Race Differences in Persistence/Desistance: A Trajectory Analysis of Serious Youthful Offenders Followed into Adulthood

Material Information

Race Differences in Persistence/Desistance: A Trajectory Analysis of Serious Youthful Offenders Followed into Adulthood
Copyright Date:


Subjects / Keywords:
Alcohols ( jstor )
Criminal justice ( jstor )
Criminals ( jstor )
Criminology ( jstor )
Hispanics ( jstor )
Juveniles ( jstor )
Nonviolent crimes ( jstor )
Parole ( jstor )
Trajectories ( jstor )
Violent crimes ( jstor )
City of Gainesville ( local )

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright John David Reitzel. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Embargo Date:
Resource Identifier:
659561098 ( OCLC )


This item is only available as the following downloads:

Full Text




Copyright 2006 by John David Reitzel 2


To my wife, Kuniko 3


ACKNOWLEDGMENTS My most important achievement was not comp leting this last hurdle of graduate study. Rather, it was the inspiration I received from th e pursuit itself. Yet, to borrow from what Darwin once remarked about mathematicians, at times over the past two years I have felt as if I were a “blind man in a dark room looking for a black cat that isn’t there.” This dissertation is not the culmination of my singular effort but of the co mbined efforts of many individuals who shined a light for me in searching for this alleged non-exis tent black cat. To all of these individuals, I offer my deepest appreciation. I would first like thank my en tire dissertation committee for their effort on my behalf. Without their generous assistance and invaluable comments, completing this task would not have been possible. Dr. Alex Piquero, my chair and mentor, is the embodiment of achievement and scholarly excellence, and his unselfishness is unrivaled. Alex’s guidance during my entire graduate career was unyielding. I can not begin to imagine how to repay him except to say that I will strive to do for my students and colleagues as Alex has done for me. Dr. Ronald Akers gave generously of his time in sharing with me hi s vast consideration of criminological theory. Through his tutelage, which extends far beyond th is project, my understanding of theory has greatly improved. Dr. Lonn Lanza-Kaduce helped me in formulating this project, particularly the theoretical framework. His sharp questions and insights into my study greatly improved the final product. I must also thank him for everything else that he did for me during my time in Gainesville, from keeping me funded in graduate school (and thus fed) to just taking time out of his busy schedule to talk about things criminological and sociologica l. Dr. Brian Stults helped to significantly improve this manuscript with his clarification of race and crime issues and on methodology. Dr. Julia Graber served as my outside committee member; however, her role was more than that. Her comments on conceptualiz ing this study and clarifying the numerous 4


components were invaluable. I also want to th ank Dr. Nicole Leeper-P iquero, who though not on my committee, provided me with a great deal of sound advice and help in pursuing my professional goals. I want to also thank my many friends and collea gues at the University of Florida. I came to cherish my time Gainesville, not because of its distinguished academic reputation, visual beauty of the campus or terrific football and sports teams, though all of these things certainly helped, but because of the community of friends that were wonderfully inspiring and generous in many subtle ways during the roller-coaster ride that defines gradua te study. Clay Hipke, Dr. Dana Fennell, Dr. John and Foster and his wife Yin, and Yuko Fuji no provided me many good times and opportunities to part with my money in our poker games. Dr. Steve Rice kept me motivated by talking shop, football and politics and was just a great friend and colleague. Dr. Leslie HoutsPicca was the talented graduate student for whom all other sociology graduates, myself included, went to for advice. She made jumping through the hoops of graduate study look easy (though it certainly is not). Helena Al den and Dr. Elizabeth Fakasis provided a friendship and many good times. I also thank Matt Nobles, Dr. Scott Magg ard, Dr. Allison Chappell, Mike Baglivio, Brad Tripp, Andrea Schoepfer, Stephanie Carmich ael, Lynn Langton, Zenta Gomez-Smith, Kristina Deak, Mike Loree, Dr. Laurel Tripp, Tyson Brown, Dr. Guillermo Rebollo-Gil, Ray Hinjosa and Dr. Melanie Hinojosa, Dr. Terry Mills, Dr. Stephen Perz, Dr. Jodi Lane, Dr. Karen Parker, Dr. Kendal Broad, Dr. Danaya Wright , Dr. Angela Gover and everyone else who simply made UF the incredible place it is. I am pr ivileged just to have crossed paths with so many prodigiously talented people. I want to give a special thank you to Sheran Flowers and Kanitra Perr y in sociology and to Hazel Phillips and Dianne Bolinger in criminol ogy at the University of Florida. Sheran and 5


Kanitra’s generosity, bright smiles and jokes (u sually at my expense) made my time in Gainesville that much more enjoyable. Hazel and Dianne continued through the completion of this dissertation to handle my consta nt last minute needs with a smile. A special thank you goes to some of my prof essors at SUNY Cortland. Were it not for their enthusiastic support and beli ef in my potential, it is doub tful I would have gotten through the doors of graduate school. Dr . Virginia Levine and Dr. Craig Little gave me the second chance to prove that I was really quite capable of academic achievement. Their endorsements when I broached the subject of graduate school certainly set me on this journey. Gilda Haines Votra was a great friend who saw me through seven sometimes unspectacular years of undergraduate work. The late Dr. Rozanne Brooks ignited my interest in sociology even though my performance in her class was less than auspic ious. Dr. Stuart Traub, Dr . Harjinder Jassal, Dr. Devereaux Kennedy, Dr. Herb Haines, Dr. Bria n Phillips, Dr. Richard Kendrick, Dr. Jamie Dangler, Betsy Zaharis, and th e late Dr. Frank Hearn nourishe d my passion for sociology and provided a wonderful learning environment that in the end prepared me very well for graduate study. Their impact upon me was more than they can imagine.. I also want to give a special thanks to my colleagues at Illinois State University, particularly, Dr. Ralph Weisheit, Dr. Jeff Walsh, Dr. Jessie Krienert, Dr. Donna Van Diver, Mrs. Pam Fuller, Dr. Ed Wells and Mrs. Susan Woolle n for their encouragement and support. I also want to thank Dr. Sesha Kethineni for providing a helpful review of earli er drafts and Dr. Tom Ellsworth for providing me additional time to comp lete this project. I also want to thank my teaching assistants, Jessica Kolk and Chris Bagnall, whose efforts in helping me when completing my dissertation c onflicted with my other res ponsibilities was invaluable. 6


I want to thank all of my current and form er students, particularly my Fall 2006 Criminal Career (Section 002) students at ISU. Havi ng the privilege to pass knowledge onto so many dynamic students is a privilege th at I do not take lightly and is an important part of why I pursued a graduate education. Now having taught over five hundred students since beginning this journey, I can say it is even bett er than I thought it would be. I am fortunate to come from a large but clos e family that was suppor tive of my academic journey even when it kept me a thousand miles away during many important family functions. I thank my sister Gina DiDonato and her husband Ralph; my nieces, Julianna and Alaina, and my nephews, Michael and Joseph; my Aunt Joanne Balzano and Uncle Ke nny Balzano, my Aunt Rose and Uncle Bob Festa, my cousins Michael Balzano and Maria Varella, Jill Balzano, Chris and Haley Balzano, and the rest my extended fam ily for providing me the greatest family one can ever wish to have. Most importantly, howev er, I thank my parents, James and Ann Reitzel, who have waited years for me to finally return their phone calls to te ll them that I actually completed this dissertation. Their love and beli ef in me never waned despite my many early childhood troubles. Any success I have had or will have is directly attributable to these truly beautiful and caring people. Any failures are certainly all mine. Last, but most certainly not least, my deep est appreciation and resp ect go to my wife, Kuniko Chijiwa, and of course, to our four cat s, Ashura, Mi-chan, Jessie, and Howie (Junior). Kuniko is easily the most talented, loving, and hardest working person I know. Were it not for her giving more to me than anyone ever has the right to ask or expect from another, it is a certainty that this dissertation would not have been completed. 7


TABLE OF CONTENTS page ACKNOWLEDGMENTS ...............................................................................................................4 LIST OF TABLES .........................................................................................................................10 LIST OF FIGURES .......................................................................................................................11 ABSTRACT ...................................................................................................................................12 CHAPTER 1 INTRODUCTION................................................................................................................. .13 Historical Overview ................................................................................................................13 Locating Persistence and Desistance ...............................................................................15 Race, Ethnicity, Persistence/Desistance ..........................................................................17 Theoretical Foundation ....................................................................................................18 Study Format ...................................................................................................................21 2 AGE, CRIMINAL CAREERS, AND POSITIVISTIC CRIMINOLOGY.............................22 Age/Crime Relationship .........................................................................................................22 Age/Crime and Career Criminals ....................................................................................24 Age/Crime Debate and Theory ........................................................................................29 3 RACE/ETHNICITY AND CRIME........................................................................................37 Historical Overview of Race, Ethnicity, and Crime ...............................................................37 Situating Hispanic/Latino Ethnicity ................................................................................40 Social Structure, Culture, and Race/Ethnicity .................................................................41 Correlates of Race and Ethnic Differences in Crime ......................................................43 Dynamic Theories of Race/Ethnicity and Crime .............................................................45 4 THEORETICAL FRAMEWORK..........................................................................................48 Four Theoretical Models .........................................................................................................48 The Criminal Career Model .............................................................................................49 Dual Developmental Taxonomy ......................................................................................54 Life-Course Theory .........................................................................................................57 General Theory ................................................................................................................59 5 LITERATURE REVIEW OF PE RSISTENCE AND DESISTANCE...................................63 Overview .................................................................................................................................63 Empirical Studies of Persistence .....................................................................................64 8


Empirical Studies of Desistance ......................................................................................75 6 DATA AND METHODS.......................................................................................................80 Data .........................................................................................................................................80 Dependent Variables .......................................................................................................82 Independent Variables.....................................................................................................83 Research Questions .........................................................................................................99 Method of Analysis...............................................................................................................101 Semi-Parametric Group Modeling ................................................................................102 Posterior Membership Probabilitie s and Bayes Information Criteria ...........................103 Current Focus................................................................................................................105 7 RESULTS...................................................................................................................... .......108 Data Analysis ........................................................................................................................108 Descriptive Statistics .....................................................................................................108 Analysis of Variance .....................................................................................................113 Multivariate Analysis ............................................................................................................113 Total Crime ....................................................................................................................115 Total Crime Full Sample ........................................................................................115 Total Crime White Sample .....................................................................................119 Total Crime Black Sample .....................................................................................122 Total Crime Hispanic Sample ................................................................................125 Non-violent Crime .........................................................................................................127 Non-violent Crime Full Sample .............................................................................128 Non-violent Crime White Sample ..........................................................................131 Non-violent Crime Black Sample ..........................................................................134 Non-violent Crime Hispanic Sample .....................................................................137 Violent Crime ................................................................................................................140 Violent Crime Full Sample ....................................................................................140 Violent Crime White Sample .................................................................................142 Violent Crime Black Sample ..................................................................................144 Violent Crime Hispanic Sample .............................................................................146 8 DISCUSSION AND CONCLUSION..................................................................................149 Discussion .............................................................................................................................149 Race/Ethnic Trajectory Differences ..............................................................................151 Risk Factor Effects ........................................................................................................151 Conclusion .....................................................................................................................156 LIST OF REFERENCES .............................................................................................................166 BIOGRAPHICAL SKETCH .......................................................................................................184 9


LIST OF TABLES Table page 7.1. Descriptives for parolees across independent measures and race/ethnicity ........................109 7.2. Mean number of offenses by crime type and race ..............................................................113 7.3. Total crime full sample multinomial logistic regression model ........................................118 7.4. Total crime white sample multinomial logistic regression model ......................................121 7.5. Total crime black sample multinomial logistic regression model ......................................123 7.6. Total crime Hispanic multinomial logistic regression model .............................................127 7.7. Non-violent crime full sample mu ltinomial logistic regression model ..............................130 7.8. Non-violent crime white sample multinomial logistic regression model ...........................133 7.9. Non-violent crime black sample multinomial logistic regression model ...........................136 7.10. Non-violent crime Hispanic sample multinomial logistic regression model ......................139 7.11. Violent crime full sample logistic regression model ..........................................................142 7.12. Violent crime white sample logistic regression model .......................................................144 7.13. Violent crime black sample logistic regression model .......................................................146 7.14. Violent crime Hispanic sample logistic regression model ..................................................148 8.1. Summary of trajectory mode ls and significant risk factors ................................................152 10


LIST OF FIGURES Figure page 1.1. Age/crime curve for robbery, burglary, and aggravated assault ...........................................23 7.1. Total crime full sample trajectory model ............................................................................115 7.2. Total crime white sample trajectory model ........................................................................119 7.3. Total crime black sample trajectory model .........................................................................122 7.4. Total crime Hispanic sample trajectory model ...................................................................126 7.5. Non-violent crime full sample trajectory model .................................................................129 7.6. Non-violent crime white sample trajectory model ..............................................................132 7.7. Non-violent crime black sample trajectory model ..............................................................135 7.8. Non-violent crime Hispanic sample trajectory model ........................................................138 7.9. Violent crime full sample trajectory model ........................................................................140 7.10. Violent crime white sample trajectory model .....................................................................143 7.11. Violent crime black sample trajectory model .....................................................................145 7.12. Violent crime Hispanic sample trajectory model ................................................................147 11


Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy RACE DIFFERENCES IN PERSISTENCE/DESI STANCE: A TRAJECTORY ANALYSIS OF SERIOUS YOUTHFUL OFFENDERS FOLLOWED INTO ADULTHOOD By John David Reitzel December 2006 Chair: Alexis R. Piquero Major Department: Criminology, Law and Society The emergence of the criminal career model ma rked a significant touchstone for academic criminology and reignited anew the long-standing age/crime debate within criminology. Since some offenders persist well past the aggregate peak age in offe nding, termed persistence, while others desist in a more normative manner as they move through their indi vidual lives it suggests that there are different types of offenders with potentially diffe rent etiological underpinnings. To date, however, the literature on pe rsistence and desistance in the context of the life-course or developmental perspective is thin and even th inner when race or ethnic differences were examined. Building on this observation, this study seeks to contribute to the crim inological literature on persistence and desistance w ithin a developmental and lifecourse context by employing a semi-parametric group trajectory modeling desi gn to examine adult violent and non-violent offending differences of a racially mixed sample of former juvenile offenders that were released from the California Youth Authority. Findings from this study showed signi ficant race and ethnic differences in persistent viol ent offending and psychological and intelligence differences in nonviolent offending. 12


