and/or criminal behaviors (LeBlanc, 1997). Unlike various criminological perspectives, developmental psychology does not nec essarily include crime as an important factor. In fact, crime may not be include d at all. Because of developmental psychologyÂ’s original focus on early childhood and adolescent development, life-course sociologists saw the need to fill the void between development and behavior throughout the life span and thus, started their studies where developmental psychology essentially ended (Piquero and Mazerolle, 2001). The first strong push for sociological life-course study began in the 1960Â’s when the relationship between social changes and the liv es of individuals began to take on new importance. The notion of studying the life-course first began with the growth of U.S. cities resulting from European migration. American sociologist, William I. Thomas, traced patterns in experiences across generations a nd promoted a Â“longitudinal approach
6 to life historyÂ” (Volkart, 1951, p. 593). ThomasÂ’ t hinking that research should focus on past, present, and future experiences of individual s (Elder, 1985) led to essays including Norman RyderÂ’s, The Cohort as a Concept in the Study of Social Chan ge (1965), which provided a new view of the relationship between soc ial changes and the behavior of various age cohorts. RyderÂ’s essay argued that diff erences in behavior, at varying ages, could be explained through historical changes in on eÂ’s life (Ryder, 1965). Building on this concept that changes occur through out oneÂ’s entire life, Glen Elder, a leading life-course sociologist, held that aging and development are continuous processes that change over time (Elder, 1985). Acco rding to Elder (1998), basic lifecourse elements include Â“multiple trajectories of i ndividuals and their developmental implicationsÂ” (p.1). Simply put, trajectories are p athways or lines of development over the life span, such as family, work, or school (Ben son, 2002). Trajectories are composed of long-term patterns and sequences of behavior and experiences that are marked by change. Trajectories can be examined by combining s tates, such as states of health, across a personÂ’s life span. Changes, or turning points, i n these states are referred to as transitions. Transitions are marked by specific lif e events that are Â“more or less abruptÂ” (Elder, 1985, p.32) and include events such as gett ing married, graduating, or acquiring a job (Sampson and Laub, 1992). These turning points result in a trajectoryÂ’s change, and in turn, lead to a change in oneÂ’s life course (Eld er, 1985; Thornberry, 1997). How a person responds to transitions is extremely importa nt, as oneÂ’s reactions determine how their trajectory may change. The theme of transitio n refers to how turning points can alter life trajectories, while the theme of trajectories in life-course leads to connections regarding childhood events and adult experiences (S ampson and Laub, 1992).
7 Applying these themes of transitions and turning po ints to the study of criminal behavior enables better understanding of why and ho w people begin, continue, and end criminal behaviors across the life span (LeBlanc an d Loeber, 1998). This criminal career, or Â“longitudinal sequence of crimes committed by an individual offender,Â” (Blumstein, Roth, & Visher, 1986, p.12) is characterized by tra jectories with elements including onset, stabilization, and desistence. Though transi tions and turning points are central to most life-course theories, many theorists disagree on the specific implications of maturation as it relates to crime (Piquero and Maze rolle, 2001). Debate ensues when determining whether a pattern of crime across the l ife-course is categorized by continuity or change (Blumstein and Cohen, 1979; Farrington, 1 986; Gottfredson and Hirschi, 1987). Continuity is characterized by stable and continuou s behavior throughout the life course. In terms of the criminological life-course perspective, continuity refers to the antisocial and/or criminal behaviors manifesting th emselves in childhood that correlate to similar behaviors displayed in adulthood (Elliott, 1985; Wolfgang, Thornberry, & Figlio, 1987). Conversely, change is characterized as behav ior that starts in one direction and then moves in another direction. Views of change st ress the fact that most juvenile offenders do not continue their offending patterns into adulthood (Robins, 1978). The challenge of life-course theory, then, is to explai n continuity and/or change in criminal involvement across the life-course. Consequently, b y borrowing from the disciplines of developmental psychology and life-course sociology, American criminologists recently began to study the links between historical age and cohort as compared to social age in the life course (Benson, 2002).
8 Theoretical Concepts Before examining the predominant concepts and theor ies within the criminological developmental/life-course perspectiv e, it is necessary to mention what exactly a life-course theory is intended to explain Most use ElderÂ’s (1985) definition to characterize the life-course as a series of Â“pathwa ys through the age differentiated life span,Â” or the interrelated paths that people take t hrough life as they age. Accordingly, life course theories explain crime as a process, develop ing from childhood into adulthood (Farrington, 1986; Loeber and LeBlanc, 1990; Sampso n and Laub, 1991; Moffitt, 1993). Farrington (2003) outlines ten widely accepted conc lusions that any life-course or developmental theory of criminal careers must be ab le to explain. These conclusions include statements regarding onset age of offending peak ages in offending and desistence, versatility of offending, continuities in behavior from childhood through adolescence, differences between preand post-adol escent offending, variability in reasons for offending, and predictions of offending duration related to age of onset. Age of onset peaks between age 8 and 14, while age of d esistence peaks between age 20 and 29. The earlier the age of onset, the more likely o ne is to have a longer period of offending. However, a majority of all offenses are committed during the late teenage years. During adolescence, a majority of offenses a re committed with peers. However, after adolescence, offenses are usually committed a lone. The largest percentage of crime is actually committed by a small percentage of the population. Regarding behavior, there are similarities in behavior from childhood through adolescence and then into adulthood. There are many different types of offending and ant isocial behaviors. Reasons for offending change with age. In addition, different t ypes of crimes are committed at
9 different age groupings. By addressing these facts in an attempt to explain offending development, risk factors for offending, and the ef fects of life events on offending, life course theories are able to explain crime in very d istinctive ways. (Farrington, 2003). Theoretical Importance Life-course theories can explain criminal behavior and delinquency in ways that other relatively static criminological theories can not. By observing offending behavior over time within the individual and taking into con sideration oneÂ’s life circumstances, life course theories are better able to explain bot h continuity and change in offending behavior through the life span. In addition, life c ourse theories can speak to the probability of future criminal behavior and delinqu ency. By studying these age-related themes over time, lif e-course research has found the age-crime relationship to be one of the most consis tent empirical findings (Thornberry, 1997). Empirical data show that a majority of crime is committed by adolescents, who then desist as they Â“age outÂ” of their late teenage years (Hirschi and Gottfredson, 1983; Farrington, 1986, 1992; Wolfgang et al., 1987; Thor nberry, 1997). Because of these findings regarding adolescents, studies in criminol ogy have tended to ignore childhood behaviors and characteristics as they relate to lat er adult behaviors and characteristics (Caspi, Elder, & Bern, 1989; Farrington, 1989; Loeb er and LeBlanc, 1990; Sampson and Laub, 1990). As a result of this adolescent focus, the implications of early childhood experience on both adolescent and adult development have not yet been well addressed by criminologists (Sampson and Laub, 1992). Recently, however, developmental and life-course th eories have found their way to the forefront of criminology-based research as a number of significant longitudinal
10 studies have been published (Farrington, 2003). Th ese include additional examinations of the GluecksÂ’ study (Sampson and Laub, 1993); the Dunedin study in New Zealand (Moffitt, Caspi, Rutter, & Silva 2001); the Causes and Correlates studies begun by the Office of Juvenile Justice and Delinquency Preventi on in Denver, Pittsburgh, and Rochester (Huizinga, Weiher, Espiritu, & Esbensen, 2003; Loeber, Farrington, Stouthamer-Loeber, Moffitt, & Caspi, 2003; Thornber ry, Lizotte, Krohn, Smith, & Porter, 2003); the Seattle Social Development Proje ct (Hawkins et al, 2003); and Tremblay and colleaguesÂ’ (Tremblay, Masse, Vitaro, & Dobkin, 2003) Montreal Longitudinal-Experiment study. Theoretical Evolution A ground-breaking study in the use of longitudinal research began in the late 1930s with the work of Glueck and Glueck (Glueck and Glue ck, 1950). The Crime Causation Study: Unraveling Juvenile Delinquency recognized t he importance of longitudinal data collection in regard to life-course research and de linquency. The Gluecks followed their sample of 500 delinquent boys and 500 non-delinquen t boys from low socio-economic neighborhoods over a nine year period. This compreh ensive data set included biological, psychological, familial, school, work, and delinque ncy indicators. These variables were measured through self-reports, parent reports, teac her reports, and official records (Sampson and Laub, 1993). Another crucial project in the history of life-cour se research was undertaken by Wolfgang, Figlio, and Sellin during the early 1960s Wolfgang et al., (1972) collected detailed information on over 10,000 boys from Phila delphia. Data were collected over an eight-year period in order to observe the longitudi nal progression of criminal behavior.
11 Original data were taken only from official records but detailed interviews conducted when the subjects were in their mid-twenties enable d the researches to make assessments regarding offending behavior from childhood through young adulthood (Wolfgang et al., 1972). More recently, Blumstein, Cohen, Roth, and Visher ( 1986) had a very notable impact in life-course criminology with their research on c areer criminals and criminal careers. The criminal career is defined as the Â“longitudinal sequence of offenses committed by an offender who has a detectable rate of offending dur ing some periodÂ” (Blumstein et al., 1986, 2). Career criminals Â“commit serious offenses at high rates and over extended periods of timeÂ” (Blumstein et al., 1986, 2). The concept of the criminal career holds value to the study of life-course theory as it also stresses the importance of longitudinal data collection. Loeber and LeBlanc (1990) were the first life-cours e theorists to use the term Â“developmental criminologyÂ” as that which focuses o n continuity and within-individual changes in offending over time. Loeber and LeBlanc highlight the occurrence and timing of life circumstances in their explanations of offe nding. By focusing on the differences between correlates and causal factors of offending, developmental criminology adds another important facet into the life-course perspe ctive (Loeber and LeBlanc, 1990). Initiating the first real theoretical debate in cri minology in decades, Sampson and Laub (1993) were pioneers in studying the sources o f continuity and change in crime over the life span. Like Loeber and LeBlanc, they focuse d on within-individual changes in offending in order to explain stability and change over the life-course.