CHAPTER 1 INTRODUCTION Historical Overview The emergence of the criminal career model (Blumstein, Cohen, Roth, and Visher, 1986) marked a significant touchstone for academic cr iminology. It reignited th e long-standing debate over the age/crime relationship while providing im portant new mechanisms in which to isolate and link together different stages of the offending career. The cr iminal career model also called attention to salient concepts regarding offending careers, such as the prevalence of those committing crime and their individual rates of offending frequency (i.e., “Lambda”) (Farrington, 2003; Blumstein, 1986). Developmental and life-course theories are exte nsions of the criminal career model (Nagin and Farrington, 1992) and along with it, embody the principle that there is continuity in offending over time such that relative offending ra tes in one age-period li nk to relative offending rates in others (Farrington, 2003; Moffitt, 1993; Nagin and Farrington, 1 992; Blumstein et al., 1986). Continuity in offending is also the hallmark of the most high-rate persistent offenders that are thought to begin their offending careers earlier, have longer careers, and to commit more crimes over comparable periods than other type s of offenders (Mazerolle et al., 2000; Moffitt, 1993, 1994). For instance, early onset of offendin g, a key predictor in the developmental and criminal career literature s, has been found to correlate to th e development of problems earlier in childhood and links to continued and often incr easing problems throughout each developmental stage of the life-course (Mazero lle et al., 2000; Moffitt, 1993, 1994). Since some offenders persist well past the a ggregate peak age in offending while others desist in a more normative manner in late adolescence/early a dulthood, it suggests that there are different types of offenders with potentially different etiological underpinnings (Moffitt, Lynum, 13


and Silva, 1994; Moffitt, 1993; Barnett, Blumstein, and Farrington, 1989). It further suggests that persistence and desistance are not simply two sides of the same co in. Rather, they are separate but related consequences of individual crimin ality. Despite their seeming importance to our understanding of crime, persistence and desistan ce are understudied topics in criminology. As Laub and Sampson (2001: 1) recently wrote, “There have been relatively few long-term studies of crime over the full life-span. Consequently, relatively little is known about desistance and, for that matter, the processes of persistent criminal behavior throughout the life course.” Building on their observation and on the recent empirical literature on persis tence and desistance, this study seeks to contribute to criminologi cal knowledge in two important ways. First, the primary objective is to investigat e differences in persistence and desistance across violent and non-violent crime types. This study seeks to determine how factors drawn from the developmental and life-course literatures predict persistent offending or desistance from offending for a sample of young adults who had pr eviously been incarcer ated as juveniles. Examining how persistence and desistance functi on across violent and non-violent offending can add to the literature by determining if they operate in similar or dissimilar ways. For example, persistence in crime has been linked to selective incapac itation. If there are, offenders that definitively stop offending because of identifiable factors, then keeping them incarcerated beyond such time that they are no longer a threat to recidivate might be counterproductive. There is also evidence that th e individuals most at-risk to reoffend are not necessarily the most serious offenders, much less violent offenders (Heilb run et al., 2000; Clear, 1988; Petersilia, Greenwood, and Lavin, 1977). In addi tion, persistence and desistance situate at the center of the most recent theoretical discussions on criminal careers. 14


Second, although the criminology and sociology literatures on race and crime are extensive, few published studies investigate raci al differences in pers istence and desistance within a life-course/developmental framewor k (Piquero, McDonald, and Parker, 2002; Ge, Donnellan, and Wenk, 2001, Elliott, 1994). There ar e even fewer that consider offending trajectories between racial groups and none that consider ethnicit y, particularly Hispanic/Latino ethnicity (Ge et al., 2001). Cons equently, an equally important objective of this study is to investigate persistence and desist ance with respect to differen ces in risk factors predicting offending trajectories across and within race and ethnicity. Locating Persistence and Desistance Most earlier studies of persis tence are predominantly investigations of recidivism, framed in such contexts as “time to first arrest” or “failure to desist” (Pet ersilia and Turner, 1993; Shover and Thompson, 1992). While these studies ar e longitudinal in nature, they only cover very short time spans, typically three years afte r release from incarcer ation. On the other hand, the empirical literature on desistance has genera lly approached it from a static perspective, employing a cutoff point in which to measure when individuals simply stop offending (Bushway et al., 2003). Or they are quali tative in nature that asked non-randomly selected samples of offenders about why or how they desist (e.g. see Maruna, 2001). Although these approaches have provided valuable insight for criminology, they have not been informative about the stability of long-term persistent offending patterns, changes in offending patterns, or persistent differences in offending. In addition, they have not shed much light on any shared patterns that explain why or how people come to desist, at least not from a systematic standpoint. From a developmental standpoint, the key featur e of persistence is its stability across the life-course for a small group of the most freque nt offenders (Moffitt, 1993). This perspective holds that there is a small group of high-rate, fr equent offenders whose offending trajectories are 15


steady across the life-course and whose offending e tiology is also stable, but that it differs from others whose offending is limited to adolesce nce. Moreover, implicit in this position is heterotypic continuity of offending, which is th e concept that persistent offenders typically exhibit increasing seriousness in offending as they age or they display a va riety of behaviors that stem from a single underlying trait. Desistence is the termination of offending. Th e criminal career model employs desistance as a way to measure career lengt h, thus providing additional information for policy construction. The extant literature on the aggregate age/crime relationship suggests that most offenders begin offending during adolescence but desist by their mid-twenties. However, to what extent individual patterns of de sistance reflect the aggregate pattern and at what poi nt do individuals that are desisting become indistinguishable from non-offenders remain empirical questions (Benda, Toombs, and Peackock, 2003; Bushway et al ., 2003). In other words, are most of these offenders getting married or finding gainful employ ment or other adult commitments that might lead them to terminate offending immediately or are other forces operating that causes them to reduce their offending frequency but not to terminat e fully until years later? As is discussed more fully below, earlier theories ha ve treated desistance as an event where people just quit offending (Bushway et al., 2003; 2001). Some researchers, however, have proposed measuring desistance as a process that leads to termination (Mar una, 2004; Bushway et al., 2003, 2001; Laub and Sampson, 2003). Since both persistence and desistance potentially influence the age/crime relationship, consideration of both processes should be taken into account. If, for instance, there is an identifiable group of high-rate pe rsistent offenders whose stability in offending can be tracked into late adulthood and whose risk factors differ from those limited to delinquency in 16


adolescence (or for those who might begin offending in adulthood), then there are compelling reasons to believe that individua l trajectories would not necessari ly comport with the aggregate age/crime curve. What accounts for the stability of offending we ll past the aggregate decline in crime after peak age, and why and how offenders come to de sist are important quest ions without sufficient answers. In this vein, developmental and life-co urse theories have taken on added importance since they specifically attend to the initia tion and termination of offending and to the implications that identification of a persiste nt offending group might have for the age/crime relationship. Conversely, when set against Gottfredson and Hirschi’s general theory, which posits that different groups of offenders do not exist but rather, variation in offending frequency is a result of underlying criminal propensity, it can help sharpen not only our understanding of the theoretical differences but of offending itsel f. Together these rece nt theories have the advantage of better addressing an array of unsettled issues about the development of criminality over time. Together, they have pressed researchers to rethink thei r conceptualizations of crime and criminality, and to consider the full le ngth of the individual offending career. Race, Ethnicity, Persistence/Desistance Noticeably absent in the empirical literature ar e studies that investig ate potential race and ethnic differences in persistence and desistance. Racial or ethnic group membership plays an inextricable role in our unders tanding of crime Whether signif ying potential socioeconomic class or residential antagonisms that correlate to race-differentiated crime rates or as a potential reflection of institutionalized racism in the crimin al justice system, it is necessary to investigate the role of race and ethnicity in differences in offending across the life-course (Walker, Spohn, and Delone, 2004). 17


Black Americans comprise less than 15% of the total population yet they make up more than 37% of those under some form of correctional supe rvision (probation, ja il, prison, parole) (Bureau of Justice Statistics, 2006) and nearly 64 % of all inmates who leave prison (Petersilia, 1999). Punishment also distributes differently along ethnic lines. For instance, just as violent and property crime rates have dropped to levels not achieved since the 1960’s, arrest and incarceration rates for Hispanic/Latino American s of any race have increased more than any other comparable demographic group, which makes minorities of any color or ethnicity over three times more likely to be involved with the criminal justice system (Petersilia, 1999). Racial and ethnic group membership has a strong bearing on how we think about crime, criminals, and criminal justice. That race and ethnic matters are among the most contentious social issues in the United States, particularly when discussions turn on biological, genetic, or cultural differences in propensity to offend or on the existence of institutionalized racism in the criminal justice system, signals to its central im portance to criminological theory. As racial and ethnic issues converge on age and crime issues , the discussions become considerably more convoluted since the effects of race may be differ ent from that of ethnicity, both of which might differ from the effects of age on propensity to offend or on the criminal justice systems response to offending. Theoretical Foundation Earlier studies of offending have traditionally focused on differentiating between offenders and non-offenders but recent theories assert that it is important to make distinctions between offenders (Chung et al., 2002; Farrington, 1996). Moreover, delineating between theoretical models that until recently have b een less than adequate in explaining racial or ethnic differences and that hold contrasting positions on the age/cr ime relationship is an important step toward reconciling unnecessary discord that has confronted criminology of late. Toward this end, there 18


are four theoretical models that inform this study: Blumstein and colleagues career criminal model, Moffitt’s dual taxonomy, Gottfredson a nd Hirschi’s general theory, and Sampson and Laub’s life-course theory. Blumstein and colleagues’ (1986) criminal car eer model is first among these perspectives since it provides important th eoretical constructs and a uni fying conceptual underpinning on which the other theories extend. However, because it structures offending around a framework specifically intended to inform crime contro l policy—with somewhat dubious implications regarding selective incapacitation—its capacity as a theory of crime is constrained (Laub and Sampson, 2003; Tittle, 1988). Moffitt’s (1993) dual taxonomy, on the other hand, offers an e xplicit theoretical typology that moves beyond policy by offering more vibran t theoretical guidance in which to measure persistence and desistance. Grounded in the neuropsychological development of antisocial behavior over the life-course, Mo ffitt’s theory provides two disc rete pathways of offending. The first, termed life-course persistent antisocial beha vior (LCP), places the ro ots of persistent antisocial behavior and offending ove r the life-course in pathologi cal neuropsychological deficits. The second, termed adolescent-limited anti-social be havior (AL), applies to the typical juvenile offender who engages in crime during adolescence but then desists upon en tering into adulthood. The roots of AL offending are found in adolescen ce where the coalescence of transient factors such as mimicking other offenders (particularly, mimicking LCP behavior) to fit in or rebelling as a result of a maturity gap, which prohibits them from engaging in behaviors that they enjoy as a result of legal or moral proscriptions. Both of which cause AL offe nders to initiate and terminate offending. 19


Although it is with the career criminal model and Moffitt’s developmental taxonomy that the current study originates, two other theories in form this research. Gottfredson and Hirschi’s (1990) general theory of crime provides a useful counter position for in terpreting the results because they deny the existence of different types/groups of offenders and they dismiss the notion that criminals have “careers” (Gottfreds on and Hirschi, 1986). Rather, they argue that stable differences in levels of self-control underpin all differen ces in offending at all ages and therefore, criminologists must investigate the fact ors correlates of self-control. Importantly, it should be mentioned here that th ere are few indicators of low self -control aside from measures of criminal and analogous behaviors and thus, Gottfredson and Hirschi’ s position is more useful to this study in what they state about criminal car eers and the non-existen ce of distinct offender groups. Lastly, Sampson and Laub’s (2004, 1993) age-graded theory of cumulative disadvantage is also relevant. To borrow from Merton (1949), it is helpful to consider their perspective as a theory of the “middle-range” since it brings to gether properties of Hi rschi’s original bonding theory and properties of labeling theory while also drawing from both developmental theories and general theory in ap plication (Sampson and Laub, 2004). It pr ovides practical constructs that can elucidate findings from a somewhat different perspective than both Moffitt or Gottfredson and Hirschi. For example, Sampson and Laub do not subscribe to the notion that there are “groups” of offenders, but they do believe that there is stability in offending across the lifecourse. They also consider cha nge in offending and that anybody, given the right convergence of circumstances, can turn away from crime (or toward it). They assert that the effects of low socioeconomic status, unemployment, and human agen cy, and most importantly, that age-graded informal social controls play a significant role shaping or changing be havior over time (Sampson 20


and Laub, 2004); all of which call attention to th e reasons why some offenders might continue offending even if the probability diminishes with age. Their theory also explains why others desist as adults. This study assumes no preference toward any of the theories. Rath er, the goal is to determine whether the outcomes support one, some, or all of the theories, and to investigate how such findings impact upon our understanding of continuity and term ination of offending, particularly in how such patterns might comport with the age/crime relationship. Study Format Beginning with Chapter 2, the historical b ackdrop of age and criminal careers was reviewed. Chapter 3 follows with a review of race issues in crime and criminal justice. Chapter 4 reviews the four theore tical perspectives informing this study. Included in this chapter is a relevant discussion about theoretic al assumptions regarding crime and criminality and what each theory claims about the age/crime relationship. Chap ter 5 is a review of th e extant literatures on persistence and desistance, resp ectively. Chapter 6 is the methods and data chapter, which includes a detailed description of the data source and collection methods. In this chapter, the four research questions that underp in this study are presented and a comprehensive account of how both the independent and dependent measures we re constructed is provided. This includes a thorough explanation of the methods employed to analyze the data and answer the research questions. Chapter 7 is an analysis of the resu lts for the violent, non-vi olent, and total crime offending models. Lastly, Chapter 8 is a discussion of the findings and includes implications for our understanding of persistence a nd desistance in offending across race/ethnicity, crime types, relevant theory, and the age/crime relations hip. The study concludes by taking account of the limitations as well as providing di rections for future research. 21


CHAPTER 2 AGE, CRIMINAL CAREERS, AND POSITIVISTIC CRIMINOLOGY Age/Crime Relationship Perhaps the most reliable observed phenome non in positive criminology is the age/crime relationship, which across historic al era, geographic locale, populat ion variation, and a variety of violent and property crime types, consistently exemplifies the same basic distributional pattern of crime across age (Piquero et al., 2003; Tittle and Grasmick, 1998; Steffe nsmeier, Allan et al., 1989; Hirschi and Gottfredson, 1983). As illustrate d below in Figure 2.1, aggregate crime rates sharply increase during the adol escence, peaking somewhere be tween 15 and 19, then decreasing steadily thereafter. (e.g., see aggravated assault in Figure 2.1) (Benda, 2003; Aguilar, 2000; Farrington, 2000, 1986; Gottfredson and Hirschi, 1988; Blumstein and Cohen, 1987). Although the exact shape of each curve differs by specific crim e type. Goring (1913) claimed that the age/crime relationship obeyed a “law of na ture” while Sampson and Laub (2003) called it “accepted wisdom.” Although crime is not exclusively the domain of the young, the primacy of its relationship to youth can hardly be understa ted. Indeed, the age/crime relati onship has ordered our thinking about crime and crim inality since the 19 th century (Cullen and Agne w, 2003; Sampson and Laub, 2003), and is now one of the most studied i ssues in criminology (Piquero, Farrington, and Blumstein, 2003; Tittle and Grasmick, 1997; St effensmeier, 1989; Tittle 1988). The age/crime relationship is also a principa l assumption on which many contemporary crime theories rest. To the extent that the age/crime relationshi p is empirical fact, co mmon theoretical accord has proved elusive. There are two immediate explan ations for this. First, the interdisciplinary composition of criminological theory inhibits a c ohesive overarching theoretical paradigm that most criminologists subscribe to like that of Darwinian evolution in biology (Cullen and Agnew, 22


2003; Montagu, 1941). While some criminologists suggest that it makes criminological theory intellectually rich (C ullen and Agnew, 2003), it does not en courage widespread consensus. Second, since theories of crime derive from a va riety of academic disciplines they are often in opposition regarding the assumptions they make about the nature of criminality and human behavior (Gottfredson and Hirschi, 1990). Theoretical discord is al so a result of uncertainty over the operationalization of concepts, which have diffe rent meanings for different disciplines. As a result, explanations of offending are often mired in complicated interactions of internal and external factors, situational determinants, time-d ependent factors, and di sciplinary differences. Alternately, some researchers have challenged that the age/crime relationship is simply unexplainable with traditional soci al science variables, thus ther e is no interaction between other variables and age (Tittle and Grasmick, 1998; Go ttfredson and Hirschi, 1990). This appears to have created an either/or scenar io of explanations while ignori ng a potential “both” explanation. Three questions emerge when attempting to explain the determinants underpinning this seemingly easy to understand relationship. First, what causes crime rates to rapidly increase Figure 2.1. Age/crime curve for robbery, burglary, and aggravated assault (Federal Bureau of Investigation, 1984; reprinted with permission) 23


during adolescence in the first place? Second, what causes them to decrease steadily as soon as they peak (Piquero, Farrington, and Blumstein, 2003; Gottfredson and Hirschi, 1990)? Last, are there meaningful differences between the age/cr ime relationships for different crime types (e.g. as Figure 2.1 depicts, crime peaks later and drops off more gradually for a ggravated assault than for other crimes) (Steffensmeier, Allan et al., 1989)? 1 For whatever advancements have been made in over a century and a half of research on crime and criminality, the unsettled age/crime ques tions remain front and center of persistent discord. Sellin (1940) forewarned that the age/crime relationship might be unsolvable in writing, “The research student who is in pursuit of an answ er to the relationship of age and crime . . . is doomed to disappointment” (Wolfgang, Figlio, and Sellin, 1972: 106). 2 He was prescient in his edict. Age/Crime and Career Criminals The age/crime relationship converges with the popular notion of the habitual and irredeemable career criminal. Criminology’s adhe rence to the existence of both ideas can be traced back to the pioneering works of early so cial positivists. Most prominent of them was Belgian social physics and statistics luminary, Adolphe Quetelet. 3 Among Quetelet’s many noteworthy contributions to posi tivism, such as developing the concept of the average man, the body mass index, and other advances in statisti cal regression, he found that crime strongly correlated to young people less than age thirty an d that age differentiated and individual’s 1 A more detailed discussion of the age/crime debate is in Chapter 4 2 This quote was reproduced from a pa ssage in Wolfgang, Figlio, and Sellin’s Delinquency in a Birth Cohort . Sellin made the original statement in an article entitled, “The Cr iminality of Youth,” which appeared in the American Law Institute in 1940. 3 Quetelet employed the term social physics ( physique sociale) to describe his positivistic social research. It is believed that Auguste Comte, who is considered the father of “sociology, “ actually coined the term for positive social science but later invented the term “sociology” since he believed that Quetelet stole the social physics term from him. 24


propensity toward property and person crimes (H ankins, 1968; Quetelet, 1968). In a particularly telling passage that is illustrative of the value of age on criminal propens ity, Quetelet (1968: 92) wrote: Of all the causes which influence the developm ent of the propensity to crime, or which diminish that propensity, age is unquestionabl y the most energetic. Indeed, it is through age that the physical powers and passions of man are developed, and their energy afterward decreases with age. Reason is developed with age, and continues to acquire power even when strength and passi on have passed their greatest vigour. Years later, Cesare Lombroso (1876) would make similar claims while simultaneously bringing to the forefront of positivistic crime re search the notion of a biologically primitive criminal, which he labeled an atavist. Writi ng that every age had its own “age-specific criminality” and that crime rates were highest for people between ages 15 and 30, he considered factors that caused juvenile crime different from that which caused atavism, attributing juvenile criminality to the “virility” of youth and “i nstinctive tendencies toward law breaking” (Lombroso-Ferro, 1944: 152). Lombroso and Quetelet viewed juvenile deli nquency largely as a byproduct of youth, not of biological or other individual traits. Yet, there is an additional important theme common to each of their positions. Both suggest that criminal propensity is pliant (for most people); that it increases as individuals enter into and pro ceed through adolescence but declines in adulthood after having fully developed the ability to reason and appreciate the consequences of actions. This is notable since distinguish ing between career criminals and normative juvenile delinquents, and in determining the extent to which criminal propensity changes or remains stable for each type of offender are core issues in criminology. In the early part of the twen tieth century, sociologists at the University of Chicago expanded the study of crime and age with their groundbreaking studies of Chicago urban life. Their numerous contributions to social scien ce included developments in sociological and 25


criminological theory, empiricism, and methodol ogy, all of which have important bearing for life-course criminology research. Of the vari ous defining features characterizing Chicago sociology, a grand and multifaceted examination of urba n social life certainly stands out. This is readily apparent by the emphasis placed on empl oying multiple methods and theories, fostering of collaborations that cut across academic discip lines, and in bringing ac tivists, journalists, and other community members into the fold of examining social problems (Bulmer, 1984). For example, Robert Park was the firs t to apply plant life ecology to th e social environment of a city, of which Ernest Burgess’ concentric zone theory emerged. Park was also the first to write about the “collective effects” of neighborhood residency and that having a high number of young males in a community contributed to its crime rates (Bulmer, 1984: 107). It was Thomas and Znaniecki’s, The Polish Peasant in Europe and America (1920) and Shaw’s, The Jack Roller (1930) that illustrated the impor tance of understanding social and criminal behavior over the life-course. For exam ple, in following the maladaptive life-course development of a petty juvenile delinquent named Stanley, Shaw employed a combination of life-history diaries (“his own stor y”), field research, background checks of official records and, an empirical analysis of Chicago communities, to capture the origins and development of criminal behavior (Bulmer, 1984). In other words, he integrated a life-cour se approach with other methodological and theoretical appro aches in investigating the forces that encouraged change in or continuation of offending behavior. In the 1930s and 40s, Sheldon and Eleanor Glueck continued this “life-course approach” by initiating an important longitudinal study entitled Unraveling Juvenile Delinquency . The Gluecks’ had matched 500 delinquent boys to 5 00 non-delinquent boys on community residency in poor neighborhoods, age, ethnicity, and “globa l intelligence” (Glueck and Glueck, 1950). The 26


goal was to understand why some boys became de linquent while others did not given similar residency in underclass neighborhood s and similar traits thought to be important in causing crime, which, as Sampson and Laub (1993) argue, is a central issue that still confronts criminological studies of delinque ncy and development. Their findi ngs contributed to theory by highlighting the differences in the effects of community factors interacting with individual characteristics in causing delinquency among some youths but not others. From a life-course or developmental perspective, their findings point ed to the need for following people over time, preferably a very long time. In the seventy plus years since the Gluecks’ study, a number of othe r critical longitudinal investigations have emerged whose efforts have come to define the underpinnings of recent criminal career, life-course, and developmenta l criminology literatures. Chief among them was Wolfgang, Figlio, and Sellin’s Philadelphia Birt h Cohort Studies (1972), which stand as the seminal works in contemporary criminology (P iquero et al., 2003; Sampson and Laub, 2003; Morris, 1972). In their investig ation of a juvenile cohort born in Philadelphia in 1949, the researchers’ found that 6% of the cohort and 18% of the delinquent subset committed nearly 52% of all the crimes by the cohort. Even furt her, less than 3% of the delinquent subset committed over 70% of the most serious felonies (Piquero et al., 2003; Wolfgang et al., 1972). Wolfgang and colleagues study sent reverberations th roughout criminology and was the catalyst for a flurry of empirical studies and new theories (Gottfredson and Hirschi, 1988). Their findings suggested that there was a sma ll group of highly active criminals who were responsible for an extraordinary amount of serious crime. In other words, their findi ngs suggested the existence of a small group of career criminals. 27


Other noteworthy longitudinal studies that were commissioned in the wake of the Wolfgang studies include the Cambridge-Somer ville Project (1978); West and Farrington’s Cambridge Youth Study (1961-present); Elliott’s National Youth Survey (1972); Montreal Longitudinal-Experimental Study (Tremblay et al ., 2003); Dunedin Multidisciplinary Health and Development Study (1972); Proj ect on Human Development in Chicago Neighborhoods (1998); Seattle Social Development Project (1981); th e trio of Office of Juvenile Justice and Delinquency Prevention (OJJDP) studies to include the Pittsburgh Youth Study (Loeber), Rochester Youth Development Study (Thornberry ), and the Denver Youth Study (Huizinga); and, Sampson and Laub’s (2005, 1993) reconstruction and update of the original Glueck data. 4 Although the scope and purpose varies for each study, they all share the underlying principle of investigating dynamic processes in the develo pment of criminal behavior over time. Since 1986, numerous critical theo retical developments have also taken place. Blumstein and colleagues’ criminal career model was the first and paved the way by providing a framework for investigating the criminal career. Hirschi a nd Gottfredson, who had firs t written an important essay in 1983 on their interpretation of the age/ crime relationship, which preceded the Blumstein and colleagues model, countered conventional ag e/crime interpretations with a new general theory of crime that set the root causes of o ffending and analogous behaviors in individual selfcontrol. Others, such as Rowe, Osgood, and Nicewander ( 1990), Le Blanc and Loeber (1990)/Loeber and Le Blanc (1998), Hawkins and Catalano (1992), Moffitt (1993), Patterson, Capaldi and Bank (1993), Patterson and Yoerger (1997), and Farrington (2000), penned theories whose features draw from the criminal ca reer model, but place offending causes in 4 Piquero, Farrington, and Blumstein (2003) provide a more detailed listing and explanation of the most notable longitudinal studies, including ones not mentioned here. Their coverage provides an excellent summary of the parameters and findings of each study. 28


developmental psychology, unobserved latent traits, or other individual risk factors that link the continuity of behavior over the life-course. Ot hers still, such as Sampson and Laub (1992, 2005), Tittle (1995) , H agan (1997) and, Thornberry (1987, 1997), and Colvin (2000) reconsidered traditional sociological theory by integrating multiple theories or by fitting existing theories within a life-course framework. Age/Crime Debate and Theory Hirschi and Gottfredson (1983) proposed th eir own explanation of the age/crime relationship arguing that it was invariant acro ss social and cultural conditions, and that sociological theories should not be tied to explaining this relationship. In 1986 and 1987, Blumstein and colleagues published findings from their research on criminal careers and what they found ran counter to many prevailing beliefs in criminology and particularly, to Hirschi and Gottfredson’s position. For instance, frequency of offending was relatively stable for active offenders and termination rates for those over 30 was low. That is, if they continued offending into their thirties they were less likely to stop. In addition, demographic factors that crosssectional data demonstrated as important di d not significantly link to offending frequency (Blumstein and Cohen, 1987). Gottfredson and Hirschi (1988) contested these fi ndings and the implications of the career criminal model in general by dismissing the utilit y of the entire model as being a fixed tool of policy such that Blumstein and colleagues had se t out to identify career criminals for policy purposes and then designed the model in a way that found exactly what they wanted to find; which was the existence of career criminals (G ottfredson and Hirschi, 1988; Tittle, 1988). Furthermore, they argued that individual comp onents of the criminal career model such as initiation and frequency did not have any theoreti cal significance and that their model—and most 29


criminological theories as well—neglect to accoun t for the generality of crime (Pratt and Cullen, 2000). Blumstein and colleagues (1988) responded to the criticisms by arguing that their position was mischaracterized regarding, among other things, the meaning of the criminal career and the value of Lambda in understanding criminality. In effect, they accused Gottfredson and Hirschi of setting up straw man type arguments in order to undermine the model’s value. This drew additional responses and rejoinders, which further opened up criminology to a renewed discussion about the role of age in crime resear ch as well as a debate over the use of cross sectional versus longitudinal methodologies to an swer core criminological questions. Nagin and Land (1993) characterized the more recent disput e as reaching a level of contention not seen since the Gluecks’ and Sutherland de bate in the mid twentieth century. Tittle (1988) offered an articulate comment ary on this academic tit-for-tat, summarizing not only where the disagreements lie, but by pointi ng out the inconsistencies in each position (in both the assumptions they make about their resp ective positions and about each other’s). He identified five basic disagreements, which in clude the validity of selective incapacitation policies, invariance of the age/crime relationship, utility of longitudinal and cross-sectional data, government “programmed” research, and the theo retical import of career criminal research (Tittle, 1988: 75). In broad terms, the early age/crime debate ce ntered on whether the underlying patterns of the age/crime curve reflect changes in participat ion or changes in frequency (Vold, Bernard, and Snipes, 2002). The traditional position was th at offending frequency among active offenders drove the shape of the curve and therefore the sh arp rise and later dec line after peak offending represented changing rates of offending of those e ngaging in criminal behavior (Blumstein et al., 30


1986). Conversely, the criminal career position sugge sts that the shape is predominantly a result of participation; that th e curve represented an influx of offend ers prior to peak and the aging out for most offenders by their mid tw enties, albeit with a small group of persistent offenders whose offending frequencies remained stable at a high rate (Blumstein et al., 1986). Somewhat of a paradigm shift occurred in the 1990’s following the publishing of numerous life-course, developmental and general theories. It now centers on career criminals vis-vis criminal propensity (Vold et al., 2002). The career criminal position has not changed. Its emphasis still hinges on changes in participation with a small group of high-rate persistent offenders. However, the ascendance of crimin al propensity and its relative hold for some regarding the age/crime relationship ha s become the prime counterpoint. Specifically, criminal propensity position is a linchpin of Gottfredson and Hirschi’s general theory and it holds that the effect of age on crime is inva riant, particularly across crime types. Therefore theories of of fending must explain the causes of differing but stable propensities of individual offenders (Sampson and Laub, 2005b; Warr, 1993; Gottfredson and Hirschi, 1990: 128). Tittle (1988: 76) points out , however, that both sides of th e debate have assumed extreme positions with little evidence (twenty years ago) to substantiate either position. Specifically, he argues that their positions are empirical questions. This need not be nor is it necessarily the case. Studies by Rowe and Tittle (1977) and Steffensmeier and collea gues (1989) are two examples that challenge this invariance assertion, at leas t from the extreme position that Gottfredson and Hirschi have taken. Rather than rehashing the debate since those differences and si milarities have already been addressed above and extensively elsewhere, (s ee Vold et al., 2002; Bl umstein et al., 1988, 1987, 1986; Gottfredson and Hirschi, 1988, 1987, 1986; Tittle, 1988) and since Moffitt’s 31


developmental position toward the age/crime curv e is quite similar to the criminal career’s position, the discussion turns to how Moffitt’s ta xonomy and Gottfredson and Hirschi general theory measure up. Both theories converge on the fact that they are at first trait-ba sed and developmental theories; at least in part (Delisi, 2001). In effect, each assumes problems with parenting or irregular socialization in childhood that leads to the development of trait-based pathological behaviors. Whether the trait is a neuropsychological deficit or lo w self-control is not necessarily important. Rather, it is the f act that early childhood problem s link to later problems in adolescence and adulthood. For Moffitt, it begins at birth or during pregnancy where problems that could be the result of the mother’s health or medical problems l ead the child to sustain verbal and executive neuropsychological deficits. Treated, individuals can overcome such deficits, but in families where the parents lack the ability to handle the child’s deficits or just do not properly socialize a child, they get worse. This eventually leads to a pathological antisocial personality disorder. Alternately, Gottfredson and Hirschi leave open the possibility that low self-control could be the result of neuropsychologi cal deficits and poor paren ting, though the emphasis for them derives from Hirschi’s earlier control theory, which places the problem nearly exclusively on parental bonds. To them, crime is “universally attractive,” and children are more easily seduced by the seeming payoff of crime. Thus, offending be havior reflects an absence of “internal (low self-control) and external (opportunity) constrai nts, which are better developed in children who have been properly socialized by their parent s (Tittle and Grasmick, 1998: 317). However, the fact remains that early childhood sets the stage for later criminal (and analogous) behavior during adolescence and adulthood. 32