12 Early Antisocial Behavior Many researchers have focused on early onset of an tisocial behaviors in children as an influential factor in later criminal behavior and delinquency. Robins (1978) was one of the first researchers to look into the correlati ons between antisocial behaviors (ASB) in children and adults. After collecting data on four male cohorts, results indicated that all types of antisocial behaviors studied in the childr en were indicators of adult antisocial behavior. Though not all male children with these b ehaviors carried them over into adulthood, Robins (1978) found that an overwhelming majority of the adults with ASB demonstrated these behaviors as children. In light of the importance of age regarding early a ntisocial behavior, life-course criminologists propose that age is a crucial compon ent in the understanding of delinquent behavior as well (Bartusch, Lynam, Moffit, & Silva, 1997). Life-course theories hypothesize that prior behavior is linked to future behavior (Nagin and Farington, 1992). Accordingly, research shows that age of onset is th e single best predictor of future offending (Farrington, Loeber, & Van Kammen, 1990). Â“The term onset typically refers to a discrete change in state namely from nonoffend er to offenderÂ” (Piquero and Chung, 2001, p.190). Age of onset is important for several reasons. Stud ies have posited that early onset is a significant indicator of later offending durin g adolescence and adulthood (Wolfgang, 1983; Loeber and LeBlanc, 1990; Sampson and Laub, 1 993). In addition, it has also been argued that the probability of oneÂ’s continuance in delinquency increases as the age of onset decreases (Piquero and Chung, 2001). However, what is actually meant by the term Â“earlyÂ” onset is not a concrete construct in the li fe-course perspective. Loeber and
13 Farrington (2000) cite early onset as delinquent be havior occurring before age thirteen; alternatively, other studies have included age thir teen in their definition of early onset (Tremblay, Masse, Vitaro, & Dobkin, 1995). Theoretical Perspectives of Early Onset In an attempt to further investigate the relationsh ip between age and onset of offending, Moffitt (1993) suggested that antisocial behavior that begins in childhood is qualitatively different than antisocial behavior or iginating during adolescence. Her developmental taxonomy (1993) breaks offenders down into two categories in order to explain why one group manifests continuity in offen ding while the other group manifests change in offending. The first group, life-course p ersistent offenders (LCPs), develop antisocial behaviors early in childhood. These anti social behaviors escalate into crime and delinquency through adolescence and into adulth ood. The second group, adolescentlimited offenders (ALs), begin and end their crimin al offending during adolescence. In addition to explaining unique aspects regarding the development of offending such as onset, life-course theories also try to exp lain the varying risk factors that affect the development of offending (Nagin and Farrington, 1992). According to Moffitt (1993), individual neurological traits and deficits interac t with an individualÂ’s social environment to produce early antisocial behavior. Neuropsycholo gical problems result from disruption in normal brain development that leads to further p sychological deficits. Neurological development can be disrupted by many factors includ ing drug and alcohol use by the mother, poor prenatal nutrition, exposure to toxins complications during delivery, inheritable neurological problems, and child abuse and neglect. Psychological characteristics that can be affected by neurologica l deficits include temperament,
14 behavior, and cognitive abilities. These deficits c ombine with social environments to promote early antisocial behaviors for life-course persistent offenders. These deficits create stability in offending through their constan t, or Â“contemporaryÂ” effects in the individual since early childhood and through the Â“c umulative continuityÂ” that develops over time. As a result, there are fewer opportuniti es for a LCP offender to learn appropriate alternatives to antisocial behavior. Unlike LCPs, adolescent-limited offenders often exh ibit abrupt changes in delinquency and do not tend to show any continuity in antisocial behaviors. ALs exhibit no history of childhood antisocial behaviors. Most ALs begin offending during puberty, peak during the later teenage years, and desist by the time they reach adulthood. Whereas LCP offending behavior originates from a neurologic al basis, AL offending behavior originates with a maturity gap and continues with s eeing the delinquency of other LCP youths. Moffitt defines this process by which ALs begin offending as Â“social mimicry.Â” Models developed by Patterson and colleagues (Patte rson, DeBaryshe, & Ramsey, 1989) also show that peers are more influential for those who onset at later ages. AL youth begin to desist when they enter legitimate adult ro les and can acquire desires through legitimate means. It is easier for AL offenders to desist from offending because they are not as affected by cumulative and contemporary cont inuity as are LCP offenders. Male Studies As referenced above, Moffitt (1993) has found evide nce suggesting that neuropsychological problems in early childhood are a risk factor for delinquency and antisocial behaviors of male offenders. Moffitt and colleagues studied data from a battery of neuropsychological tests administered to a cohor t of several hundred 13-year old New
15 Zealand males. This study produced the first eviden ce of a prospective link between early-measured neuropsychological test scores and f uture offending in males. Results proved that poor neuropsychological scores at age 1 3 were correlated with early onset of delinquency. In addition, males with the lowest sco res had the highest levels of delinquency when measured five years later. The tes ts predicted later delinquency measured though multiple sources including police, courts, and self-reports. Other research examining the influence of different risk factors on early antisocial behavior includes biological and physiological dete rminants (Brennan et al., 1995; Raine et al., 1997). Farrington and Hawkins (1991) establ ished that poor psychomotor skills were a factor influencing antisocial behaviors. Oth er risk factors that influence onset of delinquency and antisocial behavior include context ual, child, and parent risk factors. Community risk factors include access to firearms a nd drugs. Familial risk factors include parental mental illness or criminal behavio r, negative parental attitudes, and poor family management (Preski and Shelton, 2001). One such study that looks at a variety of risk fact ors regarding the development of juvenile offending is the Pittsburgh Youth Study (L oeber, Farrington, StouthamerLoeber, Moffitt, & Caspi, 1998). Participants in th is study included over 1,500 preadolescent and adolescent inner-city boys. Assessme nts that measured risk and protective factors and antisocial behaviors were administered over a ten-year period to three different samples of boys. Results showed that deli nquency was particularly related to conduct problems, as well as measures of impulsivit y, IQ, and personality. All three samples showed correlations between delinquency and child, family, and contextual explanatory variables. Having deviant peers was als o a significant predictor of onset.
16 Loeber et al. (1998) found that the probability of delinquency increased with increasing numbers of risk factors. Two additional studies eme rged using samples from the Pittsburgh Youth Study data including Moffitt and R aineÂ’s analyses of boys from the second group of assessments. Moffitt used these ass essments to look into measures including neuropsychology, impulsivity, and persona lity. Raine took a closer look at the boys from the second sample as well in regards to p sychophysiological and biological risk factors for violence. Another study that looks at the relationship betwee n delinquency and personality assessments over time was conducted by Tremblay, Ph il, Vitaro, and Dobkin (1994) in Canada. Over 1,000 boys from low socioeconomic scho ols were studied from kindergarten through age thirteen in order to test personality as a risk factor for early onset of delinquency. Personality dimensions in the assessments included anxiety, impulsivity, and reward dependence. Self-reported d elinquency scales were administered to the boys when they were between the ages of ten and thirteen. High impulsivity rated by the boysÂ’ kindergarten teachers proved to be the best indicator for future self-reported delinquency. The other two personality dimensions o f anxiety and reward dependence contributed to predicting future offending, though their correlations were not as strong as impulsivity. The above studies demonstrate how early onset in ma les is a consistent indicator of future delinquency. However, whether the relatio nship between early onset and offending is a male, and not female issue, is a topic that has had little investigatio n (Piquero and Chung, 2001). Research has shown that deviant friends, failure at school, and leisure activities are more strongly related to conduct problems for boys than for girls
17 (Stattin and Magnusson, 1995; Storvoll and Wichstro m, 2002; Werner and Silbereisen, 2003). Life-course persistent paths in males have b een found to be related to dropping out of school, impulsivity, personality traits of alien ation, acts of violence, and drug abuse (Moffitt, Caspi, Harrington, & Milne, 2002; Hanlon, Bateman, Simon, OÂ’Grady, & Carswell, 2004). In addition, MoffittÂ’s theory invo lving the interaction of neuropsychological factors and disadvantaged enviro nments has been found to be significant risk factors for boys, but not for girl s (Moffitt et al., 2002). Moffit (2002) even suggests that the issue of early onset in females m ay be nonexistent. Female Studies Because it is believed that females initiate onset at significantly later ages than males, little is known about the precursors and cor relates of early female delinquency (Hipwell, Loeber, Stouthamer, Keenan, White, & Kron eman, 2002). Most of the studies that have included females are related to various f orms of delinquency, though not very many specifically look at early onset. A majority o f these studies involving females point to various forms of abuse as indicators for deviant behavior and criminality. One such study showed a significant relationship be tween maltreatment and delinquency in females (Widom, 1989). In her Cycle of Violence study, Widom (1989) included males and females in the sample, yet found some interesting results relating to females. Though the males had higher rates of delin quency and adult criminality than the females, Widom found that the abused or neglected f emales were significantly more likely to have an arrest than the control females w ho reported no incidences of abuse or neglect. These results regarding female arrest did not hold when involving violent crime. However, Widom explains that the long-term outcomes of abuse and neglect in young
18 girls come across in different ways than in their m ale counterparts. She argues that females may be more likely to undergo depression be cause of these incidences of abuse, rather than externally display forms of violent beh avior as with males. In addition, Widom points out that females are more likely to su ffer sexual abuse than males, which may also affect long-term outcomes of behavior. A further study found that sexual abuse is a unique risk factor in females (Fergusson and Woodward, 2000; Herrera and McCloske y, 2001). In addition, the onset of puberty has been shown to be more strongly relat ed to conduct problems in females than in males (Storvoll and Wichstrom, 2002). Femal es have also been found to present higher rates of psychopathology, maltreatment histo ry, and familial risk factors than males (McCabe et al., 2002). Though these findings can be explained as risk factors for delinquency, limited research has been done associa ted with risk factors for earlier onset in females (Gibson, Piquero, & Tibbetts, 2001). Physical aggression has also been looked at as an o nset indicator of later delinquency and criminality. Boys are more likely t o display acts of violence and physical aggression than girls (Widom, 1989; Moffit t et al., 2002). However, in relation to boys, girls are more likely to display covert as opposed to overt forms of aggressive behavior (Kazdin, 1992). In addition, a form of agg ression more common to females is known as relational aggression or indirect aggressi on. Covert indicators for later delinquency include stealing, lying, and gossiping. Relational aggression refers to harming others through manipulation, or damage to r elationships such as ostracizing peers (Tiet, Wasserman, Loeber, McReynolds, & Mille r, 2001).
19 Persistent offending has been shown to begin with problem behaviors in early childhood (Gibson, Piquero, & Tibbetts, 2001). Beca use of the higher prevalence of antisocial behavior in males, much less is known ab out the outcomes of girls with behavioral problems later on in their lives (Fergus son and Woodward, 2000). Little work has been done focusing on female onset as it relate s to the number or frequency of arrests in late adolescence and early adulthood. Research o n the chronic female offender is scarce and research relating chronic female offendi ng to early onset is even scarcer (Piquero, 2000). Leve and Chamberlain (2004) found that early onset of delinquency in females leads to poorer outcomes and long-term prob lems. Cote, Zoccolillo, Tremblay, Nagin, and Vitaro (2001) have also suggested that o nset problem behaviors in females are significant predictors of later serious delinquency Pajer (1998) also found evidence supporting negative long-term outcomes stemming fro m early delinquency in females. However, she, along with many other life-course cri minologists, acknowledges that there is insufficient research and data to fully understa nd these issues. The current study examines gender differences in th e age of onset, risk factors for onset, and the effects of early onset on frequency of future offending. Data from the second generation children involved in the Pathways to Adulthood: A Three-Generation Urban Study, 1960-1994: [Baltimore, Maryland] are u sed to test the following hypotheses regarding female early onset of delinque ncy: Hypotheses 1. Female early onset occurs at a later age from that of male early onset. 2. Risk factors including delinquent peers, leisure ac tivities, and behavioral and academic problems in school are predictive of early onset in males; whereas
20 maltreatment, sexual abuse, and other familial prob lems are predictive of early onset in females. 3. Age of onset in females is predictive of future off ending.
21 Chapter Three Research Design and Methodology The data used to examine these hypotheses came from Hardy and ShapiroÂ’s study, Pathways to Adulthood: A Three-Generation Urban Stu dy, 1960-1994: [Baltimore, Maryland]. The Pathways to Adulthood study included both prospective and retrospective data on three generations of inner-ci ty families from Baltimore, Maryland. The prospective data were drawn from data collected during the National Collaborative Perinatal Project (NCPP). The retrospective data ca me from the Johns Hopkins Collaborative Perinatal Study (JHCPS). Follow-up d ata were collected approximately twenty-five years after the JHCPS ended. The NCPP was a survey of pregnant women seeking pr enatal care and delivery at Johns Hopkins Hospital during 1960-1964. The women were selected on the basis of the last digit of their hospital history number. This n umber was assigned from a central hospital file at point of first patient contact, of ten many years earlier. The survey looked at these first generation mothers and the children born to them during 1960-1965, coded as G2 (second generation), until the children were eight years old. Data on the second genreation included delivery room observations at b irth; pediatric examinations at certain ages; developmental, neurological, language, hearin g, speech, vision, motor, and physical evaluations at certain ages; psychological and beha vioral profiles; and various other tests. To qualify for the Pathways Follow-up study, secon d generation children had to have been born between 1960-1965 and completed the 7 and/or 8-year assessments of the JHCPS. Of the 2,694 second generation children elig ible for follow-up, only 1,758 completed the 67-page, standard format questionnair e. Of the 1,758 children who
22 completed the final interview, 807 were male and 95 1 were female. The second generation retrospective data from the follow-up st udy, conducted when the respondents were age 30-35, included information on aspirations education, schooling, employment, family composition, health, housing conditions, inc ome, legal problems, living arrangement, marriage, neighborhood characteristics at varying ages, reproductive history, social relationships, smoking, and substan ce abuse. In addition to the standard questionnaire, each second generation child complet ed a life history calendar as well. There are limitations that must be considered when using this data set. Of the approximately 4,000 mothers enrolled in the JHCPS, 77% were black, 22% were white and less than 1% were from other racial groups. The y were generally poor and had very low educational levels. Most of the women were empl oyed in domestic and clerical positions. Furthermore, the second generation follow-up interv iew produced only retrospective data, requiring that respondents reme mber correctly about the past and be willing to admit to their previous actions. The val idity of self-reported criminal offenses in adult samples is generally lower than in juvenil e samples (Junger-Tas & Marshall, 1999). Self-report data are also complicated by the fact that the temporal order of negative predictors and negative outcomes is imposs ible to discern without prospective data collection. These facts hinder the ability to discuss causal order. Because of these sampling limitations, one must be careful when tryi ng to generalize the data to other groups and populations. This speaks to the dataÂ’s e xternal validity as it will be difficult to generalize results to other settings. Nevertheless, the data are still very diverse in their sample characteristics and remain a representative sample for similar populations.