It is in adolescence where these two theori es divide. Moffitt echoes the criminal career position that the precipitous incline in aggregat e offending during adolescence is a result of both LCP offending steadily and the la rge influx of AL offending as a form of rebellion or the mimicking of LCP types. Conversely the decrease in the late teens/early tw enties is a result of AL offenders aging out since the status proscrip tions they were rebelling against no longer apply. From her view, the age/crime relationship necessarily reflects the existe nce of two groups that contribute to the shape of th e curve in different ways. Gottfredson and Hirschi assert that the age effect on crime is invariant across time and space, but this leads them to a different conclu sion about the age/crime relationship. For them, it demonstrates that the aggregate increase and d ecrease in offending during adolescence and early adulthood is necessarily a change in frequency across all ages even though criminal propensity does not change between individu als. It would follow that cr iminal propensity can only be explained by would vary between individuals, whic h are parenting factors that lead to stable differences in self-control. The only commonality that they seem to recognize between their position and Moffitt’s or Blumstein and colleagues is that criminals have careers in the sense that there are stable differences in offending; a point that Nagin and Land (1993: 329) called “unassailable.” Over the last two decades, several other theore tical and empirical issues have come to the forefront of the age/crime debate. Moffitt ( 1993) and Sampson and Laub (2003), for example, separately draw attention to some common pr oblems with earlier theori es. They argue that criminological theory’s response to the age/cr ime relationship has been to focus almost exclusively on juvenile criminality at the expens e of more extensive investigations of childhood and adulthood; two periods in the lifecourse that are undoubtedly important in making crime. 33


They also posit that most delinquency theories make assumptions about criminal behavior irrespective of age. By definiti on, this leads such theories to fail to link adolescent delinquency to childhood antisocial behavior or to offending in adulthood. Additionally, Moffitt (1994) suggests that the theoretical debates over juvenile offending are unnecessary. She acknowledges the validity of earlier delinquency th eories in setting the scope of j uvenile offending; that in their own way each captures important dynamics of delinquency. The main issues with earlier delinquency theories are that th ey do not anticipate change s in aggregate offending during adolescence. In other words, the massive increase in the prevalence of antisocial behavior that occurs during adolescence, which masks different types of offende rs differing offending etiologies (Moffitt, 1993). Another issue germane to the age/crime and criminological theory discussion centers on direction criminological theory is heading. This is similar to th at which confronted sociological theory fifty years ago. Some criminologists have recently asked whether general theories serve criminological theory better as a whole than behavior or event-specific theories. As Gottfredson and Hirschi (1990) and Sampson and Laub (2005) have both argued, the in flux of theories too narrowly tailored to specific be haviors or events (such as developmental theories) make it more difficult to achieve order out of seeming ever-increasing chaos. 5 To resolve this problem, they suggest that criminological theory needs to move toward more general explanations of offending. However, as with general sociological theories , general criminological theories have yet to convincingly substantiate the universality of be havioral causes across tim e, cultural boundaries, and diverse situations (Horney, 2006; Geis, 2000, Birbeck and Lafree, 1993). Such general 5 Such reasoning by Gottfredson and Hirschi could explain why they were so strident and extreme in their initial criticisms of the criminal career mode l back in the 1980’s (see Tittle’s 1988 Criminology article for a good summary of the apparent underlying motivations of each side in advancing their position). 34


theories tend to be ethnocentric and they are often dismissive of the role of law in defining behavior (Geis, 2000). Most, however , are not well suited to capture changes in behavior in their quest to explain all behavior at all times. On the other hand, it is clear that behavior or event-spec ific theories have added a dimension of complexity that i gnores the generality of deviance and thus, fail to appreciate how behaviors transcend disciplinar y understanding (Pratt and Cullen, 2000; Gottfredson and Hirschi, 1990). It seems counterintuitive to dismiss either approach, given that theories, by definition, must be fallible (and if it is not , then it is not a theory) and that the nature of human behavior— criminal or otherwise—can at once be both si mple and complex. From a positivistic standpoint, it would be unthinkable to be so dismissive of either approach without having fully pursued the empirical side of the equa tion to some logical end. In highlighting some of the persisting unanswered questions, Piquero, Farrington, and Blumstein (2003) addressed unresolved age/crime issues that bear import on criminological theory. In their review of the state of criminal career lite rature, they summarized seven unanswered questions age/crime. I add three more (questions 8-10) that I think are relevant: 1. [To what degree is] the aggregate pattern disp layed in the age/crime curve similar to—or different from—the pattern of individual car eers and is similar to—or different from—the pattern of individual careers and whether conclusions a bout individuals can be validly drawn from aggregate data? 2. How far does the observed peak of the aggregate age/crime curve a function of active offenders committing more crime or of more individuals actively offending during those peak years? 3. Within individuals, to what extent is th e slowing past the peak age a function of deceleration in continued criminal activity or stopping by some of the individuals? 4. Across individuals, how much of the age crime curve can be attributed to the arrival/initiation and departure/termination of different individuals. 5. What about the role of co-offending? 35


6. How much of the continuation of offending by lone/solo offe nders is attributable to identifying theirs as the key criminal caree rs of long duration with their co-offenders serving merely as transients with shorter careers? 7. How much of the age/crime curve for any particular crime type? 8. Could not the aggregate age/crime curve refl ect both a change in participation and a change in frequency? Cannot the polar positions coexist in explaining the relationship or do their extreme positions encourage an eith er/or choice, thereby setting up a false dichotomy? 9. Assuming the effect of age on criminality is invariant for most offenders, does this necessarily diminish the potential importa nce of investigating career criminals in contributing to our understanding of age and crime? 10. Since the age crime curve differs somewhat across crime types in both when crime peaks and how sharply it increases or decreases befo re and after peak, what is the relevancy of the differences between specific crime types for criminological theory? (for an example, see Figure 1.1) The answers to these questions, however, extend beyond arguments over the age/crime relationship and point to the uncertainty and pitfalls of the theoreti cal approaches that have come to define the debate. Many of the criticisms are empirical questions not in controvertible fact. The only thing that does seem certain is that the entire debate ne eds to move past the extreme positions. It demands answers that are more as well as a reconciliation of the theoretical tribulations, to include an acknowledgement of the validity of both general and behaviorspecific, and cross-sectional and longitudina l data in advancing theory (Tittle, 1988). 36


CHAPTER 3 RACE/ETHNICITY AND CRIME Historical Overview of Race, Ethnicity, and Crime Obviously Jim Crow no longer exists in the U.S., offici ally, but Jim Crow racism certainly does. Indeed, W.E.B Dubois’ (1903 ) oft quoted decree that the “problem of th e twentieth century is the problem of the color line” certainly endures to this day. In many critical ways, while the social and economic conditions for many black and ethnic minorities have improved over the past forty years, pervasive racism and econom ic disadvantage still exists (Thernstrom and Thernstrom, 1997). In his book, The Declining Significance of Race (1978), W.J. Wilson wrote that the subordination of some blacks and the advancemen t of others is more a function of “economic class in the modern industrial period” than race. Societal cha nges have measurably diversified the nature of the black American experien ce despite opposition on th e part of many white Americans to governmental intervention, such as with affirmative action programs that have helped ameliorate some racial and economic in equality (Bobo and Kluegal, 1993; Kluegal, 1990; Wilson, 1987). This has necessarily made discussions of race more complex since there is less of a universal black experience to speak of and because for a portion of black Americans the positive achievements of the civil rights era bene fited them well. Yet, the economic and social progress made in recent decades is necessarily te mpered by the reality that too many segments of the black population remain trapped in the pressing exigencies of concentrated poverty, de facto residential segregation, social and economic marginalization, and by their increasing overrepresentation in the criminal justice system (Sampson and Bean, 2006; Sampson and Morenoff, 2004; Walker, Spohn, and Delone, 2004; Feagin, 2000; Krivo, Peterson, Rizzo, and Reynolds, 37


1998; Kennedy, 1997; Sampson and Wilson, 1995; Massey and Denton, 1993; Eggleston and Massey, 1992; Wilson, 1987). To illustrate the interaction of race and economic improvement, it is helpful to examine changes in poverty rates and median-household income for blacks compared to whites. U.S. Census Bureau data (2006) show that betw een 1960 and 1995 the percen tage of blacks living below the poverty line decrea sed from about 55% to 32%. By 2000, the black poverty rate decreased to 22%. At the same time, household income increased for all groups. In 1971, the median household income for blacks was about 19,000 dollars compared to 32,000 for whites. By 2000, median household income had increas ed to 34,500 dollars for blacks and to 46,000 dollars for whites. However, in 2000 the racial di sparity in median household income was still as large as it was forty years prior, which is tr oubling when considering that blacks had only achieved a median household income level by the year 2000 that equaled whites 1968 median household income (U.S. Census Bureau, 2006). As informative as these numbers might be, they do not tell us about the causes of social and economic disparities confronting black Ameri cans or about the social structure that have helped maintain such dispar ities. For instance, the corre lation between socioeconomic deprivation and heightened criminal offending is one of the most widespread general findings in academic criminology (Fergusson, Swain-Campbell, and Horwood, 2004). In a rather striking finding, Sampson and Wilson (1995: 42) found that in 171 cities with popul ations greater than 100,000 the black-white racial dispar ity was so complete “that the ‘worst’ urban contexts in which whites reside are considerably better th an the average context of black communities. Research by Massey and Denton (1993), Eggers and Massey (1992), and Krivo and colleagues (1998) are some of the numerous studie s that support this general conclusion. 38


Findings from such research suggest that not only does general economic improvement not tell a complete story about changes in poverty; th at concentrated disadvantage, particularly for poor blacks, is almost wholly worse for blacks th an it is for whites. The societal changes that have improved the economic well-being for most Americans, especially those changes that brought a notable percentage of bl acks out of poverty, did not opera te evenly nor did they lessen the gap between median household incomes across race. At the same time, while crime has declin ed in many minority communities, official statistics reveal that crime is still an omnipres ent social problem. For example, the homicide rate for blacks at 24.1% is still nearly eight times hi gher than for whites at 3.6% while blacks and other minorities are still more likely to be vi ctimized than are white s (Bureau of Justice Statistics, 2004). Predominantly black urban communities continue to have the highest crime rates in many major U.S. cities. Changes in cr iminal punishment also differentially affect minorities even though punishment trends have had a dramatic effect on the entire population (Travis and Visher, 2005; Tonry, 1995). Since the 1980’s, there has been nearly a f our-fold increase in the number of people on probation, in jail or prison, or on parole in the United States (B ureau of Justice Statistics, 2006). For example, in 1980, there were 1,842,000 peopl e under formal supervision but by 2004, it increased to 6,996,500. Statistics on federal and stat e prison populations show that there were 319,598 prisoners or 139 for every 100,000 persons but by 2000 there were over 1,421,900 people in prison. While blacks comprised less than 13% of the U.S. population, over 2,149,000 or 39% of all those who were in prison, jail, or probation by years end of 1997. Incidentally, Hispanics have had the fasting growing imprisonm ent rates and now compri se nearly 16% of the total prison population (P etersilia, 2005). 39


In addition, the black imprisonment rate is near ly five times higher than it is for whites and nearly two times higher than Hispanics. Based on first arrest statistics, one in every three black men will have been imprisoned in federal or st ate facilities during th eir lifetime (Bureau of Justice Statistics, 2006; Travis, 2005 ). Some researchers estimate that nearly 10% of all black males and 3% of all Hispanic males less than thir ty years of age were in prison in the year 2000 (Travis, 2005; Beck and Harrison, 2001). Ha ney and Zimbardo (1998) and Mauer (1990) calculated that more blac k males were in prison than in co llege (see also Travis, 2005). These numbers suggest an alarming trend in punishment in the U.S. It is one that until recently has predominantly affected black Americans but is now having an increasing affect on Hispanic/Latino Americans as well. Situating Hispanic/Latino Ethnicity The literature is replete with studies on race, crime, and punishment; however, the literature on ethnicity and crime is markedly less, particularly where ethnicity intersects with race (Rice, Reitzel, and Piquer o, 2004). Sampson and Lauritsen (1997: 364) commented on the general condition of ethnicity and crime research stating that the “future picture of criminal justice processing may be closely tied to the experiences of race or ethnic groups that have heretofore been neglected by mainstream criminological research” (Rice et al., 2004). Past studies examining ethnicity and crime have tended to take an implicitly reductionist approach by subsuming Hispanics and Latinos under racial classificat ions (Cheurprakobkit, 2000). Yet, as Herbst and Walker (2001: 30) point out, the experiences of Hispanics are different in significant ways from that of blacks. Research from the polici ng literature, for example, shows that views of law enforcement practices differ in important ways by race and by ethnicity (Rice et al., 2004). In addition, in much of the literature on ethnicity and crime shows that differences between various ethnic groups of Latin origin are often obscured by the absorption of such 40


groups under broad umbrella classifications such as Hispanic or Latino (Reitzel et al., 2005; Cheurprakbobkit, 2000). It is hard to envisage that these generali zed ethnic classifications can fully capture the diverse arra y cultural mechanisms that differentiate numerous ethnic Hispanic/Latino minority groups in the U.S. and which hold evident differences from the effects of skin color. The potential dilemma when employing broad ca tegories is that generalizations made from data on one Hispanic/Latino group do not necessari ly apply to other Hispanic/Latino groups (e.g. Mexican Americans in Texas vs. Cuban American s in Miami). Incidentally, in a development that portends well for crimi nological research, recent cha nges to the Census now allow individual multiple self-identif ication of racial and ethnic ba ckground. Classification categories thus allow for identify-specific combinations whereby individuals can report up to four racial/ethnic classifications (Rice et al., 2004; Logan, 2003). While it would be ideal to determine specific racial and ethnic combinations , much of the available data on crime does not allow such analyses beyond these umbrella groups. As in other studies on race, ethnicity and crime, the data for this study also does not allow for ideal intra-ethnic and racial classifications. That said, there is still much to be learned about Hispanic/Latinos and differences using such a classification scheme and it is better than not making a distinction at all. Social Structure, Culture, and Race/Ethnicity For a time, criminological theory on race and ethnicity was lacking with respect to the changing landscape of social advancement for blacks. A decade ago Sampson and Wilson (1995: 37) asserted that theore tical discussions of race have been “mired in controversy and silence” and that many criminologists had only offered simplistic explanations for the racial link to crime out of fear of being labeled a racist (see also Wilson, 1987). Yet, there is another potential reason 41


aside from the threat of being labeled racist for criminological theory’s sluggish response, which Bursik (1988) touches on in his soci al disorganization theory article. In the 1970’s and 1980’s, when the social st atus for blacks and other minorities were beginning to generally improve, in many communities of color crime rates remained high or were rising. In seeking to explai n this apparent paradox, researcher s turned their at tention toward structural and cultural th eories of place. However, the stru ctural view collided with an ongoing broader dialogue concerning structural-level theori es such as those in the tradition of social disorganization (Bursik, 1988; Briar and Piliavin, 1965; R obinson, 1950; Shaw and McKay, 1942) vs. individual-level theories such as Mert on’s anomie theory, and more recently strain theories (Agnew, 2001; Bursik, 1988). The nexus of this dilemma stemmed from wh at Robinson termed the ecological correlation (otherwise understood as an “ecological fallacy”) when examining the properties of to infer upon properties of individuals (Bursi k, 1988). Its exact bearing on stru ctural explanations of the race/crime correlation stemmed from the fact that structural theories had lacked widespread acceptance in criminology. This seems to have also limited earlier structural explanations of race differences in offending, at least where many crim inologists were not inc lined to provide such explanations because of practical theoretical considerations. In retrospect, it appears that individual theories such as r outine activities theory (Cohen and Felson, 1979), learning theories (Akers, 1977), and strain theories (Agnew, 200 1, 1999) were more popular, even if they too received criticism. In a simple r elucidation, the resi stance toward fully exploring structural factors was as likely to be a result of apprehensi veness to macro-level expl anations as it was to fear of being labeled a racist. 42


Cultural theories have also been invoked to explain racial variations in crime rates (Anderson, 1999). Cultural theories generally su ggest that the cultu re of a group leads individuals to accept violence or illegal behavior and internalize such attitudes toward these behaviors because of extreme deprivation and marginalization in their communities (Briar and Piliavin, 1965; Cloward and Ohlin, 1960; Wolf gang and Ferracuti, 1967; Blau and Blau, 1982; Anderson, 1999). Because blacks and other minorities are poor and subject to more extreme forms of persistent deprivation that other racial groups are not, they are more likely to handle disputes and public social relations in ways that might counter traditional societal norms even if they subscribe to middle-class values (Ande rson, 1999). Anderson calls it “the code of the street,” where the values of the street (in inner cities ) govern public behavior, to include violence and normative interactions that have their own peculia r significance. Correlates of Race and Et hnic Differences in Crime The study of race/ethnicity, crime, and punishme nt situates within commingling of factors that differentiate both race/ethni c-specific crime and imprisonme nt rates (Walker et al., 2004; Krivo and Petersen, 2000). Sorting out how thes e effects operate across the crime and control spectrum is important to a ny study on race/ethnicity. Criminological efforts to understand aggregat e crime differences across race/ethnicity generally fall into three areas; group-based and individual differences in arrest rates and groupbased differences in punishment (or attitude s thereof) (Cohn, Barkan, and Halteman, 1991). The main thrust of race/ethnic differences in crime have typically come from macro-structural and cultural explanations including society structural-level factors such as economic inequality and relative deprivation, extreme poverty, unemploymen t and, human and social capital; community factors such as residential se gregation and community disorganization; family structure and processes including single-parent and female-headed households; a nd also, the impact of drugs 43


and gangs (Clear, Waring, and Scully, 2005; Wa lker et al., 2004; Kri vo and Petersen, 2000; Parker and McCall, 1999; Sampson and Wilson, 1995; Harer and Steffensmeier, 1992; Bursik, 1988; Blau and Blau, 1982; Shaw and McKay, 1942). Moreover, recent cultural and structural theories of race have garnered a great deal of attention by prevailing over individual explanations. Theories by Sampson and Wilson (1995), Sampson and Bean (2006), and Anderson (1999) address factors such as communi ty organization (i.e. co llective efficacy) and the culture of the inner city that seem to hol d important keys to understanding race and ethnic differences in offending. On the crime control side, however, many of the e xplanations of racial differences in arrest and imprisonment rates given in the past have their roots in th e historical legacy of racial discrimination and institutionali zed bias in the criminal justice system, including potential discrimination in court or criminal processing proceedings (Pope, Lovell, and Hsia, 2002; Cohn et al., 1991; Klein, Petersilia, a nd Turner, 1990), differential law enforcement tactics targeting minority communities (Reitzel and Piquero, 200 6, Walker, 2001; Weitzer and Tuch, 1999) and crime control policies differentially affecting minority populations (T onry, 1995; Petersilia, 1983). Yet, some have argued that victimization data undermine claims of systemic racism since they consistently show black over-involvement in crime (Gottfredson and Hirschi, 1990). Empirical studies, which investigat e the correlates of specific theories, would therefore need to include a range of factors that address both race-based differen ces in offending, and differences in arrests and imprisonment, including differential treatment across the criminal justice system. All of which would inhere whether such factors affect different raci al and ethnic groups in similar or dissimilar ways (Krivo and Petersen, 2000). 44


Dynamic Theories of Race/Ethnicity and Crime Moffitt, Gottfredson and Hirschi, and Sampson and Laub all consider race/ethnicity in their theories. Moffitt (1994) takes up the diffe rences in sub-group offending, which illustrate that the crime rates for blacks, particularly young black males, is higher than for whites. She asserts the race differences in crime rates are identifiable by higher prev alence rates of black adolescent limited and life-cour se persistent offenders. There are more black life-course persistent o ffenders (i.e. higher within race percentages) because of institutionalized prejudice and poverty. Moreover, because blacks are much more likely to be poor, they are more likely to suffer pr enatal problems as an outcome of exposure to environmental toxins and coupled with mothers’ poor nutritional habits, it places black children more at risk for maladaptive development. Th e higher rates of poverty also lead to more instances of family disruption and weaker bonds, trouble in school, unemployment, or underemployment, and other outcomes of poverty a nd marginalization that predispose children to “aggressive interpersonal beha vior.” Moffitt (1994: 39) stated, “for poor black children, the snowball of cumulative continuity is anticipated to begin rolling earlie r, and it rolls faster downhill.” In other words, persistent differen ces in crime are a matter of both degree and number. For Moffitt, the potential race and ethnic differences in persistent behavior derive from the higher rates and more extreme forms of soci al, economic and health problems that blacks and other minorities face compared to whites. This, in turn, leads to higher rates of pathological neuropsychological problems and other conditions that predict life-course persistent offending. Sampson and Laub (2004, 1993) generally take a similar position on race as Moffitt; however, they put more emphasis on informal social controls rather th an neuropsychological deficits. They suggest that because of weak bonds to family, authorities, and school, juvenile delinquents who become ensnared in the system are labeled, which also has a cumulative effect 45


throughout their lives. In other wo rds, deficits in life pile up faster, which make it tough to “change” away from crime/criminality. The over-representation of blacks in official crime statistics would therefore be a result of blacks being more likely to suffer from pressing poverty and all that it encompasses. Consistent with their self-control thesis a nd their deliberate “tension” causing style, Gottfredson and Hirschi (1990: 151) disagree with cultural and st ructural theories, systemic racism, and opportunity differences as reasons for racial variati on in crime on the premise that victimization data refute such explanations. They suggest that research on racial differences in crime should focus on difference in child-rearing practices such as “m onitoring, recognizing, and correcting evidence of antisocial behavior ” (Gottfredson and Hirschi, 1990: 153). At a time when the more controversial theoreti cal problems concerning race, structure, and culture, seems to have subsided, the empirical literature has yet to produce any overwhelmingly convincing explanations of race differences in crime (Gottfredson and Hirschi, 1990). Nevertheless, two things are clear. First, minority Americans, particularly blacks, continue to confront the impediments of discrimination insi de and outside the criminal justice system. Second, racial (or ethnic) group membership alone cannot fully explain the intricate milieu of the American experience. Overall, theories and empirical studies of race and crime would lead one to hypothesize that blacks and to a lesser degree Hispanics as being more likely to persist in offending than would whites. This is not a hypothesis bourn out of speculation but ba sed on a large body of evidence. The concentration of poverty and disa dvantage, which correlate to higher rates of offending and arrests for minorities, suggests that active offenders of color face additional burdens or more extreme forms of burdens that l ead to race differentiated crime rates. Thus, the 46


predictors used in this study’s models are ones th at have been found to affect all offenders, but not equally. For example, family disruption and psychological status factors should have more influence on black and Hispanic offending rates th an whites simply because they are more likely to affect minorities (i.e. family disruption is mo re likely to occur in black and Hispanic families than white families). They are more likely to incur more extreme forms of negative factors than whites are. 47


CHAPTER 4 THEORETICAL FRAMEWORK Four Theoretical Models Concerned with the high crime rates in New York City in the 1970s, two engineers, AviItzhak and Shinnar (1973) (see also Shinnar and Shinnar, 1975) created a relatively simple stochastic model to predict the reductive e ffects of incapacitation on crime (Nagin and Land, 1993: 329). In turn, their predictive model partly served as the model for the development of the criminal career model. In 1986, a National Research Council panel of experts, which comprised a group of nationally recognized criminologist s headed by Alfred Blumstein, came together to develop the criminal career model. From this NRC collaboration, the scholars’ produced a two-volume publication, Criminal Careers and “Career Criminals,” that offered a new model for understanding criminal offending over time (B lumstein, Cohen, Roth, and Visher, 1986). The NRC panel addressed important questions about two separate, bu t related concepts: criminal careers and career criminals (Blumstein et al., 1986). This came on the heels of rapidly increasing imprisonment rates in the Unite d States—doubling between 1972 and 1983—and the crime rate surpassing 13 million for index crimes. The heightened political focus on the national crime problem and the seemingly vain incapacita tion policies for reducing crime rates led policymakers and academics to join in developing new measurement tools for policy development. The applicability of the criminal career model wa s critical in a variety of ways. It provided academic criminology with a model for classifying and measuring the criminal career by emphasizing the need to investigate different dimensions of offending such as prevalence, frequency, duration, initiation, and termination (Farrington, et al., 2003). It also provided a 48


complimentary tool for identifying career crimin als. Under this definition, a unique sub-group of offenders that exemplify the most frequent and stable criminal behavior across the life-course. Insofar as the criminal career model is a prospective model for examining the dimensions of criminal career, it can be considered a lif e-course or developmental model. However, Blumstein and colleagues (1988) argue that it sh ould not be considered a theory. Where the criminal career model has an admitted policy orientation and clear expl anatory limitations, lifecourse and developmental theories better addr ess behavioral causes that link childhood and adolescent development to adulthood behavior. Laub and Sampson (2001) suggest elsewhere that the criminal career framework may be stagnant in terms of advancing our knowledge of criminal behavior because of its specific policy focus. The criminal career model is, nevertheless, an important starting point for this study de spite its limited theoretical vitality. The Criminal Career Model Blumstein colleagues (1986) define the crimin al career as the longitudinal pattern of crimes by individuals from the time they commit their first crime to the time they commit their last (see also Piquero et al., 2003). Career criminals, on the ot her hand, are the highest rate, persistent offenders who contri bute most to aggregate crime ra tes (Blumstein et al., 1986: ix ). Policy implications notwithstanding, the importance of distinguishing between the two is critical to the task of understanding the model’s utility and structuring research around it (Blumstein et al, 1988). 1 The former is the primary focal point for which the authors engaged in the NRC panel while the latter is a potential outcome of the form er. Indeed, the very idea of a career criminal 1 In the past, Blumstein and colleagues have argued that Gottfredson and Hirschi “obfuscate” the contextual differences between the criminal career and career cr iminal constructs (see Blumstein et al., 1988). 49


has a longstanding history in criminology, extend ing back to Quetelet in 1831 and underscored by Wolfgang and colleagues in 1972. Piquero and colleagues (2003: 377) state that the criminal career model “recognizes that individuals start their criminal activity at some age, engage in crime at some individual crime rate, commit a mixture of crimes, and eventually stop.” It does not propose, as Gottfredson and Hirschi (1988) have suggested in the past, that offenders make a living through criminal activity, such as with a more conventional understand ing of a career (although so me might). Nor does it suggest any “particular patterned trajectory” (Blumstein, Cohen, and Farrington, 1988b: 60). Rather, the criminal career model serves as a framework for systematically arranging research around sequential criminal even ts (Piquero et al., 2003: 378, Blumstein et al., 1988). The criminal career model encompasses four key com ponents: participation, fr equency, career length, and seriousness, as well as a number of other im portant dimensions, such as prevalence and cooffending. Each dimension is discussed below. Blumstein and colleagues define participa tion as the “distinction between those who engage in crime and those who do not” or “the frac tion of a population that is criminally active” before some designated age or during some par ticular observation period (Blumstein et al., 1986: 1, 17). Assessing participation is dependent upon the scope of criminal acts under consideration and sensitive to the leng th of observation. For example, measur ing participation for more serious criminal acts such as violent offenses shoul d decrease the number of active offenders while employing longer observational periods should increas e the number of participants, since more offenders that only commit crime rarely and mo re offenders that initiate offending during the observational period will be included (Blumstein et al., 1986 : 18). Blumstein and colleagues found that the most glaring disp arities were between males and females (1986). In a 1989 study, 50


Farrington found that 96% of the sample had repor ted committing at least one offense from a list of ten offenses by age 32. Frequency is an individual’s offense ra te and represented by the Greek letter “ ” (Lambda) (Blumstein et al., 1986: 55). Since frequency cente rs only on active offe nders identified through the participation measure, it is useful the think of aggregate crime rates as being partitioned into two components, participation and frequency; for which frequency derives from an active criminal subset of any given population (Piquero et al., 2003). Moreover, because of the relative irregularity of individual offendi ng for most offenders, the distribu tions of individual frequencies are highly skewed. Blumstein and colleagues (1986: 94) report that the “median offender engages in only a few crimes per year, but the most active 10 percent of offenders commit crimes at rates that may exceed 100 per ye ar.” Ultimately, this has led mo st researchers to concentrate on chronic offenders (Piquero et al., 2003: 379; see also Greenwood and Turner, 1987; Blumstein, Farrington, and Moitra, 1985; Chaiken and Chaiken, 1982). Duration is the length of time between the initiation into criminal offending and its cessation. It encompasses initiation (age of onset), termination, and persistence. Duration and its ancillary components have signifi cant implications with respect to modification of criminal careers (Blumstein et al., 1986). Typically measured as the commission of at least one crime per year, high-rate offenders may commit as much as one-hundred crimes pe r year and low rate offender might skip years intermittently (Blumstein et al, 1986). Past research shows that age of onset typically occurs between 8 and 15 years of age (Farrington, 2003). There is also a link between early onset of offending and the durati on and frequency of a criminal career. Generally, those who begin o ffending earlier have longer and more intense careers (i.e., they commit more crimes when active) (Farrington, 2003; Le Blanc and Frechette, 51