23 Dependent Variables Age of First Arrest, Booking, or Charge. According to Piquero and Chung (p. 190, 2001), Â“the term onset typically refers to a d iscrete change in state namely from nonoffender to offender.Â” Onset can be measured usi ng many different methods and factors. Onset has been operationalized in differen t studies as first offense, first arrest, and first conviction. Onset has been measured by of ficial, retrospective, and prospective methods (Piquero and Chung, 2001). It must be noted that there are different domains o f offending in these data that can be used to assess age of onset. These domains v ary as they relate to gender. For this analysis, only self-reported age of first arrest wa s used. Measurement validity issues are raised when using self-reported age of first arrest to measure onset. The results may be biased because age of first arrest does not take in to account the fact that respondents may have done previous delinquent acts for which they w ere not arrested. Therefore, the onset age discovered in these analyses may, in fact, be o lder than onset age using other domains. In addition, measurement problems arise wh en dealing with any type of memory recall. A combination of self-reports and of ficial arrest records would give the best results when asking about the age of first arr est. However, official arrest records are not available in the data set. The first outcome measure in this study is age of first arrest (i.e., the age in years that the respondent was first arrest, booked, or ch arged by an authority of the law). At the follow-up interview, second generation children wer e asked a single-item question regarding the age at which they were first booked, charged, or arrested for breaking the
24 law. Of the sample, 29.3% (515) reported having bee n booked, charged, or arrested. Of these 515 respondents, 74.4% (383) were male and 25 .6% (132) were female. Independent Variables Independent variables include family adversity, chi ld abuse, neurologicalcognitive, drug use, school deviance, deviant peer associations, and stand-alone risk factors including gender and frequency of arrest. T hese variables from the NCPP, JHCPS, and follow-up survey were chosen based on re search linking their relationship to early onset of delinquency. Family Adversity. Mothers (G1) were asked questions regarding their age, whether or not they were receiving public assistanc e, level of education, marital status, and income at the time of their childÂ’s (G2) birth. Mothers age was labeled young (1) if they gave birth at age seventeen or younger, or non -young (0) if they gave birth at age eighteen or older. Public assistance was recoded in to (0) not receiving public assistance and (1) receiving public assistance. Income was con ceptualized by using Farrington and LoeberÂ’s risk factor paradigm (1999), with the lowe st 25% of income indicating low income. The variable was dichotomized to differenti ate those with low income (1) from those with non-low income (0). Marital status was d ichotomized into (0) married when their child (G2) was born and (1) single. An educat ional scale was used to measure mothers educational level at the time of their chil dÂ’s birth, with (1) less than eighth grade through (6) graduate work. The family adversity items (motherÂ’s age, public as sistance, educational level, marital status, and income) were assessed for scala bility using a principal components factor analysis (Appendix A). Among females, the fa mily adversity items formed a single
25 factor solution (eigenvalue = 1.874) with loadings that ranged from .570 to .644. Among males, these items also formed a single factor solu tion (eigenvalue = 1.781) with loading that ranged from .495 to .720. Final scores on the family adversity scale range from zero to ten. Child Abuse. Family conflict tactics were conceptualized by ask ing how the respondents interacted with family members during d isagreements. A scale of physical abuse during disagreements included whether or not the parent threatened to hit or throw something, threw something, pushed or shoved, slapp ed or spanked, kicked, bit, hit, beat up, burned or scalded, or threatened or used a knif e or gun on the child. These items were assessed for scalability using a principal componen ts factor analysis (Appendix B). Among females, the family-conflict tactics items fo rmed a single factor solution (eigenvalue = 6.068) with loadings that ranged from .588 to .821. Among males, these items also formed a single factor solution (eigenva lue = 4.266) with loading that ranged from .461 to .708. Six frequency indicators include never, once, twice, sometimes, frequently, and most of the time. These indicators were used as discrete variables with 0 = never and 6 = most of the time. In this additive scale, higher scores indicate higher levels of parental physical abuse. Respondent score s range from zero to forty-four. From the subsample of respondentsÂ’ reporting an arrest, 82% (444) indicated receiving some form of parental physical abuse on the conflict tac tics scale. Sexual abuse is conceptualized as inappropriate sex ual behavior committed by a person responsible for the care of the child. From the subsample of respondents reporting an arrest, 8.5% (44) reported being sexually abused The sexual abuse indicator and the family conflict tactics scale were assessed for sca lability using a principal components
26 factor analysis (Appendix C). Among females, sexual abuse and family conflict tactics scale items formed a single factor solution (eigenv alue = 1.290) with loadings of .803. Among males, these items also formed a single facto r solution (eigenvalue = 1.109) with loadings of .745. Final scores on the child abuse s cale range from zero to 45. Neurological-cognitive. Neuro-cognitive factors include birth weight, verb al IQ score, WRAT reading score, WRAT arithmetic score, a nd WRAT spelling score. Birth weight was obtained at time of childÂ’s birth. Using the World Health OrganizationÂ’s standard low birth weight cutoff of < 2500 grams, low birth weight was conceptualized by the childÂ’s weight being < 2500 grams (5.8 pound s). This variable was recoded to 1 (low birth weight) and 0 (non low birth weight). Fr om the subsample of those reporting an arrest, 13.0% (67) of respondents had a low birt h weight. Verbal IQ was assessed at age 7 using the WISC verb al IQ score. Verbal IQ was conceptualized by using Farrington and LoeberÂ’s ris k factor paradigm (1999), with the lowest 25% of scores (poor scores) indicating low v erbal IQ. The variable was dichotomized to differentiate those children with l ow verbal IQ (1) from those with nonlow verbal IQs (0). From the subsample of responden tsÂ’ reporting an arrest, 21.7% (112) had a verbal IQ in the lowest 25 percentile. The Wide Range Achievement Test (WRAT) is a screen ing test that can be administered to measure the development of reading, spelling, and arithmetic skills. Also assessed at age 7, WRAT scores on each of the three tests (reading, spelling, and arithmetic) were conceptualized by using Farrington and LoeberÂ’s risk factor paradigm (1999), with the lowest 25% of scores (poor scores) indicating low WRAT scores. The
27 variable was dichotomized to differentiate those ch ildren with low WRAT scores (1) from those with non-low WRAT scores (0). The neuro-cognitive factors (birth weight, verbal IQ, and WRAT scores) were assessed for scalability using a principal componen ts factor analysis (Appendix D). Among females, the neuro-cognitive items formed a s ingle factor solution (eigenvalue = 2.527) with loadings that ranged from .466 to .828. Among males, these items also formed a single factor solution (eigenvalue = 2.239 ) with loading that ranged from .303 to .779. Final scores on the neuro-cognitive scale range from zero to four. Drug Use. Drug use is conceptualized as the extent to which the respondents indicated whether they had ever used illegal drugs including marijuana, cocaine, heroin, methadone, opiates, and glue. Involvement with thes e drugs was assessed by asking respondents, Â“Have you ever used at least one of th ese drugs to get high, or for other mental effects, or more than was prescribed or for longer than the doctor wanted you to?Â” From the subsample of respondentsÂ’ reporting an arr est, 89.1% (459) reported marijuana use; 60.2% (310) reported cocaine use; 39.6% (204) reported heroin use; 8.0% (41) reported methadone use; and 15.0% (77) reported opi ate use. Answers to questions regarding substance use have b een shown to demonstrate high reliability in studies of urban and rural yout h of diverse races (Starfield, 1997). However, in reviewing the validity of self-reported drug use, Magura and Kang (1996) found that marijuana was more frequently admitted t han the use of other drugs. The six drugs were assessed for scalability using a princip al components factor analysis (Appendix E). Among females, the drug items formed a single factor solution (eigenvalue = 4.488) with loadings that ranged from .920 to .96 8. Among males, these items also
28 formed a single factor solution (eigenvalue = 4.701 ) with loading that ranged from .961 to .978. The five drug variables (marijuana, cocain e, heroin, methadone, and opiates) were all separately dichotomized to differentiate t hose children who had used the individual drug (1) from those who had not used the individual drug (0). An additive index is employed measuring the frequency in which the respondents engaged in drug use. Final respondent scores on the additive scale range from one to ten. School Deviance School deviance is conceptualized as the extent t o which the respondents participated in behaviors while during school hours that were against school rules. The following three items were used to creat e a subscale representing school deviance: getting into trouble with teachers/princi pal for misbehaving in grade school, getting into trouble at school for fighting, and wh ether or not the respondent was expelled/suspended. The items were assessed for sca lability using a principal components factor analysis (Appendix F). Among females, the sc hool deviance items formed a single factor solution (eigenvalue = 1.499) with loadings that ranged from .449 to .851. Among males, these items also formed a single factor solu tion (eigenvalue = 1.636) with loading that ranged from .676 to .780. An additive index is employed measuring the frequency with which the respondents engaged in school devian ce. Final respondent scores on the additive scale range from zero to three. From the s ubsample of respondents reporting an arrest (515), 40.6% (209) reported misbehaving in s chool; 62.7% (323) reported fighting; and 78.8% reported being suspended or expelled. Deviant Peer Associations Deviant peer association is conceptualized as the extent to which respondents spent time with friends who participated in deviant (substance use) and/or criminal activity. Responden ts were asked to think about close
29 friends they hung around with when they were younge r. They were then asked whether most, some, or none of their friends did certain ac tivities including involvement with criminal activity, smoking, drug use, and drinking. Responses to these four questions are combined to create a scale of deviant peer associat ion. The items were assessed for scalability using a principal components factor ana lysis (Appendix G). Among females, the deviant peer association items formed a single factor solution (eigenvalue = 1.893) with loadings that ranged from .595 to .826. Among males, these items also formed a single factor solution (eigenvalue = 2.300) with lo ading that ranged from .669 to .816. The three frequency indicators were used as discret e variables with 0 = no friends, 1 = some friends, and 2 = most friends. In this additiv e scale, higher responses indicate a larger number of respondentsÂ’ friends involved in d eviant activities. Final respondent scores on the additive scale range from four to twe lve. From the subsample of respondents reporting an arre st, 28.9% (149) reported having no friends involved in criminal activities, while 49.9% (257) and 20.6% (106) reported having some friends or most friends involv ed in criminal activities, respectively. From the subsample of respondents reporting an arre st, 8.9% (46) of the sample reported having no friends who smoked, while 37.7% (194) and 53.0% (273) reported having some friends or most friends who smoked, respective ly. From the subsample of respondents reporting an arrest, 23.9% (123) of the sample reported having no friends who used drugs, while 47.2% (243) and 27.2% (140) r eported having some friends or most friends who used drugs, respectively. From the subsample of respondents reporting an arrest, 12.6% (65) of the sample reported having no friends drank alcohol, while
30 48.7% (251) and 37.7% (194) reported having some fr iends or most friends who drank alcohol, respectively. Stand alone risk factors used in this study as inde pendent variables include gender and frequency of arrest. Frequency of arrest is con ceptualized as the number of times an individual was arrested during the duration of the study. The number of arrests for a single respondent varies from one arrest to thirty arrests. Table 1 presents descriptive data on the variables employed in this study. Table 1. Description of the Data Variables Mean Std Dev Minimum Maximum Sex .744 .437 .00 1.00 Mom Age .210 .408 .00 1.00 Mom PubAssist .107 .309 .00 1.00 Mom Educat 2.745 1.084 1.0 0 6.00 Mom MarStat .324 .469 .00 1.00 Mom Income .143 .351 .00 1.00 Fam Con Tact 10.922 9.002 .00 44.00 Sex Abuse .087 .282 .00 1.00 WRAT Spell .249 .433 .00 1.00 WRAT Read .286 .452 .00 1.00 WRAT Arith .227 .419 .00 1.00 Birth Weight .130 .337 .00 1.00 Verbal IQ .219 .414 .00 1.00 Drug Scale 3.422 2.314 1.0 0 10.00 School Dev 2.715 1.115 .00 3.00 Dev Peer Assoc 8.626 2.087 4.00 12.00 Age of Arrest 20.008 4.950 7.00 32.00 Freq of Arrest 3.759 4.168 1.00 30.00 Analytic Plan This study addresses three interrelated questions regarding female onset of delinquency. The first question asks if there is ev idence of early onset of delinquency in females and, if so, at what age this onset occurs a nd if it differs from the age of onset in males. Second, what factors predict early onset of delinquency in females and are these factors similar to or different from those predicti ve of male early onset. Finally, is an
31 early age of onset related to the frequency of arre st in females as it has been found to be related in males? The first research question is answered by determin ing when age of onset, as defined by arrest, occurs. To address this question both the relative frequency and cumulative frequency distribution of arrest prevale nce by age for males and females is examined and compared. It is necessary to determine what constitutes evidence of Â“early onsetÂ” in females. For instance, the presence of a distinct group that onsets at age thirteen or younger as has been observed in males. Failing t hat, evidence of early onset in females could also be provided by the presence of a distinc t group that onsets earlier than most other females though at a different age than what h as been observed for males. The second research question is answered by determi ning if there are gender differences in risk factors predicting onset. To do so, this study follows a three-step process. First, the means and standard deviations o f the various independent variables are compared across different age groups for both males and females. Second, independent ttests are employed to assess whether or not there a re significant differences on these factors both between and within males and females o f different age groupings. The effects of gender and other explanatory factors on age of first arrest are examined in multivariate Ordinary Least Squares (OLS) regressio n models. Then, logistic regression models are run to determine whether risk factors fo r onset are similar or different across combinations of gender and age. In these logistic regression models, the dependent variable, age at first arrest, is dichotomized in t he male model as (0) first arrest at age fourteen or older, and (1) first arrest at age thir teen or younger; and in the female model as (0) first arrest at age sixteen or older, and (1 ) first arrest at age fifteen or younger.