1989). Since there are potential caus al differences that affect intensity of offending, it might be necessary to make distinctions between high rate and low rate persistent offenders, or at a minimum, to investigate whether risk factors op erate similarly or differently across different offending rates. The termination of a criminal career has consid erable salience as well but the definition of what constitutes desistance from a criminal car eer is unclear. Individua ls may commit crime at different rates but than also st op in different ways and for diffe rent reasons. Some offenders might slowly deescalate, eventually stopping. Ot hers might intermittently commit more or less crimes while showing a pattern of desistance, while others sti ll might abruptly stop. Laub and Sampson point out that desistan ce, although addressed widely in the theoretical literature, is treated as an event where people just stop; what smokers might call going “cold turkey.” Most empirical studies have also treated desistance as a “discrete state” (Bushway, Piquero, Broidy, Cauffman, and Mazerolle, 2001: 491-492). Howeve r, some researchers have recommended treating it as a process [that cu lminates in termination from offending] (Bushway et al., 2001; Maruna, 2000). In other words, it is a process wh ere the pattern of desis ting overtime is itself desistance. Part of the reason why termination of a crim inal career is important , besides the obvious theoretical implications, stems fr om the reality that criminologi sts have yet to agree on the meaning of desistance or how to measure it . Sampson and Laub (2001) argued that defining desistance has been hindered by conceptual a nd measurement problems. Piquero et al. (2003: 379) noted, “There is much less research on the dur ation of criminal care ers, primarily because of the difficulty involved in determining the true end of an individual’s criminal career.” More 52


recently, Piquero and colleagues (2004) report th at the average career duration among a sample of serious offenders was about seventeen years. The seriousness/crime-type dimension centers on the varying offense types committed by active offenders. Differentiating serious or pred atory criminal offenders from less serious offenders has particular implications for theory and policy (Blumstein et al., 1986: 76). Equally important is the pattern of offenses for a ny particular offender. Questions about dynamic processes such as offending specialization or the tendency for offenders to move to more serious offense types as offending continues (Piquero et al., 2003: 380), and escalation, which is the propensity for criminals to progress toward committing more serious crimes over time, hold considerable meaning in the criminal career model. Research has shown a tendency for offenders to escalate in crime seriousness and to specia lize within broadly defined crime types (e.g. property crime or violent crime) even though they may “engage in a great diversity of crime types” (Piquero et al., 2003: 380; Blumstein et al., 1986: 94). The co-offending feature situates the offender within groups or accomplices in offending. In Volume II of Criminal Careers and “Career Criminals , Reiss suggests that positive reductive effects on crime depends not only on addressing i ndividual offenders but also on diminishing the status of the offender within th e group and that of the other member’s behaviors (Reiss, 1986). From a policy standpoint, the co-offending dimens ion is important, es pecially with younger (juvenile) offenders since research has shown th at co-offending is more typical in adolescence than in adulthood (Reiss, 1986). Furthermore, co-offending seems to decrease as offenders’ age, not as a result of incapacitati on. Piquero and colleagues (200 3: 380) stated, “Although the decline in co-offending might, at fi rst glance, be attributed to co-offenders dropping out, it seems to occur because males change from co-offending in their teenage years to lone offending in their 53


twenties.” In other research, Reiss and Farrington (1991) found that co-offending decreased with age, not because of “selective attrition” but b ecause the offenders themselves changed. They further concluded that there is a tendency to either offend solo or to co-offend and that the most persistent offenders did both about equal (Reiss a nd Farrington, 1991). Dual Developmental Taxonomy In 1993, Terrie Moffitt presented a dual taxonomic theory in order to settle what she argues are two logically inconsistent facts. The first is that antisocial behavior is a stable trait over the life course and the second is that its prevalen ce alters over time; temporarily increasing sharply during adolescence (Moffitt, 1993). To Moffitt, because there is such a high prevalence of delinquency during adolescence, it masks two e tiologically discrete groups. The first, life-course persistent antisocial behavior (LCP), predicts that the coll ective effects of neuropsychological problems during childhood interacts with criminogenic environments (that could be a result of having an antisocial child in the first place), whic h produces a pathological antisocial personality or continuity in antisocial behavi or across the life-course. The second, adolescence-limited antisocial behavior (AL), predicts that for most people, antisocial behavior is a result of typical youthful rebellion and status discord, and it is limited to adolescence. The defining characteristic of the LCP is stability across ones lifespan. The seeds of antisocial behavior are present at or soon afte r birth and there is a marked gradation in seriousness of antisocial behavior as the child develops. For exampl e, they might bite or kick often at age 4 but by age 16, they are stealing cars. Other children/adolescents might commit many of the same crimes as the LCP depending on the situation or a sp ecific conflation of conditions, however, for the LCP; there is a defin itive gradation in maladaptive behavior that is underpinned by a single antisocial tendency, termed heterotypic continuity (Moffitt, 1993). 54


Moffitt asks what explains the variability of antisocial behavior in this typology. Her answer is that the gradation in seriousness links to neuropsychological factors, which are outcomes of early (even prenatal) developm ental problems. The disruption of neural development of the fetus during pregnancy, infa ncy, or early childhood development eventually leads to a host of negative outco mes later in life. She cites a number of studies that link neuropsychological deficits w ith complications during pregna ncy, physical abnormalities and fetus or mother’s exposure to environmental toxins (Moffitt, 1993). In fact, she argues, the relationship between neuropsychol ogical problems and antisocial behavior is among the most robust findings in developmental criminology (Moffitt, 1993: 102). Three are two primary types of neuropsychologica l deficits that correlate with persistent antisocial behavior. The first is verbal func tioning, which centers on problems with listening, reading, and expression of speech (Moffitt, 1993:103). The second is an executive function deficit. Typically, those with executive function deficits suff er from “comportmental learning disabilities,” which are i ndicative of impulsivity or inattentiveness. According to her theory, neuropsychological defi cits do not necessarily translate into lifecourse persistent antisocial be havior since many children who suffer from such deficits go on to lead normal, law-abiding lives. The interaction of bad poor parenting of a child with neuropsychological deficits eventually leads to pathological conditions. She writes that the challenge of coping with such a si tuation can trigger a series of disastrous interactions between parent and child, which may vary by social class or other grouplevel factors (i.e. factors that contribute to bad parenting) but ultimately se rves as the mechanism for continued offending. Her second typology, adolescence-limited anti-social behavior, is distinguished by discontinuity or intermittency. AL types are typically delinquent during their teens when 55


aggregate offending rates are highest but do not typical ly exhibit evidence of persistent antisocial behavior beyond childhood. Although discontinuity is the “hallmark” of the AL type, it does not undermine the fact that they often commit very serious crimes or a large number of crimes during adolescence (1993: 113). By definition, br ief offending career of an AL is limited to adolescence and is often inconsiste nt in that there are usually periods of inactivity. Moreover, AL types are primarily distinguished by their persona lity. They show signs of more rational thought prior to committing crimes. They are also oppor tunists and may make choices dependent on a cost/benefit analysis. AL types also differ from LCP types in bot h when and why they initiate and terminate offending. The pathway to AL offending begins with imitation of LCP behavior and, because AL types are more sensitive to the gap in maturity such that their sexual drives, and financial and social dependency upon parents, conflicts with th e status proscriptions that are placed upon them (e.g. age restrictions on such behaviors). The di sjuncture ultimately reinforces delinquency and leads to rebellion as a means to show independenc e. However, the AL typology suggests that this period of antisocial behavior is short-lived and often inconsistent. AL types eventually assume conventional roles as the gap in their status diminishes because they reach the legal age to make important decisions for themselves and to engage in behaviors that are so cially, if not legally prohibited when they are minors. Moreover, the structure and culture of modern U.S. society pressures them to take on roles that are reflect ive of adulthood during this period and coincides in part with their maturational development. In short, Moffitt’s typology pr edicates the existence of two distinct types of offenders and that, despite some potential similarities in behavior during adolescence, each type of offender offends for very different reasons. 56


Life-Course Theory Sampson and Laub’s (2004, 1993) life-course theory centers on the structural and process risk factors, formal and informal social controls, and age-graded life events that affect criminal offending. Though they recognize that there can be stability between ear ly childhood antisocial behavior, adolescent delinquency, and adult criminality, they argue that throughout the lifecourse, important age-graded and non age-related lif e events can alter an individual’s trajectory toward or away from crime (Sampson and Laub, 2004; Benda, 2003; Uggen and Staff, 2001; Warr, 1998). For example, a high-rate offender might get married, which increases their stake in conforming to societal rules. It could also be that someone who abides by the rules might experience a traumatic event that sets them on a course of law-breaking. Recently, the two scholars have recently come to claim that they conceive their theory as occupying a middle ground in the realm of criminol ogical theory because it does not predict the existence of groups as do many developmental th eories. That is, the sa me theoretical model holds for all offenders. Alternat ively, it also requires more than simply age in determining offending probabilities (see Sampson and Laub, 2003). There are three main features to Sampson and Laub’s age-graded theory, which partly builds upon a reconsideration of Hi rschi’s control theory in a dynamic combination with labeling theory. First, the social bonds element c onsiders the significance of bonding with others, especially intimate others and t hose in authority, on criminal pr opensity. Important informal and formal social institutions such as family, school, marriage, and work affect individual’s social bonds differently at different points across the life -course. They also posit that “early (and distal) precursors to adult crime (e.g. conduct disorder, low self-control) are mediated in developmental pathways by key age-graded institutions of informal and formal social control, especially in the 57


transition to adulthood (e.g., via employment, military service, official sanctions)” (Sampson and Laub, 1993). Second, Sampson and Laub unite “continuity and change within the context of a sociological understanding of crime through life” (Sampson and Laub, 1993). This feature delineates how trajectories and transitions interact to produce turning points in the life-course (Sampson and Laub, 1993). Again, even though they agree with the no tion that there is behavioral stability over time, they contend that turning points can modify trajectories (or criminal propensity) thereby redirecting one’s path. Importantly, they conceptualize a turning point as being “incremental and age-related progressions and events, which carry forward or set in motion dynamic processes that shape future outcomes” (Sampson and Laub, 1993). This final feature offers a diffe rent lens for viewing trait-base d explanations of persistent offending. Here, Sampson and Laub reconsider both labeling theory and Hirs chi’s (1969) control theory within a developmental framework. They begin by questioning why psychological traits such as temperament or activity level, which ar e typically thought to be primarily rooted in biology, show relatively low stab ility over time whereas aggressi on, which is thought to be less biologically based, shows relativ ely high stability. Their answ er lies in the response to aggression such that because of its antisocial effects aggression tends to be met with severe retaliatory actions. It also centers on the cumu lative disadvantage from the progression of negative outcomes of antisocial interaction. When maladaptive interactions begin in childhood, the negative effects accumulate over time, which th en produces negative developmental effects. In the process, both cumulative continuity and interactional conti nuity are sustained. Sampson and Laub also borrow from Nagin and Paternoster’s (1991) state dependence argument—which is that past participation in crim e is a strong predictor of future participation— 58


but employ it in a developmental framework. Wher eas Nagin and Paternoste r suggest that prior criminal offending has a strong and direct beha vioral effect on the probability of future offending, Sampson and Laub offer that prior de linquency systematically erodes individual’s social bonds to various institutions and to society in adulthood, in turn, affecting adult criminal offending (Sampson and Laub, 1993; Na gin and Paternoster, 1991). More recently, however, Laub and Sampson (2004) extend their theory to include personal human agency and the effects of routine activit ies on crime. They assert that the underlying causes of crime are the same for all people, though there may be a single pathway or multiple pathways to either persistence or desistance. And, even if the specific manifestations are different in leading to different types of crime, the pathways lead ing toward or away from crime result from the convergence or absence of info rmal social control mechanisms, human agency, and routine activities. In other words, individual s will commit crime when there is an absence of these mechanisms but when present, such as at a particular turning point in life; it will put them on a trajectory toward desi stance (Laub and Sampson, 2004). General Theory Gottfredson and Hirschi are, perhaps, the two biggest critics of the criminal career/developmental/life-course perspectives, and if not, they are at least the most vocal. This becomes apparent when considering their general theory along a theoretical continuum with theirs’ on one end, developmental and criminal care er approach at the ot her end, and life-course and integrated theories somewhere near the middle. For Gottfredson and Hirschi, explaining the age/crime relationship does not require elaborate theory or sophisticated longitudinal data such as in the case of developmental and life-course theories. Also, they summarily dismiss the need to differentiate between prevalence and frequenc y. They argue that there is not much to be gained by making such a distinction. 59


Gottfredson and Hirschi assume that when child ren are not properly so cialized to resist immediate gratification, they will be more likely to indulge in criminal or analogous behaviors. This is because crime typically offers immediatel y gratifying rewards and requires little planning or proficiency (Pratt and Cullen, 2000; Gottf redson and Hirschi, 1990). They also focus on behaviors analogous to crime such as drinking, smoking, drug use, phys icality, and thrill seeking, which they claim predict criminal behavior. Thou gh it is not clear exactly when such behaviors would occur, even when presented with opportunit y, their argument that such behaviors predict or are predicted by some underlying trait is suspect. In fact, those most likely to engage in such behaviors do because of one all-inclusive trait— low self-control—that, accordingly, is the root cause of all crime and analogous behaviors. Low self-control can reflect the absence or ine fficiency of socializing institutions (e.g. the family and schools) to properly socialize child ren before they reach offending age or it can indicate neuropsychological impairments (Mitchell a nd MacKenzie, 2006). Their conceptualization of low self-c ontrol draws partly fr om the classical school perspective that humans are all rational actors concerned primar ily with their own self-interest as well as Hirschi’s (1969) earlier control theory, which places the emphasis on institutions and bonds in socializing children. This seems to point to two reciprocating el ements of low self-control. First, low self-control signals an individual’s inability to choo se pro-social, legal behavior. Second, it also indicates patholog ical deficiencies in resist ing the seduction of crime or immediate gratification when cr iminal opportunities present them selves (Pratt and Cullen, 2000). As a property of the offender, low self-control is a trait established in early childhood, usually by age eight, and even though it is believed to reflect a stable, underlying propensity toward crime, it does not necessarily follow that crime will automatically result (Pratt and Cullen, 2000; 60


Gottfredson and Hirschi, 1990). Ra ther, Gottfredson and Hirschi ar gue that impulsivity, thrillseeking, short temperedness, self-centeredness, inclination toward physic ality, and opportunism characterize low self-control individuals and that such individuals always be more likely to take advantage of opportunities when they present themselves (Mitchell and MacKenzie, 2006; Gottfredson and Hirschi, 1990; Piquero and Bouffard, Forthcoming ). Gottfredson and Hirschi’s theory has rece ived a prodigious amount of attention by criminologists, however, with th is attention has come increase d, and at times, very sharp criticism (Piquero and Bouffard, Forthcoming; Geis, 2000). Some of the more prominent criticisms suggest that general theory is tautological (Akers, 2000), that it dismisses Hirschi’s earlier bonding theory or subsumes indivi dual self-contro l under the attachment bond (Longshore, Chang, Hsieh, and Messina, 2004; Akers, 2000); that it contradi cts itself in defining and applying criminal behavior (Geis, 2000); that it is not portable across different situations, cultures, or crime types, part icularly white-collar crime (H orney, 2006; Simpson and Piquero, 2002; Geis, 2000; Reed and Yeager, 1996). In othe r words, it is not general. Also, withinindividual levels of self-cont rol and between-indivi dual levels of self-control are stable throughout the life-course irrespective of whet her or not individuals’ are committing crime (Mitchell and Mackenzie, 2006; Turner and Piqu ero, 2002; Akers, 2000); and finally, that they make dubious assumptions based on questionabl e or lone studies (Geis, 2000; Warr, 1993). Despite that their theory is, perhaps, the most empirically investigated over the last decade, overarching all these criticisms or linking to all of them at least, is the problem of how to measure self-control, which clear ly stems from its conceptualization. Gottfredson and Hirschi suggest that it should be measured with behavi oral variables indicative of low-self control because of the problems with self -report attitudinal measures, such as Grasmick and colleagues 61


self-control index (Piquero and Bouffard, Forthcoming; Grasmick, Tittle, Bursik, and Arneklev, 1993). However, disagreement over measuremen t methodology has led to further blurring of what self-control actually capture s, even though most studies seem to find some support for it (regardless of how it is operati onalized). These problems have subsequently led Hirschi to submit a modified version of lo w self-control (Piquero and Bouffard, Forthcoming; Hirschi, 2004), which appears to capture not just individual self-control, but draws from Hirschi’s original bonding theory, ra tional choice, and the expe riential effect. Hirschi now asserts that self control is basically the ability (or tendency) to co nsider the full array costs of specific behaviors, which in turn, shifts the focus from “long-term im plications of the act to its broader and often contemporaneous implications” (Hirschi, 2004: 543). 62


CHAPTER 5 LITERATURE REVIEW OF PERSISTENCE AND DESISTANCE Overview Research on persistent offending and desistan ce from offending within the criminal career framework is growing. Identifying risk factors for those most likely to persist and desist bear import on theoretical and policy considerations. For example, identifying a career criminal subgroup would lend credence to developmental theori sts’ contentions that the distribution of the age/crime curve is chiefly due to this small group of high-rate offenders. At the same time, it would challenge the propensity theorists’ cont ention that the distribut ion of crime by age is simply a function of age by calling into question th eir lack of explanation of different types of criminals. From a policy standpoint, identifying individu al persistent offenders and related risk factors can help shape prevention or intervention measures before such an offender becomes ensconced in a lifetime of crime, this review of the research has identified nearly a dozen studies that examine the causes a nd correlates of life-course persis tent offending. Most have found some evidence of association between early childhood aggression, neurops ychological impairments or poor childhood environments and later criminal be havior. Many of the studies have examined age of onset and offending, although the results are mixed. This might be partly an outcome of differing conceptualizations of the starting point. For example, prior research by Moffitt (1993) and Dean and colleagues (1996) posit a somewhat different conception of age of onset. They characterize early delinquency as occurring at the beginning of adolescence, closer to 1314 year age range than 17-18 year range. Ge and colleagues (2001) conceptualizati on of early onset as beginning pr ior to age 17 and that late onset occurs after age 17 is a re lative distinction. Indee d, it is certainly early and late onset as 63


conceptualized. However, with respect to the developmental framework put forth by Moffitt and others, which might not detect important and nuanced differences between certain types of offenders, it could be that a c onceptualization based upon an earlier age might yield stronger results. In short, what follows is a review of the important research on pe rsistence and desistance that has emerged in the extant literature re cently. Many of these studies focus, not only on persistence and desistance, but on th e issues of age and age/onset that have come to characterize this line of research. Empirical Studies of Persistence Paternoster analyzed a sample of 1,600 s ophomore and junior high school students with respect to relatively minor delinquent offenses (marijuana or alcohol use, petty theft, and vandalism). He was interested in the effects of two types of sanction threats; absolute and restrictive deterrence. Absolute deterrence is a sa nction threat that deters people from first time or continued offending and restri ctive deterrence is that which might dampen or attenuate the rate of offending for participants’ cu rrent offending (Paternoster 1986: 291). A longitudinal panel design of students from nine high schools in the Columbia, South Carolina area who were surveyed at the beginni ng of their freshman and sophomore years was employed. The survey instrument included exogenous background factors pertaining to participation and frequency such as gender, family structure, and welfare assistance. The models also included variables relating to opportunity (e.g. parental s upervision, peer delinquency) to non-delinquency, bonding, informal sanctions, formal legal and moral beliefs in opposition to specific delinquent acts (Paternoster, 1986: 294). No explanatory factors achieved statistical si gnificance across the four offenses except for prior criminal participation. This might refl ect different causal mechanisms operating for different types of offenses or a latent delinquency construct oper ating across delinquency types. 64


His analysis did, however, uncove r two distinct groups based on whether respondents had prior experience in a given offense or not, which re lates directly to the onset of offending and decisions to persist or desist thereafter. A broad interpretati on of the findings suggests that persistence in and desistance from offendi ng is influenced by the onset of offending. Nagin and Farrington (1992) studied whether the inverse relationshi p between onset age and persistent offending remained after controlli ng for unobserved persistent heterogeneity. They found that onset age and offending persistence coul d be attributed to persistent heterogeneity, which is actually consistent with Gottfredson and Hirschi’s position. They also found that onset did not vary with intelligence or a daring disposition while parental behavior showed a definite effect on the probability of participation, which correlated invers ely with onset age. Similarly mixed, parental separation and interaction meas ures of parental separation and parental criminality differed in both direction and significance. Specifically, parental separation was positively correlated with onset age while the interaction term was negatively correlated with onset. The researchers also looked at the effects of poor child-rearing behavior of the parents, which they found to be positive and significant with early onset but insignificant for late onset. In 1994, Moffitt, Lynam, and Silva also inve stigated the relationship of onset age, neuropsychological risk factors, and male delinquency using a bi rth cohort aged 13-18 from the Dunedin Multidisciplinary Health and Development Study. They attempted to determine the predictive strength of neuropsychol ogical test scores on delinque nt behavior and whether poor neuropsychological status predicts delinquency commenced or accelerated after age 13, whether it indicated early onset age, and whether de linquency was stable across ages 13, 15, and 18 (Moffitt, Lynam, and Silva, 1994). 65


Their analyses found a link between early neur opsychological status and later delinquency and that the driving factor in IQ on later delinquency was poor verbal ability. Regarding persistence, they reported that early poor neuropsychological st atus predicted later delinquency and that delinquency was stable over time. However, it should be noted that the age range for this study was censored at age 18 and therefore, while it does not necessarily diminish their findings, it would be useful to st udy the stability of offending or antisocial behavior through later adulthood. Dean and colleagues (1996) studied the crim inal propensities of discrete groups of offenders and their persistence in crime using a cohort of offenders past age 16. The sample for the study was drawn from a population of releasee s from the North Carolina Division of Youth training schools from 1988 to 1989. They found so me support for the typological position (i.e. Moffitt and Patterson’s theories) as well as some support for the criminal propensity position (i.e. Gottfredson and Hirschi’s theory). For example, they found that higher numbers of prior adjudication increased the risk of criminal persistence in the late adjudication group but it decreased for the early adjudication group. They also found that child abuse was crim inogenic for the early first adjudication group but not for the late group. As such, the non-fi ndings for the later age groups seemingly lend support to the criminal propensity hypothesis. They did not find s upport for the other variables in the model such as early learni ng disabilities or single parent households (Dean et al., 1996). Bartusch and colleagues (1997) pitted Moffitt’s taxonomy versus Gottfredson and Hirschi’s general theory. Recognizing, as others have, the similarities betw een the theories, they suggest that for developmental theories to bett er explain crime than more parsimonious general theories, they must be “demonstrably superi or” in empirically explaining the same data 66


(Bartusch et al., 1997: 17). The authors examined underlying structures of risk factors that impact upon antisocial behavior by garnered from a variety of different reporting sources. They tested whether one underlying trait that is ag e independent versus two traits derived from Moffitt’s theory. They found stronger support for Mo ffitt’s since the same trait did not predict offending in childhood and adolescence as Gottfre dson and Hirschi would suggest. Moreover, they found that the patterns of behavior that spring from underlying pr opensity were important for childhood but that adolescents were influenced more by peer factors. Finally, they also found support that earlier signs of antisoc ial behavior linked to more violent behavior in adolescence while late starters (in this case , those who began in adolescence) predicted non-violent offending. Thus, they f ound support for the notion that underlying traits matter; however, they also argue that qualitative differences ar e meaningful. That is, whereas Gottfredson and Hirschi suggest that low self-control explains all crime at all ages whereas from Moffitt’s perspective, this might be true for some types of offenders but other factors are important, particularly for those who begin offe nding in adolescents. In short, they found support for the idea of persistence through adolescence for a particular type of offender; an early starter offender more prone to violence who suffe rs from neuropsychological problems. Aguilar and colleagues (2000) i nvestigated age of onset a nd neuropsychological to in investigating persistence. Their sample of 180 participants of normative and non-normative development in a high-risk urban population of firstborn children and their mothers was drawn from an ongoing 20-year longitudinal study (Agui lar, Sroufe, Egeland, and Carlson, 2000: 112). Independent measures comprised a wide variety of neuropsychologi cal scale measures such as the Brazelton Behavioral and Neurological Assessme nt Scale, Casey Infant Temperament Scale, 67


Wechlser Preschool and Primary Intelligence Sc ale, and the Peabody Individual Achievement Test. Their analysis yielded important findings . They found support for Moffitt’s typology in that there can be distinctions made between LCP and AL behavior types. However, perhaps more importantly, the strongest effects were not adol escent temperament or neuropsychological factors but rather, measures of psychosoc ial history. Furthermore, they reported that neuropsychological functions within the first four years were not significant in distinguishing differences in antisocial groups, which is inconsistent with earlier studies. This held even in the presence of numerous different measures such as early temperament predictors. Psychosocial environment measures, however, did predict risk probabilit ies differences between LCP and AL types (Aguilar et al., 2000). In short, their findings seem to suggest th at early psychosocial environment (such as family problems) seem to play more of a role in predicting antisocial be havior in the short run. Given that the researchers did not employ a longer time period for their models, it is not possible to infer beyond the young adolescent years in pr edicting antisocial beha vior. Likewise, since they did not look at the relationship between antisocial behavior during the formative years and later criminal behavior, no conclusions beyond the presence of two distinct groups can be drawn. Piquero and White (2003) studied the relati onship between neurops ychological factors (cognitive ability) and life-course persistent offending but among bl ack residents in Philadelphia (from birth to the late 30s). Da ta analyzed for the study came from the Philadelphia subset of the National Collaborative Perinatal Project (NCPP). The basis of the study was to test empirically Moffitt’s developmental taxonomy in predicting lif e-course persistent offending. The dependent variable was operationalized in several ways such as ‘earliness of onset’, ‘individual stability 68


across developmental stages’, ‘rate of offendi ng’ while the independent predictors included measures of ‘cognitive ability or neuropsychological risk’. Employing a series of logistic regression models, they found substantial support for Moffitt’s typology. For example, males, participan ts who had more disciplinary problems, and those born to single mothers were more likely to persist in offending over time. These findings held even after operationalizing alternative measures of persistent offending, and for different measures of cognitive ability. Put differently, those who had high disciplinary problems and low cognitive ability scores were more likely to persist in offending compared to those who had higher cognitive ability scores and less disciplinary problems. In a ddition, they reported that that their findings replicated those of Kratzer a nd Hodgin’s (1999) study of Swedish longitudinal cohort in differentiating effects of similar measures between life-co urse persistent offenders from other offenders and non-offenders alike (Piquero and White, 2003). Recently, Piquero and colleagues (2001) suggested the importance of studying incapacitation time in longitudinal studies of criminal offending. They argue that most longitudinal studies of behavi or have not accounted for the amount of time individuals are imprisoned, which is essential for accurately m easuring within-individual offending differences. Not accounting for exposure time can lead to po ssible underestimates of individual offending rates since the highest rate offe nders are likely to have been incapacitated. Using the CYA data, they analyzed a sample of male parol ees from the California Youth Authority. The researchers first examined arrest rates fo r the entire sample as a whole. They then investigated trends in arrest ra tes with and without adjustments for exposure time. Lastly, they analyzed arrests by utilizing late nt class models (see also Nagin and Land, 1993) to account for potential arrest rate heterogeneity across the sample (Piquero et al., 2001). Their analysis 69


identified a number of key findi ngs. First, they found that ther e is evidence of a relationship between exposure time and criminal trajectories with exposure time having a greater impact on younger parolees than older ones. This suggests th at arrest rates are higher at younger than older ages. Second, their latent class analysis reveal ed that 92% of the population reached the highest arrest activity in the la te teens through early twenties. Howe ver, after controlling for exposure time, the percentage of those ha ving higher activity up to age 33 d eclined to 72%. They note that this finding is important since the remaining 28% were still pers istent in offending. Lastly, the researchers also argued that controlling for expos ure time should yield the greatest proportion of persistent offenders however, when controls were eliminated from the model the percentage of persistent offenders decreased to 7%; a fi nding that suggests the importance of exposure adjusting for exposure time (Piquero et al., 2001: 69). Interestingly, very few studies, to date, have examined race differences in persistent offending. Only three studies were found that have done so and none employed trajectory models to study within-individual differences. The first, by Elliott (1994) examines data from the National Youth Study. The NYS is a longitudinal study of a probability sample of 1,725 youths between the ages of 11-17. He reported that the la st interviews of the sa mple were conducted in 1993 when the individuals were between 27 and 33 years old and had a total of nine waves. Background information on the youths were obtained fr om both self-reports and official records and official data were obtained on parents or primary caretakers. In studying the self-reports of whites versus non-whites, he found that ever-prevalence to age 24 for whites and non-whites were similar, indicating very little, if any, substantive race differences in propensity for violence. He also found that any differences between races were not enough to explain the -1 differe nces in arrest rates for viol ent offense over the adolescent 70


years. Although, the disjuncture be tween self-reports and official data diminished in the late twenties (Elliott, 1994: 18). However, with rega rd to persistence in violent offending, blacks were more likely to persist in offending into adulthood than were whites. He concluded that variations in career le ngth and “spacing of the career ove r the lifespan” accounted for the significant change in prevalence race differences in violence in adult hood (Elliott, 1994: 19). The second study examining race differen ces was by Ge, Donnellan, and Wenk (2001). The researchers investigated persistent offending among young males using California Youth Authority data. Employing a 20-year longitudinal study of 4,146 wards committed to the Deuel Vocational Institution (DVI) of the CYA between 1964 and 1965, they analyzed the effects of family environment, cognitive ability, and early anti-social behavior factors in predicting later persistent offending. Specifically, the researchers analyzed 2,363 of the original 4,146 wards for which they had complete records. Detailed da ta was had been obtaine d from a combination of self-report data, case worker inte rviews, and official records. Th ere were a number of key factors employed in the study such as measures of drug/alcohol abuse, family environment, cognitive ability, antisocial tendencies, onset age, age upon leaving school, and offending rates (Ge et al., 2001). Their analysis identified that family environment was significantly related to age at first arrest and the frequency of arre st prior to age 17. In addition, while alcohol or drug abuse was not related to onset age, drug a buse was significant and positive in predicting frequency of arrest before age 17. In other words, as the authors note, alcohol and drug abuse may not be as important in predicting when someone begins to offend, but it is related to late onset offending frequency. Antisocial behavior was significant in predicting bot h age of onset and offending frequency before age 17. 71