32 To answer the third and final research question, th is study examines whether or not there are gender differences in effect of early onset on the frequency of subsequent arrests using negative binomial regression. Negativ e binomial regression is necessary because the frequency of arrest is a highly skewed, discrete, non-negative integer (Long and Freese, 2003). Separate models are examined for males and females as this allows an examination of the effects of age of onset on the f requency of arrest across genders A test of equality among coefficients (Brame et al., 1998) is used to estimate the statistical significance of any observed differences in coeffic ients between males and females.
33 Chapter Four Results The purpose of the current section is to examine se veral multivariate models that predict age of first arrest, risk factors for arres t, and frequency of arrest. The following analyses speak to my primary research questions: 1) Is there evidence of early onset in females and, if so, what age does this occur and is this age different from males? 2) Are there gender differences in risk factors predicting early onset? 3) Is an early age of onset related to frequency of arrest in females? Ordinary least squares regression was utilized to model the effects of sex and other risk factors on age of first arrest. Logistic regression was applied to model the effects of age and other r isk factors on Â“earlyÂ” onset of arrest by sex. Lastly, negative binomial regression was then employed to predict frequency of arrest for both males and females. It is first necessary to determine if Â“earlyÂ” onset does or does not occur in females, and if so, at what age it occurs in female s relative to males. First this study assesses the proportion of males and females arrest ed by age. As seen in Table 2, 36 (9.5%) males in the sample were initially arrested at age thirteen or younger. Though thirteen is used as an indicator of early onset amo ng males (Farrington, 2000), only one female in the sample was arrested at age thirteen o r younger. Additionally, only five females in the sample were arrested at age fourteen or younger. Fourteen female respondents in the sample indicated an age of first arrest at age 15 or younger. However, similar in proportion to early onset males in these data, 14 (10.9%) females in the sample
34 were initially arrested at age fifteen or younger. It may very well be that there is no Â“earlyÂ” onset among females as has been claimed by others (Silverthorn and Frick, 1999). Conversely, it may also be the case that female Â“ea rlyÂ” onset occurs later than early onset in males. It may be suggested then, that females do indeed have an Â“earlyÂ” onset, yet it occurs later than early onset in males. Furthermor e, age fifteen is chosen as the cut point for Â“earlyÂ” onset in females to also account for th e typical threshold in age that often needs to be passed in order for females to be forma lly processed in the criminal justice system. Table 2. Age at First Arrest 13 14 15 16 17 MALES N 36 52 72 118 259 % 9.5% 13.8% 19.1% 31.3% 68.7% FEMALES N 1 5 14 20 109 % 0.8% 3.9% 10.9% 15.5% 84.5% Since an Â‘earlyÂ’ age of onset in females may have b een discovered, the current study was then able to assess the bivariate correla tion to determine the degree of association between both age of first arrest, and f requency of arrest, and the available risk factors in these data. Results presented in Appendi x H represent the associations of family adversity, child abuse, neuro-cognitive, dru g use, school deviance, deviant peer associations, age of arrest, and frequency of arres t factors for both males and females. Among females, age and frequency of arrest are not significantly correlated with any of the variables. For the males, frequency of arrest (r = -.436, p < .01), WRAT reading score (r = -.104, p < .05), drug use scale (r = -.179, p < .01), school deviance (r = -.163,
35 p < .01), and deviant peer associations (r = -.296, p < .01) all yield significant inverse correlations to age of arrest; whereas, family adve rsity factors, sex abuse, family conflict tactics, and the other neuro-cognitive factors (i.e ., WRAT spelling score, WRAT arithmetic score, birth weight, and verbal IQ) are not. Variables that positively correlate with frequency of arrest in males included their mo therÂ’s marital status (r = .158, p < .01), WRAT reading (r = .243, p < .01), WRAT ar ithmetic (r = .174, p < .01), Verbal IQ (r = .154, p < .01), drug scale (r = .206, p < 01), school deviance (r = .216, p < .01), and deviant peer association (r = .258, p < .01). M other educational level (r = -.218, p < .01) is also correlated with frequency of arres t in males, but in an inverse relationship. The remaining family adversity factors (i.e., mothe rÂ’s age, public assistance, and income), sexual abuse, family conflict tactics, and WRAT spelling score did not significantly correlate with frequency of arrest. Modeling Age of First Arrest To answer the second research question, the effect of gender and other explanatory factors on age of first arrest are exam ined in multivariate Ordinary Least Squares (OLS) regression models (See Table 3). Mode l 1 includes sex; whereas, Models 2 and 3 are regressed by gender. In Model 1, sex ha s a significant effect on age of first arrest (b = -2.889, p < .05). This result indicates that sex is a significant predictor of age of first arrest such that males are more likely to have an earlier onset than females. Interestingly, the child abuse indicator is signifi cant (b = -.326, p < .05), yet indicates that age of first arrest decreases with increased report s of sexual and physical abuse. The effect of deviant peers is significant (b = -.502, p < .05), indicating that age at first arrest decreases as the number of deviant friends increase s. Among males (Model 2), deviant
36 peer association remains significant (b = -.538, p < .05). This result again indicates that a male offenderÂ’s age of first arrest decreases as th eir deviant peer associations increase. Among females (Model 3), none of the effects are si gnificant in predicting age of first arrest. Table 3. Ordinary Least Squares Regression Predicti ng Age of First Arrest Variables Model 1 Model 2 Model 3 b SE b SE B SE Sex -2.195* .560 Family Adversity -.061 .098 -.030 .117 .169 .188 Child Abuse .326* .160 .222 .223 .379 .245 Neuro Cognitive .018 .071 -.028 .083 .110 .1 49 Drug Use -.199 .108 -.201 .119 -.120 .2 75 School Deviance -.307 .218 -.368 .265 -.153 418 Deviant Peers -.502* .125 -.538* .143* -.395 269 Constant 27.507* 1.088 25.795* 1.270 26.034 2.263 R .143 .098 .058 F 9.406* 5.311* .975 p < .05 Comparing Across Means The next series of analyses explore differences in the individual risk factors for early onset of arrest (mother age, mother public as sistance, mother education, mother marital status, family conflict tactics, sexual abu se, WRAT spelling score, WRAT reading score, WRAT arithmetic, birth weight, verba l IQ, drug use scale, school deviance, deviant peer associations, and frequency of arrest) across males and females. Ttests are performed to examine any statistically si gnificant differences in the means of these variables both within and between males and f emales by their age of first arrest grouping (i.e., Â“earlyÂ”, Â“normalÂ”, and Â“adultÂ”). Th ese analyses are reported in Tables 4a and 4b. Table 4a reports the means and standard dev iations, while Table 4b reports only statistically significant differences in these risk factors across the fifteen comparisons.
37 The comparisons relevant to the current study are b etween early onset males (males 13) and Â“earlyÂ” onset females (females 15). A significantly greater proportion of Â“earlyÂ” onset females report a prior experience of sexual abuse than do early onset males (mean = .231 vs .028, respectively; t = -2.37 5, p < .05). Conversely, early onset males report a greater number of delinquent peers t han do Â“earlyÂ” onset females (mean = 10.059 vs 8.500, respectively; t = 2.370, p < .05). That is, males are more likely to report being friends with others who were involv ed in criminal behaviors, drinking, smoking, and drug use. In addition, early onset mal es have a significantly higher mean number of arrests (mean = 7.879 vs 2.071, respectiv ely; t = 3.623, p < .05) than Â“earlyÂ” onset females. Males age thirteen and younger avera ge almost eight arrests, while females age fifteen and younger average slightly ov er two arrests. Though not significant, early onset males had higher mean scor es on the family conflict, school deviance, and drug scales than Â“earlyÂ” onset female s. Comparisons among females reveal that Â“earlyÂ” onset females had higher mean scores r egarding motherÂ’s age and income than Â“normalÂ” onset females (i.e., age 16-17). Ther e were no significant differences among the variables between Â“earlyÂ” and Â“lateÂ” onse t females (i.e., 18). However, Â“lateÂ” onset females reported higher mean scores on the family conflict tactics scale than Â“normalÂ” onset females. Modeling Â“EarlyÂ” Onset of Arrest in Males and Femal es The study now attempts to identify significant risk factors of Â“earlyÂ” onset of arrest for males and females and determine the exte nt, if any, to which these risk factors are common versus gender-specific. To do so Â“early onsetÂ” is regressed onto several key risk factors in separate logistic regression analys es for males and females (See Table 5).
38 Table 4a. Mean Comparisons by Age and Gender MALES FEMALES Early Onset Normal Onset Adult Onse t Early Onset Normal Onse t Adult Onset (Age 13) (Age 14-17) (Age 18) (Age 15) (Age 16-17) (Age 18) Variables Mean StdDev Mean StdDev Mean StdDev Mean StdDev Mean Std Dev Mean StdDev MomAge .278 .454 .186 .391 .197 .399 .1 43 .363 .546 .522 .212 .410 MomPubAssist .167 .378 .100 .301 .080 .27 2 .231 .439 .364 .505 .096 .296 MomEducat 2.639 1.222 2.768 1.115 2.848 1.0 17 2.385 .870 2.182 1.079 2.664 1.137 MomMarStat .417 .500 .288 .455 .289 .454 .308 .480 .455 .522 .385 .489 MonIncome .206 .410 .111 .316 .099 .299 .077 .277 .500 .527 .188 .392 FamConfTact 11.171 11.597 11.001 7.864 9.978 8.097 10.154 8.934 6.636 7.724 13.221 10.708 SexualAbuse .028 .167 .036 .186 .045 .207 .231 .439 .182 .405 .221 .417 WRAT Spell .303 .467 .270 .446 .263 .441 .231 .439 .182 .405 .167 .375 WRAT Read .455 .506 .342 .477 .259 .439 .231 .439 .182 .405 .235 .426 WRAT Arith .424 .502 .198 .400 .237 .426 .231 .439 .364 .405 .157 .365 BirthWeight .114 .323 .132 .341 .132 .3 39 .071 .267 .091 .505 .135 .343 Verbal IQ .343 .482 .232 .424 .217 .413 .286 .469 .272 .302 .164 .372 DrugScale 4.606 2.794 3.869 2.522 3.049 2.153 3. 583 2.575 3.091 .467 3.167 1.923 SchoolDev 2.972 1.183 3.062 .938 2.702 1.078 2 .357 1.151 2.546 2.296 2.330 1.208 DevPeerAssoc 10.059 2.044 9.391 1.945 8.280 1.993 8.500 2.139 8.546 1.375 8.111 2.015 FreqArrest 7.879 5.862 6.518 5.160 2.658 2.673 2 .071 1.686 3.182 1.293 2.077 2.468
39 Table 4b. Significant Differences in Individual Ris k Factors by Gender Male Male Male Female Fe male Female Early Onset Normal O nset Late Onset Ear ly Onset Normal Onset Late Onset Males Early Onset 13 (Early > Normal) WRAT Arith (Early > Late) Dev Peer Assoc WRAT Reading WRAT Arith Freq of Arrest Drug Scale (F > M) Sexual Abuse (M > F) Dev Peer Assoc Freq of Arrest (M > F) Deviant Peer Assoc Frequency of Arrest (F > M) Sexual Abuse (M > F) Deviant Peer Assoc School Deviance Verbal IQ WRAT Reading WRAT Arithmetic Frequency of Arrest Drug Scale Males Normal Onset 14-17 (Normal > Late) Dev Peer Assoc School Deviance Freq of Arrest Drug Scale (F > M) Sexual Abuse (M > F) School Deviance Freq of Arrest (F > M) Sexual Abuse Mom Age Mom Public Assist Mom Income (M > F) Frequency of Arrest (F > M) Sexual Abuse (M > F) Deviant Peer Assoc School Deviance Frequency of Arrest Drug Scale Males Late Onset 18 (F > M) Sexual Abuse (F > M) Sexual Abuse Mom Age Mom Public Assist Mom Income (M > F) Mom Education (F > M) Sexual Abuse Family Conflict Tact Mom Income (M > F) School Deviance WRAT Spelling Females Early Onset 15 (Normal > Early) Mom Age Mom Income Females Normal Onset 16-17 (Normal > Late) Mom Age Mom Public Assist Mom Income (Late > Normal) Family Conflict Tact
40 To determine whether risk factors for onset are sim ilar or different across different age groupings and gender, models are now examined using logistic regression analyses. Because of the related proportions of age and arres t in the current study, as seen in Table2, age thirteen and below is considered early onset in males; whereas, age fifteen and below is considered early onset in females. Thi rty-six male respondents were age thirteen and younger at the time of their first arr est, while 14 female respondents were age fifteen and younger. Table 5 illustrates the re sults of logistic regression predicting first arrest at age thirteen and younger among male s and first arrest at age fifteen and younger among females. Among early onset males, age of first arrest is a f unction of the neuro-cognitive scale (b = .128, Exp(b) = 1.137) and the deviant pe er association scale (b = .434, Exp(b) = 1.544). From these results, it can be determined that a one-unit increase in the neurocognitive scale indicates a 13.7% increase in the o dds of being arrested at or younger than 13 years old among males. Therefore, the standardiz ed neuro-cognitive scale (i.e., birth weight, verbal IQ, WRAT reading score, WRAT spellin g score, and WRAT arithmetic score) is a predictor of first arrest in early onse t males. This result indicates that lower scores on these neuro-cognitive factors is a predic tor of first arrest in early onset males. Additionally, it can be determined that a one-unit increase in the deviant peer association scale indicates a 54.4% increase in the odds of bei ng arrested at or younger than 13 years old among males. This result shows that a higher nu mber of deviant friends is a predictor of first arrest in early onset males. Therefore, th e standardized deviant peer association scale (i.e., friends who are involved in criminal a ctivity, drinking, smoking, and using drugs) is a predictor of first arrest in early onse t males. Furthermore, the male early onset
41 model explains 16.3% of the variance. The same anal yses described above are repeated for the early onset females (i.e., 15). As see in Table 5, Model 2, none of the varia bles are significant. It can be concluded from these analyses that devian t peer association remains a significant predictor of first arrest for males acr oss various multivariate models. Additionally, the neuro-cognitive scale is a predic tor of first arrest in early onset males. Unfortunately, among females, none of the effects w ere significant in predicting age of first arrest. Table 5. Logistic Regression Predicting Early Onset of Arrest by Gender Male Early Onset Female Early Onset ( 13) ( 15) Variable Model 1 Model 2 b SE Exp(b) b SE Exp(b) Family Adversity .120 .083 1.128 -.062 .140 .94 0 Child Abuse -.076 .167 .927 -.053 .174 .948 Neuro Cognitive .128* .062 1.137 .102 .095 1.108 Drug Scale .142 .084 1.153 .169 .178 1.1 84 School Deviance -.177 .133 1.544 .122 .198 1.130 Deviant Peer Assoc .434* .223 .837 .068 .297 1.070 Constant -6.621* 1.321 .001 -4.003* 1.756 .018 23.264* 3.409 Nagelkerke R .163 .069 p < .05 Modeling Â“EarlyÂ” Onset as a Predictor of Frequency of Arrest by Gender To answer the third and final research question, n egative binomial analyses are used to determine whether or not there are gender d ifferences in the effect of early onset on the frequency of subsequent arrests (See Table 5 ). Table 5 includes two models, one for males and the other for females. For males, Â“earlyÂ” onset is, as expected, a signifi cant predictor of the frequency of arrest (b = .461, Model 1); for Â“earlyÂ” onset males the predicted mean number of arrests
42 is 58.6% greater than that for Â“normalÂ” and Â“adultÂ” onset males. The frequency of arrest among males is also a function of the neuro-cogniti ve scale (b = .049), school deviance scale (b = .146), and deviant peer association scal e (b = .087). For every one-unit increase in the neuro-cognitive scale, a male offenderÂ’s pre dicted mean number of arrests increases by 5.1%. This result indicates that early onset males with lower scores on neuro-cognitive factors (i.e., birth weight, verbal IQ, and WRAT scores) have higher predicted mean numbers of arrest than Â“normalÂ” or Â“ lateÂ” onset males. In addition, for every one-unit increase in the school deviance scal e, early onset males increase their predicted mean number of arrests by 15.7%. This res ult indicates that early onset males with higher incidences of school deviance (i.e. mis behavior, fighting, and expulsion/suspension) have higher predicted mean nu mbers of arrest. Finally, every oneunit increase in the deviant peer association scale increases a male offenderÂ’s mean number of predicted arrests by 9.1%, indicating tha t early onset males with more deviant peers have higher predicted mean numbers of arrest. Contrary to the studyÂ’s third hypothesis, Â“earlyÂ” o nset among females is not a significant predictor of frequency of arrest. Howev er, the frequency of arrest in females is a function of the child abuse scale (b = -.110) and the drug use scale (b = .161). Interestingly, a one-unit increase in the child abu se scale decreases a femaleÂ’s mean number of arrests by 10.4%, indicating that females who report higher incidences of child abuse have lower predicted mean numbers of arrest. Additionally, for every one-unit increase in the drug scale, a femaleÂ’s mean number of arrests increases by 17.5%. This result indicates that for Â“earlyÂ” onset females, th e predicted mean number of arrests is 17.5% greater than that for Â“normalÂ” and Â“adultÂ” on set females.
43 Because differences between males and females are o f primary interest in the current study, an analysis is performed using an eq uality of coefficients test (Clogg, Petkova, & Haritou, 1995; Brame et al., 1998). Thes e analyses are performed for every variable, even those showing insignificance in thei r gender-specific models. This test examines the differences between the coefficients a cross males and females. As expected, though insignificant as a predictor of arrest frequ ency in the male-only negative binomial regression analysis, the drug use scale yielded a s ignificant (z 1.96) z-score of 2.556. In addition, though not significant in the female-only model, age of first arrest proved significant (z 1.96) with a z-score of 1.83. These results indica te that both drug use and age of first arrest have significantly stronger eff ects for early onset males than for Â“earlyÂ” onset females. From the negative binomial analyses, it can be concluded that early onset of arrest is a predictor for subsequent arrests in males. However, early onset of arrest is not a predictor of subsequent arrests in females. Table 6. Negative Binomial Regression Predicting Fr equency of Arrest by Gender M ale Early Onset Female Early Ons et ( 13) ( 15) Variable Model 1 Model 2 b SE % b SE % Z Family Adversity .030 .020 3.1 .020 .035 2.0 .248 Child Abuse -.017 .040 -1.6 -.110* .044 -10.4 1.5 6 Neuro Cognitive .049* .014 5.1 .020 .028 2.1 .926 Drug Scale .027 .021 2.8 .161* .048 17.5 2.55 6 Dev Peer Assoc .087* .026 9.1 .059 .048 6. 1 .513 School Deviance .146* .048 15.7 .064 .079 6.6 887 Â“EarlyÂ” Onset .461* .157 58.6 -.161 .301 -14.9 1 .83 Alpha .400* .049 .297 .085 63.21* 20.78* Psuedo R .041 .052 p < .05
44 Chapter 5 Conclusion and Discussion The purposes of this study are threefold; (1) to ex plore gender differences in age of onset of arrest for delinquency, (2) to explore gender differences in the correlates of Â“earlyÂ” onset, and (3) to explore gender difference s in the effects of age of onset on the frequency of arrest using a variety of multivariate analyses. The age of onset of delinquency and antisocial behavior has been found to be a principal indicator for further delinquency as males mature through their life-cour se (Farrington, et al., 1990). However, even though female rates of delinquency ha ve steadily increased over the past several years (McCabe et al., 2002), research on th e age of onset, risk factors for onset, and later criminality have focused almost entirely on male populations (Piquero and Chung, 2001). Because most believe that female deli nquency is only slightly varied from male delinquency, if at all, females have been gene rally left out of delinquency studies (Hoyt and Scherer, 1998). However, male and female populations are not necessarily similar and therefore, results from male studies of delinquency can not be generalized to females. Based on this, the current study focuses o n three interrelated questions regarding female Â“earlyÂ” onset, gender differences in risk fa ctors of onset, and the effects of age of onset as it relates to frequency of arrest in femal es. The data utilized to examine these research questio ns included both prospective and retrospective data collected during 1960-1994 o n inner-city families from Baltimore, Maryland. Of the original sample, only 1,758 childr en born during 1960-1965 completed
45 assessments through age 7 and/or 8, and were theref ore eligible for follow-up assessments at age 30-35. Of the respondents who co mpleted the final assessments, 515 reported the age at which they were first booked, c harged, or arrested by an authority of the law. It is these 515 respondents that are inclu ded in the following bivariate and multivariate model summaries. It was first predicted that females would display a n Â“earlyÂ” onset age of delinquency, yet at a later age from that of early male onset. An analysis assessing the proportion and corresponding age of first arrest fo r males and females suggested that females do indeed have an Â“earlyÂ” onset (See Table 2), yet one that occurs later than early onset in males. From this analysis, age thirt een and below was considered early onset in males; whereas, age fifteen and below was considered Â“earlyÂ” onset in females. Because an Â“earlyÂ” age of onset in females was disc overed from the proportional analysis (See Table 2), it was then necessary to mo ve on to the second research question. It was hypothesized that risk factors predicting ea rly onset of arrest would vary across males and females. Bivariate correlations (See Appe ndix A) were then evaluated to determine the degree of association between both ag e of first arrest and frequency of arrest, and the available risk factors in these dat a. When exploring variable correlations among females, age of first arrest and frequency of arrest were not significantly correlated with any of the predictor variables. In support of the current body of literature regard ing early onset (Moffitt, 1993; Loeber et al., 1998), OLS regression analyses provi ded that sex, child abuse, and deviant peer associations all had significant effects when predicting age of first arrest. As expected, there was a unique effect of sex on age o f fist arrest, even when controlling for
46 all other factors. Males were also more likely than females to have an earlier age of first arrest. Additionally, as expected, deviant peers pr ovided a significant effect in predicting first arrest in males. Among females, none of the e ffects were significant in predicting age of first arrest. Though the first multivariate analyses did not supp ort the second hypothesis, t-tests were then performed to examine any statisti cally significant differences in the means of the risk factors both within and between g ender. In support of the current body of literature regarding early onset (Moffitt, 1993; Loeber et al., 1998; Storvoll and Wichstrom, 2002; Werner and Silbereisen, 2003), the se t-tests indicated that a significantly greater proportion of early onset fem ales report a prior experience of child abuse (i.e sexual or physical) than do early onset males; whereas, early onset males report significantly greater numbers of deviant peers than do Â“earlyÂ” onset females. Additionally, as expected, early onset males have a significantly higher number of arrests than do Â“earlyÂ” onset females. The data were again separated by gender to identify significant risk factors of Â“earlyÂ” onset of arrest for males and females and t o determine the extent, if any, to which these factors are common versus gender-specific. Lo gistic regression analyses provided that neuro-cognitive factors and deviant peer assoc iations were significant predictors of early onset in males (i.e., first arrest 13). These results are supported by extant literat ure regarding neuro-cognitive factors as early onset pr edictors of delinquency in males (Farrington and Hawkins, 1991; Moffitt, 1993; Raine et al., 1997), as well as deviant peer associations (Loeber et al., 1998). Conversely, sup port was not found for the second hypothesis regarding risk factors predicting Â“early Â” onset in females (i.e., first arrest
47 15). Therefore, though anticipated indicators prove d significant for the males when predicting first arrest 13, none of the predictors used in the model to pr edict Â“earlyÂ” onset were significant for the females when predict ing first arrest 15. Though significant results were not found in the lo gistic regression analysis among females, multivariate analyses were then empl oyed to answer the third and final research question regarding female Â“earlyÂ” onset as a predictor of frequency of arrest. Among males, negative binomial regression analysis supports the current body of lifecourse literature. Results provide that an early ag e of onset, neuro-cognitive factors, and deviant peer associations are significant predictor s of frequency of arrest in males (Moffitt, 1993; Piquero and Chung, 2001; Werner and Silbereisen, 2003). However, the results of the current study provide no support for the hypothesis that an Â“earlyÂ” age of onset is related to frequency of arrest in females. Negative binomial regression analysis found that fo r every one-unit increase in the drug use scale, a femaleÂ’s mean number of predi cted arrests increased by 17.5%. This result indicates that Â“earlyÂ” onset girls who engag e in higher amounts of drug use increase their predicted mean number of arrests by 17.5% more than do Â“normalÂ” and Â“lateÂ” onset females. Interestingly, the current bo dy of literature regarding early onset and risk factors frequently cites drug use as an indica tor for males, but not so much for females. However, in support of this finding, Broid y and Agnew (1997) suggest that females are more likely than males to engage in sel f-destructive forms of behavior such as drug use. Another interesting finding regarding the effect of child abuse on frequency of arrest resulted in an opposite effect of what wa s hypothesized. Contrary to prior research (Widom, 1989; Fergusson and Woodward, 2000 ; Herrera and McCloskey,
48 2001), child abuse had a significant negative effec t when predicting frequency of arrest in females. That is, the more sexual and/or physical a buse a female reported, the lower her reported frequency of arrest. A possible explanation for this inconsistent findin g involves adjustment problems following sexual and physical abuse. Though the lik elihood of adverse effects on a childÂ’s emotional and behavioral development due to child abuse has been well documented, the types of processes children use in adjustment following their abuse is quite varied (Wolfe, Gentile & Wolfe, 1989). In add ition to externalizing behaviors such as delinquency, internalizing problems such as with drawal, depression, and post traumatic stress disorder are also common in childr en following abuse (Allen & Tarnowksi, 1989; Wolfe et al., 1989; Cerezo & Frias 1994; Johnson and Kenkel, 1991; Wolfe, Sas, Wekerle, 1994; Spacccarelli & King, 199 5; Feiring, Taska, & Lewis, 1998). Moreover, females are more likely than males to int ernalize problems (as opposed to externalize), in general. Therefore, it might be su ggested that the females, age fifteen and younger, in the current study dealt with their abus e by means of internalization, rather than external displays of delinquency. The three interrelated questions regarding female o nset of delinquency in the current study include (1) whether or not there is e vidence of Â“earlyÂ” onset of delinquency in females and, if so, at what age this occurs and if this age differs from males; (2) if there are similar or different factors that predict early onset in males and females; and, (3) if an Â“earlyÂ” age of onset is related to frequency of arrest in females. Early onset for among females was not defined at age 13 as it has been done in males, but rather, at ag e 15. However, even at age 15, no risk factors predicted Â“earlyÂ” onset among females.
49 Additionally, Â“earlyÂ” onset as defined in this stud y had no predictive value itself among females. The above conditions are both necessary wi thin the criminological literature for Â“earlyÂ” onset to be meaningful. Several implications can be drawn from the above re sults. The variables in the current study were not significant predictors of Â“e arlyÂ” onset (as measured by arrest) in females. Furthermore, age of Â“earlyÂ” onset was not a significant predictor of subsequent arrests in females. As stated above, both of these conditions are essential in order for the term Â“earlyÂ” onset to be applied to females as it h as already been applied to males. Research has found certain predictor variables to l ower the age of antisocial and delinquent behavior in males (Moffitt, 1993; Loeber et al. 1998; Stattin and Magnusson, 2001). Additionally, research has also found that a n early age of onset regarding these behaviors is a primary indicator for further antiso cial behavior, delinquency, and subsequent criminality (Nagin and Farrington, 1992; Sampson and Laub, 1993). If these conditions are not met (i.e. variables that predict Â“earlyÂ” onset and early onset predicting future delinquency and offending), the term Â“earlyÂ” onset can not be applied to female populations. AgnewÂ’s Strain theory can possibly be used to expla in gendered responses regarding coping mechanisms in males and females. E motional responses to strain differ across males and females. Broidy and Agnew (1997) g o on to suggest that both males and females experience various forms of anger in respon se to strain. However, males are more likely than females to engage in various forms of crime (i.e., violent crime) in response to strain (Piquero and Sealock, 2004). Res earchers propose that this response in males is the result of differences in a number of f actors which include coping
50 mechanisms, social support, and social controls (Gi ordana, Cernkovich, & Pugh, 1986; Steffensmeier and Allan, 1996; Broidy and Agnew, 19 97). On the other hand, females are more likely to respond to strain with internalizing reactions such as depression and anxiety (Leadbetter, Blatt, and Quinlan, 1995; Broi dy, 2001; Cyranowski, Frank, Young, & Shear, 2000). While this study has shed some light on an under-re searched area within criminology, there were some limitations that shoul d be noted. First, self-reported arrest data were used in the current study. Measures of ar rest may be compromised due to possible police bias and/or discrimination. There a re multiple non-legal factors than can influence an officerÂ’s decision regarding whether o r not to make an arrest including suspectsÂ’ age and gender (Sealock & Simpson, 1998). Feyerherm (1980) suggests that females are less likely to be arrested than males. However, girls are more likely to be arrested for minor property offenses and status off enses than their male counterparts (Chesney-Lind & Sheldon, 1992). One study found tha t seriousness of police contact, the number of prior contacts with police, and whether o r not the officer actually witnessed the offense had positive significant effects on fem ale arrest (Sealock & Simpson, 1998). Additionally, prior police contact was more importa nt in the arrest decision for females than for males. Other studies indicate that females are less likely to be arrested if they show signs of fear or regret when in contact with an officer, rather than signs of noncompliance or insubordination (Visher, 1983). Visher explains, ho wever, that this type of police conduct is more often offered to white females. He goes on to suggest that black females would have higher levels of arrest because they may not d isplay similar behaviors as the white
51 females when dealing with a mostly male, white poli ce-force. Harris (1977) also suggests that poorer females will be arrested at higher leve ls than middle-class females. Though the current study does not include race as an indic ator, it is important to note the majority of the mothers (77%) enrolled in the JHCPS, and thu s, their children, are AfricanAmerican. Furthermore, using arrest as a measure of Â“earlyÂ” o nset during the 1970s further limits its use as a dependent variable due to a per iod effect. Such a period effect is a likely artifact of events that were going on when t he study participants were entering their adolescent ages. The 1970s were a time when the juv enile justice system was taking steps to decriminalize status offenses, deinstitutionaliz e runaways, and divert both female and male non-serious offenders (Holden and Kapler, 1995 ). As such, female adolescents during this period may have experienced a particula r advantage relative to males from this movement to divert, deinstitutionalize, and de criminalize the juvenile and status offenders in the form of especially low probabiliti es of arrest. As a result of this period effect, the data in the current study are likely to be artificially censored by the non-arrest of female offenders such that the current studyÂ’s m easure of Â“earlyÂ” onset is not an accurate representation of their true onset of deli nquency. Another limitation related to operationalization is that the concept of Â“earlyÂ” onset and what exactly is meant by that term is not a con crete issue. Reasons for determining when early onset occurs, to whom those who onset ea rly are being compared, and who makes this distinction and why, are all questions t hat criminal career researchers find difficult to answer. Some studies seem to distingui sh what determines early onset solely through statistical procedures in order to make dat a analyses more convenient. This
52 however, is not necessarily a correct representatio n of reality. Other studies, bound by methodological limitations, define Â“earlyÂ” onset ac cording to the age of participants when the study began. For example, if the youngest parti cipants in a study were 12-years old, the researcher might define early onset as any part icipant who indicated at the first data collection point that they had already been arreste d. Regardless of the rationales provided, early-onset research is still considered relatively novel, and therefore, will continue to formulate this definition of Â“earlyÂ” on set more concretely as research continues in this field. In regard to other sampling limitations, the follow -up study data were collected retrospectively. Survey questions probed respondent s about events in their childhood and adolescence. Respondents were expected to accuratel y remember past events. To the extent respondents had difficulty remembering speci fic events that had occurred years earlier, the quality and validity of the data is co mpromised. Consequently, the ability to make causal inferences and generalizations is hinde red because of this methodological issue. Furthermore, the large standard errors in th e multivariate analyses are very likely a function of the small cell sizes. As a result of th e small sample size of those with an Â“earlyÂ” onset, there is an increased probability of Type II error. This limitation could be alleviated with a larger sample of adolescents who exhibit Â“earlyÂ” onset behaviors. Lastly, the results of the current analysis are not generalizable to all segments of the population. A majority of the mothers enrolled in the JHCPS were African-American, generally poor, and indicated low educational attai nment levels. Even though the data were collected on a large sample of mothers and inc luded over 1,700 youth in the final assessment, it was not a random sample. Nevertheles s, there was sufficient variability in
53 the data, and the high-risk sample is consistent wi th populations that are of particular interests to criminologists. In an effort to address some of these limitations, future research should concentrate on several issues. First, future resear ch is needed to provide a more sound and defensible operationalization of early onset. C riminologists have found it especially challenging to delineate an acceptable definition o f Â“earlyÂ” onset. If Â“earlyÂ” onset can not be concretely defined, it can not therefore, be con cretely determined whether or not delinquent and/or criminal behavior falls into this group. While theoretical advances might be able to elucidate a justifiable age that c onstitutes early, it seems more reasonable to rely on empirical data. Perhaps stati stical deviations could provide a reference, as was done in the current study. Howeve r, future researchers should consider whether other statistical deviations might be more reasonable (Raine, 1993). Furthermore, females make up a smaller grouping of detained and adjudicated juvenile delinquents compared to males, yet female rates of delinquency have steadily increased over the past several years (McCabe et al ., 2002). Increases in female arrests markedly surpassed those of males for most of the l ast decade (Bureau of Justice Statistics, 1999). It is necessary to better unders tand why female rates of delinquency are increasing at their current rates. Criminologists a cknowledge that there is insufficient research and data to fully understand these issues relating to female early onset (Pajer, 1998), yet reasons for the increases of female deli nquency and criminality still elude researchers. There is an absolute need to better un derstand early onset and the risk factors for delinquency and antisocial behavior in females. To do so, the range of behaviors needs to extend beyond just those behaviors that ha ve been found to be typical problem
54 indicators for males and include indicators includi ng covert forms of aggressive behavior and other internalizing problems such as withdrawal depression, and post traumatic stress disorder.
55 References Allen, D.M., & Tarnowski, K.J. (1989). Depressive c haracteristics of physically abused children. Journal of Abnormal Child Psychology, 17, 1-11. Bartusch, D. R. J., Lynam, D.R., Moffitt, T.E., & S ilva, P.A. (1997). Is age important? Testing a general versus a developmental theory of antisocial behavior. Criminology, 35 (1), 13-48. Benson, M.L. (2002). Crime and the life course: an introduction. Los Angeles, California: Roxbury Publishing Company. Blumstein, A., & Cohen, J. (1979). Estimation of in dividual crime rates from arrest records. Journal of Criminal Law and Criminology, 70, 561-585. Blumstein, A., Cohen, J., Roth, J.A., & Visher, C.A (Eds.). (1986). Criminal careers and Â‘career criminals.Â’ WASHINGTON, D.C: National Academy Press. Brame, R., Paternoster, R., Mazerolle, P., & Piquer o, A. (1998). Testing for the equality of maximum likelihood regression coefficients betwe en two independent samples. Journal of Quantitative Criminology, 14, 245-262. Broidy, L. (2001). A test of general strain theory. Criminology, 39, 9-35. Broidy, L., & Agnew, R. (1997). Gender and crime: A general strain theory perspective. Journal of Research in Crime and Delinquency, 34 (3), 275-306. Carpendale, J. I. M. (2000). Kohlberg and Piaget on stages and moral reasoning. Developmental Review, 20 (2), 181-205. Caspi, A., Elder, G.H. Jr., Bern, D.J. (1987). Movi ng against the world: life-course patterns of explosive children. Developmental Psychology, 23 (2), 308-313. Cerezo, M.A., & Frias, D. (1994). Emotional and cog nitive adjustment in abused children. Child Abuse and Neglect, 18, 923-932. Chesney-Lind, M., & Okamoto, S.K. (2001). Gender ma tters: patterns in girlsÂ’ delinquency and gender responsive programming. Journal of Forensic Psychology Practice, 1 (3), 1-28. Chesney-Lind, M., & Shelden, R.G. (1992). Girls, delinquency, and juvenile justice. Pacific Grove, CA: Brooks/Cole.
56 Clogg, C., Petkova, E., & Haritou, A. (1995). Stati stical methods for comparing regression coefficients between models. American Journal of Sociology, 100, 1261-1293. Cote, S., Zoccolillo, M., Tremblay, R.E., Nagin, D. & Vitaro, F. (2001). Predicting girlsÂ’ conduct disorder in adolescence from childhood traj ectories of disruptive behaviors. Journal of the American Academy of Child Adolescent Psychiatry, 40 (6), 678-684. Cyranowski, J.M., Frank, E., Young, E., & Shear, M. K. (2000). Adolescent onset of the gender difference in lifetime rates of major depres sion. Archives of General Psychiatry, 57, 21-27. DiNapoli, P.P. (2003). Guns and dolls: An explorati on of violent behavior in girls. Advances in Nursing Science, 46 (2), 140-148. Elder, G.H., Jr. (Ed.). (1985). Life course dynamics. New York: Cornell University Press. Elder, G.H., Jr. (1994). Time, human agency, and so cial change. Social Psychology Quarterly, 57 (1), 4-15. Elder, G.H., Jr. (1998). The life-course as develop mental theory. Child Development, 69 (1), 1-12. Elliot, D.S., Huizinga, D., & Ageton, S.S. (1985). Explaining delinquency and drug use. Beverly Hills, CA: Sage. Farrington, D.P. (1986). Age and crime. In M. Tonry & N. Morris (Eds), Crime and Justice Review (pp. 29-90). Chicago: University of Chicago Press Farrington, D.P. (1989). Early predictors of adoles cent aggression and adult violence. Violence and Victims, 4, 79-100. Farrington, D.P. (1990). Implication of criminal ca reer research for the prevention of offending. Journal of Adolescence, 13 (2), 93-113. Farrington, D.P. (1992) Explaining the beginning, p rogress, and ending of antisocial behavior from birth to adulthood. In J. McCord (Ed. ), Facts, Frameworks, and Forecasts. Advances in Criminological Theory, Vol. 3 (pp. 253 -286). New Brunswick, NJ: Transaction Publishers. Farrington, D.P. (2003). Developmental and life-cou rse criminology: key theoretical and empirical issues Â– The 2002 Sutherland Award Ad dress. Criminology, 41 (2), 221-255.
57 Farrington, D.P., & Hawkins, J.D. (1991). Predictin g participation, early onset, and later persistence in officially recorded offending. Criminal Behavior and Mental Health, 1 (1), 1-33. Farrington, D.P., Loeber, R., & Van Kammen, W.B. (1 990). Long-term outcomes of hyperactivity-impulsivity-attention deficit and con duct problems in childhood. In L.N. Robins and M. Rutter (Eds.), Straight ad devious pathways from childhood to adulthood (pp. 62-81). New York: Cambridge University Press. Feiring, C., Taska, L., & Lewis, M. (1998). The rol e of shame and attributional style in chikldrenÂ’s and adolescentsÂ’ adaptation to sexual a buse. Child Maltreatment, 3, 129-142. Fergusson, D.M., & Woodward, L.J. (2000). Education al, psychosocial, and sexual outcomes in girls with conduct problems in early ad olescence. Journal of Child Psychology and Psychiatry, 41 (6), 779-792. Feyerherm, W. (1980). Gender differences in delinqu ency: quantity and quality. In L.H. Bowker (Ed.), Women and Crime in America (pp. 82-92). New York: Macmillan. Gibson, C.L., Piquero, A.R., & Tibbetts, S.G. (2000 ). Assessing the relationship between maternal cigarette smoking during pregnancy and age at first police contact. Justice Quarterly, 17 (3), 519-542. Gibson, C.L., Piquero, A.R., & Tibbetts, S.G. (2001 ). The contribution of family adversity and verbal IQ to criminal behavior. International Journal of Offender Therapy and Comparative Criminology, 45 (5), 574-592. Gibson, C.L., & Tibbetts, S.G. (2000). A biosocial interaction in predicting early onset of offending. Psychological Reports, 86, 509-518. Giordano, P., Cernkovich, S.A., & Pugh, M.D. (1986) Friendships and delinquency. American Journal of Sociology, 91 (5), 1170-1202. Glueck, S., & Glueck, E.T. (1990). Unraveling juvenile delinquency Cambridge, MA: Harvard University Press. Gottfredson, M., & Hirschi, T. (1987). The methodol ogical adequacy of longitudinal research on crime. Criminology, 25 (3), 581-614. Hanlon, T.E., Bateman, R.W., Simon, B.D., OÂ’Grady, K.E., & Carswell, S.B. (2004). Antecedents and correlates of deviant activity in u rban youth manifesting behavioral problems. The Journal of Primary Prevention, 24 (3), 285-309. Harris, A.R. (1977). Sex and theories of deviance: toward a functional theory of deviant typescripts. American Sociological Review, 42, 3-16.
58 Herrera, V. M., & McCloskey, L. A. (2001). Gender d ifferences in the risk for delinquency among youth exposed to family violence. Child Abuse and Neglect, 25, 1037-1051. Hirschi, T., & Gottfredson, M. (1993). Age and the explanation of crime. American Journal of Sociology, 89 (3), 522-584. Hogan, John D. ; In: Encyclopedia of Psychology, Vol. 3. pp. 9-13. Washi ngton, DC: American Psychological Association, 2000. Holden, G.A., & Kapler, R.A. (1995). Deinstitutiona lizing status offenders: a record of progress. Journal of the Office of Juvenile Justice and Delin quency Prevention, 2 (2), 3-11. Hoyt, S., & Scherer, D.G. (1998). Female juvenile d elinquency: misunderstood by the juvenile justice system, neglected by social scienc e. Law and Human Behavior, 22 (1), 81-107. Hipwell, A. E., Loeber, R., Stouthamer-Loeber, M., Kennan, K., White, H. R., & Kroneman, L. (2002). Characteristics of girls wit h early onset disruptive and antisocial behavior. Criminal Behavior and Mental Health, 12, 99-118. Huizinga, D., Weiher, A.W., Espiritu, R., & Esbense n, F. (2003). Delinquency and crime: some highlights from the Denver Youth Study. In T.P Thornberry & M.D. Krohn (Eds.), Taking Stock of Delinquency: An Overview of Finding s from Contemporary Longitudinal Studies (pp. 47-91). New York: Kluwer/Penum. Jeglum Bartusch, D.R., Lynam, D.R., Moffitt, T.E., & Silva, P.A. (1997). Criminology, 35 (1), 13-48. Jenkins, S.M., Buboltz, W.C. Jr., & Schwartz, J.P. (2005). Differentiation of self and psychosocial development. Contemporary Family Therapy: An International Journal, 27 (2), 251-261. Johnson, B.K., & Kenkel, M.B. (1991). Stress, copin g, and adjustment in female adolescent incest victims. Child Abuse and Neglect, 15, 293-305. Junger Tas, J., & Marshall, H.I. (1999). The self-r eport methodology in crime research. In M. Tonry (Ed.), Crime and Justice: A Review of Research (pp. 291-367). Chicago, IL: University of Chicago Press. Keenan, K., & Shaw, D. (1997). Developmental and so cial influences on young girlsÂ’ early problem behavior. Psychological Bulletin, 121 (1) 95-113.
59 Lahey, B.B., Moffitt, T.E., & Caspi, A. (Eds.). (20 03). Causes of conduct disorder and juvenile delinquency. New York: The Guilford Press. Laub, J.H. (2004). The Life-Course of Criminology i n the United States: The American Society of Criminology 2003 Presidential Address. Criminology, 42 (1), 1-21. Leadbeater, B.J., Blatt, S.J., Quinlan, D.M. (1995) Gender-linked vulnerability to depressive symptoms, stress, and problem behaviors in adolescents. Journal of Research on Adolescence, 5, 1-29. Le Blanc, M., & Loeber, R. (1998). Developmental cr iminology updated. In M. Tonry (Ed.), Crime and Justice, Vol. 23 (pp. 115-198). Chicago: University of Chica go Press. Leve, L. D., & Chamberlain, P. (2004). Female juven ile offenders: defining an earlyonset pathway for delinquency. Journal of Child and Family Studies, 13 (4), 439452. Lewis, D.O., Yeager, C.A., Cobham-Portorreal, C.S., Klein, N., Showalter, C., & Anthony, A. (1991). A follow-up of female delinquen ts: Maternal contributions to the perpetuation of deviance. Journal of American Academy of Child and Adolescent Psychiatry, 30 (2), 197-201. Loeber, R., & Faqrrington, D.P. (1999). Serious and violent juvenile offenders: Risk factors and successful interventions Thousand Oaks, CA: Sage. Loeber, R., & Farrington, D.P. (2000). Young childr en who commit crime: Epidemiology, developmental origins, risk factors, early interventions, and policy implications. Development and Psychopathology, 12, 737-762. Loeber, R., Farrington, D.P., Stouthamer-Loeber, M. Moffitt, T.E., & Caspi, A. (1998). The development of male offending: Key findings fro m the first decade of the Pittsburgh Youth Study. Studies on Crime and Crime Prevention, 7 (2), 141-171. Loeber, R., Farrington, D.P., Stouthamer-Loeber, M. Moffitt, T.E., Caspi, A., White, A. W., Wei, E.H., & Beyers, J.M. (2003). The developme nt of male offending: Key findings from form fourteen years of the Pittsburgh Youth Study. In T.P. Thornberry and M.D. Krohn (Eds.), Taking Stock of Delinquency: An Overview of Findings from Contemporary Longitudinal Studies (pp. 93-136). New York: Kluwer/Plenum. Loeber, R., & LeBlanc, M. (1990). Toward a developm ental criminology. ? 375-473. Loeber, R., Wung, P., Keenan, K., Giroux, B., Stout hamer-Loeber, M., & Van Kammen, W.B. (1993). Developmental pathways in disruptive c hild behavior. Development and Psychopathology, 5, 101-132.
60 Long, J., & Freese, S. (2003). Regression models for categorical dependent variabl es using STATA. College Station, TX : STATA Press Mason, W.A., & Windle, M. (2001). Delinquency risk as a function of number of early onset problem behaviors. International Journal of Offender Therapy and Comparative Criminology, 45 (4), 436-448. Mazerolle, P., Brame, R., Paternoster, R., Piquero, A., & Dean, C. (2000). Onset age, persistence, and offending versatility: comparisons across gender. Criminology, 38 (4), 1143-1172. McCabe, K. M., Lansing, A. E., Garland, A., & Hough R. (2002). Gender differences in psychopathology, functional impairme nt, and familial risk factors among adjudicated delinquents. Journal of American Academy of Child and Adolescent Psychiatry, 41 (7), 860-867. Moffitt, T.E. (1993). Adolescent-limited and life-c ourse persistent antisocial behavior: A developmental taxonomy. Psychological Review, 100, 674-701. Moffitt, T.E. (2001). Adolescent-limited and life-c ourse persistent antisocial behavior. In A. Piquero & P. Mazerolle (Eds.), Life-course criminology: Contemporary and classic readings. Samford, CT: Wadsworth/Thomson Learning. Moffitt, T.E., & Caspi, A. (2001). Childhood predic tors differentiate life-course persistent and adolescent-limited antisocial pathwa ys among males and females. Development and Psychopathology, 13, 355-375. Moffitt, T.E., Caspi, A., Dickson, N., Silva, P., & Stanton, W. (1996). Childhood-onset versus adolescent-onset antisocial conduct problems in males: Natural history from ages 3 to 18 years. Development and Psychopathology, 8, 399-424. Moffitt, T. E., Caspi, A., Harrington, H., Milne, B J. (2002). Males on the lifecourse persistent and adolescence-limited antisocia l pathways: follow up at age 26. Developmental Psychopathology, 14, 179-207. Moffitt, T.E., Caspi, A., Rutter, M., & Silva, P.A. (2001). Sex differences in antisocial behavior. United Kingdom: Cambridge University Press. Moffitt, T.E., Lynam, D.R., & Silva, P.A. (1994). N europsychological tests predicting persistent male delinquency. Criminology, 32 (2), 277-289. Nagin, D.S., & Farrington, D.P. (1992). The onset a nd persistence of offending. Criminology, 30 111-140.
61 Nagin, D.S., Farrington, D.P., & Moffitt, T.E. (199 5). Life-course trajectories of different types of offenders. Criminology, 33 (1), 111-124. Pajer, K. A. (1998). What happens to Â“badÂ” girls? A review of the adult outcomes of antisocial adolescent girls. American Journal of Psychiatry, 155 (7), 862-870. Patterson, G.R., DeBaryshe, B.D., & Ramsey, E. (198 9). A developmental perspective on antisocial behavior. American Psychologist, 44 (2), 329-335. Patterson, G.R., Forgatch, M.S., Yoerger, K.L., & M iller, M.S. (1998). Variables that initiate and maintain an early-onset trajectory for juvenile offending. Development and Psychopathology, 10, 531-547. Piquero, A.R. (2000). Assessing the relationships b etween gender, chronicity, seriousness, and offense skewness in criminal offen ding. Journal of Criminal Justice, 28, 103-115. Piquero, A.R. (2001). Testing MoffittÂ’s neuropsycho logical variation hypothesis for the prediction if life-course persistent offending. Psychology, Crime & Law, 7, 193215. Piquero, A.R. & Brezina, T. (2001). Testing Moffitt Â’s account of adolescence-limited delinquency. Criminology, 39 (2), 353-370. Piquero, A. R., & Chung, H. L. (2001). On the relat ionships between gender, early onset, and the seriousness of offending. Journal of Criminal Justice, 29, 189-206. Piquero, A.R., & Mazerolle, P. (Eds.). (2001). Life course criminology: contemporary and classic readings. Belmont, CA: Wadsworth. Piquero, A.R., & Tibbetts, S. (1999). The impact of pre/perinatal disturbances and disadvantaged familial environment in predicting cr iminal offending. Studies on Crime and Crime Prevention, 8 (1), 52-70. Piquero, N.L., & Sealock, M.D. (2004). Gender and g eneral strain theory: A preliminary test of Broidy and AgnewÂ’s gender/GST hypotheses. Justice Quarterly, 21 (1), 125-158. Preski, S., & Shelton, D. (2001). The role of conte xtual, child, and parent factors in predicting criminal outcomes in adolescence. Issue in Mental Health Nursing, 22, 197-205. Raine, A. (1993). The psychopathology of crime: criminal behavior as a clinical disorder. San Diego, CA: Academic Press.
62 Raine, A., Brennan, P., & Mednick, S.A. (1997). Int eraction between birth complications and early maternal rejection in predisposing indivi duals to adult violence: Specificity to serious, early-onset violence. American Journal of Psychiatry, 154 (9), 1265-1271. Richters, J.E., & Cicchetti, D. (1993). Toward a de velopmental perspective on conduct disorder. Development and Psychopathology, 5, 1-4. Robins, L.N. (1978). Sturdy childhood predictors of adult antisocial behavior: replications form longitudinal studies. Psychological Medicine, 8, 611-622. Ryder, N.B. (1965). The cohort as a concept in the study of social change. American Sociological Review, 30, 843-861. Sampson, R.J., & Laub, J.H. (1990). Crime and devia nce over the life course. American Sociological Review, 55 (5), 609-27. Sampson, R.J., & Laub, J.H. (1992). Crime and devia nce in the life course. Annual Review of Sociology, 18, 63-84. Sampson, R.J., & Laub, J.H. (1993) Crime in the making: Pathways and turning points through life. Cambridge, MA: Harvard University Press. Sealock, M.D., & Simpson, S.S. (1998). Unraveling b ias in arrest decisions: the role of juvenile offender type-scripts. Justice Quarterly, 15 (3), 427-257. Silverthorn, P., & Frick, P.J. (1999). Developmenta l pathways to antisocial behavior: The delayed-onset pathway in girls. Development and Psychopathology, 11, 101-126. Spaccarelli, S. (1995). Measuring abuse stress and negative cognitive appraisals in child sexual abuse: validity data on two new scales. Journal of Abnormal Child Psychology, 23, 703-727. Stattin, H., & Magnusson, D. (1995). Onset of offic ial delinquency: its co-occurrence in time with educational, behavioral, and interpersona l problems. British Journal of Criminology, 35 (3), 417-449. Steffensmeier, D. & Allan, E. (1996). Gender and cr ime: toward a gendered theory of female offending. Annual Review of Sociology, 22, 459-487. Storvoll, E. E., & Wichstrom, L. (2002). Do the ris k factors associated with conduct problems in adolescents vary according to gender? Journal of Adolescence, 25, 183-202.
63 Taylor, J., Iacono, W.G., & McGue, M. (2000). Evide nce for a genetic etiology of earlyonset delinquency. Journal of Abnormal Psychology, 109 (4), 634-643. Tibbetts, S. G., & Piquero, A. R. (1999). The influ ence of gender, low birth weight, and disadvantaged environment in predicting early onset of offending: a test of MoffittÂ’s interactional hypothesis. Criminology, 37 (4), 843-877. Tiet, Q. Q., Wasserman, G. A., Loeber, R., McReynol ds, L. S., & Miller, L. S. (2001). Developmental and sex differences in types of conduct problems. Journal of Child and Family Studies, 10 (2), 181-197. Thornberry, T.P. (1997). Introduction: some advanta ges of developmental and life-course perspectives for the study of crime and delinquency In T.P. Thornberry (Ed.), Developmental Theories of Crime and De3linquency (pp. 1-10). New Brunswick, NJ: Transaction Publishers. Thornberry, T.P., Lizotte, A.J., Krohn, M.D., Smith D.A., & Porter, P.K. (2003). Causes and consequences of delinquency: findings from the Rochester Youth Development study. In T.P. Thornberry and M.D. Kroh n (Eds.), Taking Stock in Delinquency: An Overview of Findings from Contempor ary Longitudinal Studies (pp. 11-46). New York: Kluwer/Plunem. Tremblay, R.E., Phil, R.O., Vitaro, F., & Dobkin, P .L. (1994). Predicting early onset of male antisocial behavior from preschool behavior. Archives of General Psychiatry, 51 (9), 732-739. Tremblay, R.E., Masse, L.C., Vitaro, F., & Dobkin, P.L. (1995). The impact of friendsÂ’ deviant behavior on early onset of delinquency: lon gitudinal data from 6 to 13 years of age. Development and Psychopathology, 7, 649-667. Tremblay, R.E., Vitaro, F., Nagin, D., Pagani, L., & Seguin, J.R. (2003). The Montreal Longitudinal and Experimental study: rediscovering the power of descriptions. In T.P. Thornberry and M.D. Krohn (Eds.), Taking Stock in Delinquency: An Overview of Findings from Contemporary Longitudinal Studies (pp. 205-254). New York: Kluwer-Plenum. U.S. Bureau of Justice Statistics (1999). Criminal offending statistics. U.S. Departments of Justice, Office of Justice Programs: Washington, D.C. Retrieved April 9, 2006. Visher, C.A. (1983). Gender, police arrest decision s, and notions of chivalry. Criminology, 21, 5-28.
64 Volkart, E.H. (Ed.), Social Behavior and Personality: Contributions of W .I. Thomas to Theory and Research. New York: Social Science Research Council. Werner, N.E., & Silbereisen, R. K. (2003). Family r elationship quality and contact with deviant peers as predictors of adolescent problem b ehaviors: The moderating role of gender. Journal of Adolescent Research, 18 (5), 454-480. Widom, C.S. (1989). The cycle of violence. Science, 244, 160-166. Wolfe, D.A., Sas, L., & Wekerle, C. (1994). Factors associated with the development of posttraumatic stress disorder among child victims o f sexual abuse. Development and Psychopathology, 6, 165-181. Wolfe, V.V., Gentile, L, Wolfe, D.A. (1989). The im pact of sexual abuse on children: A PTSD formulation. Behavior Therapy, 20, 215-228. Wolfgang, M.E., Figlio, R.M., & Sellin, T. (1972). Delinquency in a birth cohort. Chicago: University of Chicago Press. Wolfgang, M.E., Thornberry, T.P., & Figlio, R.M. (1 987). From boy to man, from delinquency to crime. Chicago: University of Chicago Press. Zoccolillo, M. (1993). Gender and the development o f conduct disorder. Development and Psychopathology, 5, 65-78.
66 Appendix A: Family Adversity Factor Analysis by Gen der Family Adversity Items Â– FEMALES Factor Load ing 1. MotherÂ’s age at time of childÂ’s birth .644 2. Mother receiving public assistance at time of ch ildÂ’s birth .621 3. MotherÂ’s educational level at time of childÂ’s bi rth -.581 4. MotherÂ’s marital status at time of childÂ’s birth .641 5. MotherÂ’s income at time of childÂ’s birth .570 Eigenvalue = 1.874 Variance = 37.5 Family Adversity Items Â– MALES Factor Loadin g 1. MotherÂ’s age at time of childÂ’s birth .631 2. Mother receiving public assistance at time of ch ildÂ’s birth .562 3. MotherÂ’s educational level at time of childÂ’s bi rth -.552 4. MotherÂ’s marital status at time of childÂ’s birth .720 5. MotherÂ’s income at time of childÂ’s birth .495 Eigenvalue = 1.781 Variance = 35.6
67 Appendix B: Family Conflict Tactics Factor Analysis by Gender Family Conflict Tactics Scale Items Â– FEMALES Factor Loading 1. Threaten to hit or throw things .659 2. Throw, smash, hit, kick things .770 3. Throw something at child .816 4. Push, grab, or shove child .772 5. Slap or spank child .588 6. Kick, bite, or hit child with fist .821 7. (Try to) or hit child with something .780 8. Beat up child .787 9. Burn or scald child .671 10. Threaten child with knife or gun .752 11. Use knife or gun on child .718 Eigenvalue = 6.068 Variance = 55.2
68 Appendix B: (Continued) Family Conflict Tactics Scale Items MALES Factor Loading 1. Threaten to hit or throw things .694 2. Throw, smash, hit, kick things .643 3. Throw something at child .699 4. Push, grab, or shove child .689 5. Slap or spank child .493 6. Kick, bite, or hit child with fist .708 7. (Try to) or hit child with something .695 8. Beat up child .659 9. Burn or scald child .461 10. Threaten child with knife or gun .538 11. Use knife or gun on child .492 Eigenvalue = 4.266 Variance = 38.8
69 Appendix C: Child Abuse Factor Analysis by Gender Child Abuse Items FEMALES Factor Loading 1. Sexual Abuse .803 2. Family Conflict Tactics Scale .803 Eigenvalue = 1.290 Variance = 64.5 Child Abuse Items Â– MALES Factor Loading 1. Sexual Abuse .745 2. Family Conflict Tactics Scale .745 Eigenvalue = 1.109 Variance = 55.4
70 Appendix D: Neuro-Cognitive Factor Analysis by Gend er Neuro-Cognitive Items FEMALES Factor Loading 1. WRAT Spelling score .780 2. WRAT Reading score .828 3. WRAT Arithmetic score .803 4. Birth weight .466 5. Verbal IQ .609 Eigenvalue = 2.527 Variance = 50.5 Neuro-Cognitive Items MALES Factor Loadin g 1. WRAT Spelling score .711 2. WRAT Reading score .799 3. WRAT Arithmetic score .779 4. Birth weight .303 5. Verbal IQ .628 Eigenvalue = 2.239 Variance = 44.7
71 Appendix E: Drug Use Factor Analysis by Gender Drug Use Items FEMALES Factor Loading 1. Marijuana .964 2. Cocaine .961 3. Heroin .968 4. Methadone .922 5. Opiates .920 Eigenvalue = 4.488 Variance = 89.8 Drug Use Items MALES Factor Loading 1. Marijuana .978 2. Cocaine .964 3. Heroin .967 4. Methadone .961 5. Opiates .978 Eigenvalue = 4.701 Variance = 94.0
72 Appendix F: School Deviance Factor Analysis by Gend er School Deviance Items Â– FEMALES Factor Loading 1. Frequently misbehave in school .851 2. In trouble for fighting at school .757 2. Ever suspended/expelled from school .449 Eigenvalue = 1.499 Variance = 50.0 School Deviance Items MALES Factor Loading 1. Frequently misbehave in school .676 2. In trouble for fighting at school .755 2. Ever suspended/expelled from school .780 Eigenvalue = 1.636 Variance = 54.5
73 Appendix E: Deviant Peer Association Factor Analysi s by Gender Deviant Peer Association Items FEMALES Factor Loa ding 1. Friends who are involved in crime .595 2. Friends who smoke .657 3. Friends who use drugs .652 4. Friends who drink .828 Eigenvalue = 1.893 Variance = 47.3 Deviant Peer Association Items MALES Factor Loading 1. Friends who are involved in crime .669 2. Friends who smoke .816 3. Friends who use drugs .752 4. Friends who drink .788 Eigenvalue = 2.300 Variance = 57.5
74 Apendix H: Bivariate Correlations by Sex (Females A bove the Main Diagonal / Males Below the Main Diago nal) AgeArrest FreqArr MomAge MomPub MomEdu MomMar MomInc FamConTa ct SexAbuse AgeArrest -.093 -.064 -.114 .022 .043 .045 .162 .071 FreqArrest -.436** .124 .139 -.040 -.009 .024 -.045 -.040 MomAge .011 .090 .143 -.333** .439** .107 .068 .057 MomPubAs -.085 .057 .086 -.255** .136 .311** .058 .120 MomEduca .062 -.218** .219** -.217** -.179* -.090 -.161 .004 MomMarStat -.051 .158** .425** .257** -.101* .286** .048 .181* MomIncome -.033 .081 .075 .158** -.145** .211** -.062 .190* FamConTact -.014 .042 .155* .017 -.020 .046 .020 .290** SexAbuse .079 -.057 -.001 -.021 -.063 -.017 -.078 .109* WRAT Sp .038 .094 .098 .150** -.173** .048 .025 -.062 -.034 WRAT Re -.104* .243** .089 .202** -.181** .090 .067 -.078 .019 WRAT Ar -.007 .174** .062 .094 -.160** .087 -.028 -.130* -. 021 BirthWeight .004 .023 .110* .032 -.067 .080 .050 -.020 -.001 VerbalIQ -.051 .154** .095 -.048 -.155** .025 .081 -.076 .01 5 DrugScale -.179** .206** .053 .014 -.111* -.044 -.011 .094 .2 07** SchDev -.163** .216** -.070 .019 -.126* .010 -.054 .114* 043 DevPeer -.296** .250** .041 .028 -.052 .076 .010 .161** .07 7 ** p < .01 p < .05
75 Appendix H: (Continued) WRATSp WRATRe WRATAr BirthWeight VerbalIQ DrugScal e SchDev DevPeer AgeArrest .061 .035 -.079 .126 -.091 .004 -.006 -.0 93 FreqArrest .062 .109 .143 -.046 .139 .218** .048 .2 84** MomAge .127 .001 .114 -.041 .063 .150 .063 .145 MomPubAs .105 .096 .267** .122 .098 .120 .147 .110 MomEduca -.277** -.207* -.284* -.264** -.185* -.036 -.183* -.103 MomMarStat .094 -.015 .077 .132 .067 .157 .165 .029 MomIncome -.009 .057 .088 .094 -.078 .226* .069 .04 6 FamConTact .087 .053 .029 .091 -.055 .153 .217** .2 59** SexAbuse .008 .073 -.003 .080 -.064 .360** .130 .24 4** WRAT Sp .559** .454** .255** .402** .071 .330** .043 WRAT Re .468** .633** .294** .303** .073 .203* .0 24 WRAT Ar .414** .497** .243** .386** .046 .250** 060 BirthWeight .153** .155** .096 .126 -.087 .220* 033 VerbalIQ .222** .356** .391** .118* -.107 .155 .0 94 DrugScale -.041 .035 .095 -.054 -.033 .107 .480** SchoolDev .009 .075 .099 -.040 -.002 .227** .269* DevPeer -.075 -.023 -.052 -.120* -.015 .349** .137 ** p < .01 p < .05