In their second model, the re searchers employed a longitudina l design in examining factors related to persistent offending. Using arrest fr equency amongst different age periods, they found that cognitive ability was signi ficant and negative in predicti ng persistence amongst all age groups except age 33 and over. In other words, they found that higher cognitive ability was related to lower rates of persistence. Antisocia l behavior was positive and significant for only the youngest age period (18-20) and none of the family environment scales were significant across any of the age periods. They propose that the lack of significance for the family environmental scales may suggest that an adverse family envi ronment as a juvenile do es not influence adult criminality but may be responsible for “launching” juveniles in offending at earlier ages and that leaving school earlier may have a more pronounced effect on later criminality. Importantly, and relative to this current unde rtaking, Ge and colleagues also examined racial differences in offending over time and unc overed some interesting findings. Up to age 21, there were no statistically significant differences in arrest frequencies between whites and blacks, but past 21 years old, blacks were arrested more frequently. Similarly, Hispanics were also arrested more frequently than whites but only after age 25. Asians did not differ in arrest frequency from whites. Up to this point, the above studies examined persistence using several different analytical schemes. However, none employed semi-param etric group modeling (SGM) developed by Nagin and Land (1993), (see also Nagin 1999, Nagin and Tremblay, 1999, and Jones, Nagin, and Roeder, 2001; Nagin, 2005). Chung, Hill, Hawkins, Gilchrist, and Nagin (2002) did utilize SGM in their investigation of childhood predictors of offense trajecto ries on a sample of 808 youths drawn from the Seattle Soci al Development Project. 72


In their study, they identified five offens e trajectories: non-offenders, late onsetters, desisters, escalators, and chronic offenders. Th ey also employed multinomial logistic regression models to estimate childhood predic tors (ages 10-12) to delineate the groups. Their analyses indicated support for developmental theories. For example, they f ound that the trajectory for the most chronic group of offenders comported with life-course persisting be havior and early onset of criminality while the late onset group shared characteristics of adolescent limited behaviors. However, the late onset group did not desist by age 21 and they showed substantive signs of heterogeneity between them. Moreover, many of them continued offending, drinking, fighting, and committing other such minor offenses after adolescence. With respect to desistance, the group identi fied as a desisting group shared offending seriousness characteristics at younger ages but by age 21, they had desisted from offending altogether. Interestingly, further examination of these groups showed that the desisters, although also being early onsetters, were influenced by having fewer anti-social peers, healthier school attachments, and having less availability to dr ugs in the neighborhood. Lastly, their multinomial logistic regression revealed significant differences in initial cr ime seriousness between all groups and consistency amongst independent measures in predicting crime escalation (Chung et al., 2002). In their study of California Youth Authority (CYA) wards data, Lattimore and colleagues (2004) examined arrest frequencies amongst 3,586 young, paroled offenders in California. Using random samples of parolees releas ed in the 1980s, the researchers investigated specific factors associated with past criminal history including individual and familial factors in predicting individual offending trajectories. For example, because the CYA data had collected data on 73


offenders from a number of different sources , the researchers had access to data linked to different levels of analysis such as indi vidual, family, and demographic variables. Their analysis elicited a nu mber of significant findings. They found that risk factors associated with prior antisocial behavior in predicting post-release arrest and its variance were evidence of prior violent behavi or and alcohol use (Lattimore, MacDonald, Piquero, Linster, and Visher, 2004:47). Particularly noteworthy was the negative sign of the coefficients, which indicated that those who had evid ence of prior violence and alcohol abuse were less likely to be rearrested than those with no prior evidence of either. More specifically, offenders who had prior alcohol problems had an 11% lower expected pos t-release arrest rate and a concomitant 13% lower variance to the expected value ratio, all else equal. Applicable to persistence and desistance, they found no significant effect for earlier sibling criminality or for most other familial or person al factors (e.g. history of physical or sexual abuse). They attributed the null effects to a ma sking effect of earlier extensive] controls on criminal experience in the models they employed. Findings were significant for school dropout, race/ethnicity, and geographic location. For inst ance, the school dropout effect was positive, suggesting that higher dropout rates increased th e probability of post-release frequency and its variance by 5 percent. They reported that the race effect show ed Whites having a 28 percent and Hispanics having a 17 percent lower expected fre quency while Blacks had higher expected arrest frequencies. Finally, their analys is showed a small but significant effect fo r age at release in predicting arrest frequencies. They found that a one-year increase in the average age at the time of release from the CYA increased the three-year post-release arrest rate by 2 percent, which they argue that “the sample appears to be so rting itself into low and high frequency offenders” (Lattimore et al., 2004). 74


In sum, while studies of pers istence continue to show up in the literature, there are a number of important points that should be discussed. First, there has been a lack of attention in race differences in developmental criminal behavior. There are only a few in the current literature. Both studies found that blacks pers isted longer than whites. Specifically, Ge and colleagues found that Hispanics pe rsisted longer than whites. Ho wever, neither study examined persistence/desistance utilizing race-specific tr ajectory models. Equally important, most studies on persistence and desistance have failed to control for exposure time or mortality. Accounting for exposure time and mortality are important consider ations in longitudinal studies of this nature since they potentially underestim ate individual offending rates (P iquero et al., 2001). This can lead to premature conclusions. La stly, more studies need to be done on persistence in general. Empirical Studies of Desistance Desistance from offending is one of the le ast understood processes in criminology (Laub and Sampson, 2001). Questions abou t why offenders start offending or continue offending far surpass questions abou t why they stop (Laub and Samps on, 2001). One might think that the dearth of empirical studies re flects a lack of interest; how ever, Sampson and Laub (2001: 2) point out, criminological theories are anything but “silent.” Fr om Matza’s (1964) maturational reform to Akers’ (1977) differe ntial association with law-abiding peers and now, Sampson and Laub’s (1993) own turning points in the life-course; desistance has been given ample theoretical attention. The question becomes, why such little empirical at tention to comport with the theoretical side of the equation? The most likel y reason appears to be th at measuring desistance has been hindered by unclear conceptualizat ion and methodological inadequacies (Laub and Sampson, 2001). Bushway and colleagues (2001: 129) wrote “Despite the theoretical and policy importance of understanding why people stop o ffending, we do not have robust conceptual models or rich empirical inve stigations of desistance.” 75


Desistance also has important implications to the age-crime phenomenon. In the aggregate, crime declines with age (Sampson and Laub, 2001, Gottfredson and Hirschi, 1990), however, the developmental approach suggests that is why we must study within-indi vidual differences over time, since it might not be that crime declines with age for everyone. Instead, it could remain stable or increase for a small percentage of the population (Moffitt, 1993). From this perspective, understanding the age-crime relationship mean s not only understanding risk factors for persistence, but those risk fact ors that would predict who desist s as well. Some have suggested that the risk factors predicti ng desistance are the converse of those predicting persistence (Sampson and Laub, 2001; Le Blanc and Loeber, 1993). However, that remains an empirical question. In their review of the desi stance literature, Laub and Samp son (2001) identified several areas within criminology that have empirically investigated desistance. They also identify researchers who have attempted to wrestle with conceptualizing it. Im mediately, however, the problem arises in the broadly varying definitions of desistance. Several researchers including Clarke and Cornish (1985), Loeber and Le Bl anc (1990), Farrington and Hawkins (1991), Shover (1996), Warr (1998), Uggen and Piliavin (1998), Bushway and colleagues (2001) and Maruna (2001) have all defined desistance in different, if similar ways. In line with the developmental approach and in agreement with Laub and Sampson, a fitting conceptualization of de sistance posited by Bushway a nd colleagues (2001) is to study desistance as a process. That is, contrary to ear lier studies of study desistance, which treat it as a discrete event that just happens, researchers shou ld investigate it as a process that whereby the frequency of offending slows down over time (Bushway et al., 2001). Alth ough this is certainly a 76


compelling perspective, because of its novelty it has yet to elicit many empirical studies analyzing desistance in this way. One of the earliest studies that addressed de sistance from a developmental perspective was the Glueck’s Unraveling Juvenile Delinquenc y Study (1950). The Gluecks’ followed 510 male reformatory juveniles over 15 years and found that the percentage of thos e desisting increased by twenty percent. Likewise, Wolfgang, Thornberry and Figlio (1987) studied a cohort sample of Philadelphia up to age 30. Their analysis revealed stability in offending during the juvenile years but decreased after age 16, however, they also found that the mean of offenses committed remained relatively stable into adulthood. In add ition, they looked at onset age relative to the probabilities of becoming delinquent and found that blacks had higher probabilities of initiating into offending earlier and persis ting longer, while whites initiate d later and desisted earlier (Wolfgang, Thornberry and Figlio, 1987). Paternoster (1989) studied decisions to partic ipate or desist from four types of common delinquency. As covered ear lier in the persistence literature review, he ex amined data from nine public high schools in the area of a mid-sized southern city. His analysis found that concerning relatively minor offenses (marijuana use, alcohol, petty theft, and vandalism), decisions to desist from offending were offense specific and not re lated to any sanction th reats. Moreover, in examining the effects of changes in moral tolera nce and moral beliefs, he found that changes in moral tolerance were significantly associated with decisions to desist. That is, those who developed lower tolerances for offe nding over the four measures were more likely to desist from offending than those who did not (Paternoster, 1989). Loeber and colleagues (1991) examined a variet y of different criminal career dimensions including initiation, escalation, and desistance, using data from the Pittsburgh Youth Study. They 77


reported that low disruptive behaviors such as scores on a low physical aggression scale, low oppositional defiant symptoms, and low attention deficit/hyperactivity scale were associated with desistance in offending. Moreover, they were related to higher scores on attitudes toward schooling. When they broke the sample down by br oad age groups (younger, middle, older), they found that amongst younger juveniles, they found that shyness was al so correlated with desistance from offending while for the middle and older groups trustworthiness, low truancy, good school motivation, caretaker enjoyment of child, good relationships with caretaker, and caretaker discipline correlated with desistance. The researchers also employed models that pa rtitioned desisters (“st able non-delinquents”) from de-escalators (Loeber et al., 1991: 73). Their findings re vealed that 51% of the deescalators at younger ages showed signs of aggressive behavior whereas only 10% of the desisters were categorized as aggr essive. In short, their findings suggest that across the three age groupings, low disruptive behavior , good educational achievement, negative attitudes toward problem behaviors, association with conforming p eers, and positive interactions with caretakers were important predictors in juveniles desisting from deli nquency (Loeber et al., 1991). Uggen and Staff (2000) examined work as a turning point toward desistance from crime using data from the National Supported Work Demonstration Project between 1975 and 1977. They analyzed the data on over 3,000 particip ants from nine U.S. cities. Employing random assignment, they assigned participants to either an experimental or cont rol group to investigate whether an experiment that gave jobs to adult offenders serv ed as a turning point toward desistance. The researchers found that work for t hose over age 26 did in fact serve as a turning point for life course persistent offenders. This held even for those gain ing marginal employment. 78


The findings support Sampson and Laub’s (1993, 2003) contention that informal social controls and human agency are important influences in redirecting offenders away from crime. 79


CHAPTER 6 DATA AND METHODS Data The data for the study comes from the National Institute of Justice Data Resource Program study entitled “Continuity and Change in Cr iminal Offending by California Youth Authority Parolees Released 1965” (CCCO) (Piquer o, Brame, Mazerolle, and Haapanen, 2001). 1 CCCO is part of a larger project on the o ffending careers of serious juvenile offenders who were incarcerated in the California Youth Aut hority juvenile justice sy stem (Piquero et al., 2002; Haapanen, 1990). The data comprise a partic ularly criminal juvenile offending population that was housed at the CYA Preston facility. Information was collected on 524 individual juvenile offenders paroled by 1984. The parolees ra nged in age from their late teens to early twenties at the time of their rele ase—a period in the life-course that researchers have termed “emerging adulthood”—and were tracked for seven years thereafter (Pique ro et al., 2001). Upon incarceration, the CYA compiles comp rehensive background and incarceration records on all wards. Information on each individu al comes from a variety of sources including youth authority electronic data, ward hard-copy mast er files, California Department of Justice (CDOJ) criminal history files, and the Californi a Department of Vital Statistics. Relevant background information was collected such as childhood living arrangement; family structure; whether the family received welfare; and parent and sibling criminality. Records on each ward’s childhood prior to incarceration include age at first arrest and at first imprisonment, 1 The data used in this study are twenty years old. “Period-specific factors” that may no longer have influence on contemporary patterns need to be acknowledged in interpreting results (see Piquero et al., 2002). Nevertheless, the author agrees with others who have used ‘aged data’ that the findings from studies such studies can be squared with the results from other empirical studies that utilize more contemporary data (see Piquero et al., 2002: 146 for a more detailed reasoning; also see Laub et al., 1998) . 80


psychological, educational, and intelligence evaluations, prior drug and alc ohol use, and juvenile offending history. Comprehensive incarceration data were also provided on each ward including their length of confinement to the CYA, number of escape at tempts, gang affiliation, whether or not they had known enemies, and whether or not they re ceived vocational or educational training. 2 CYA researchers also compiled post-release data th at included whether wards were assigned to probation or parole, had any psyc hological or medical afflictions and if so, what type of treatment was required, whether they were depe ndent upon drugs or alcohol , and whether or not they required treatment for substance addictions. Detailed time-dependent data were also coll ected each year for seven years after release from the CYA. The data contain information on their offending behaviors including violent and non-violent crime arrests. Arrest s for violent crime include murder, rape, aggravated assault, robbery, kidnapping, and extortion. Arrests for n on-violent offenses include grand theft and grand theft of an automobile, bu rglary, receiving stolen propert y, and forgery. Data on drug and alcohol use/abuse were collected at each year and were categorized by the type of drugs used, if any, such as heroin, mind altering substances, or upper/downers. Information on post-release drug use was also collected at each of the se ven time periods. Post-release data includes measures thought to affect the ward’s reentry in to the community such as whether they were gainfully employed, whether they were marrie d, cohabitating, or singl e, and their living arrangements such as with family, relatives, or in other institutions. Finally, data from the CDOJ rap sheets are provided on the amount of “exposure time” or time spent on the street (i.e. not incarcerated) du ring each of the seven waves. Exposure time is 2 Not all the data noted above are included in the analytic models. 81


coded within each period as free for the numbe r of months not spent in jail/prison/CYA detainment or they are coded as under correc tional supervision. Piquero and colleagues (2002: 147) explain exposure time coding in this way “an individual who was in prison for eight months during a particular year would be coded as having exposure time equal to four months.” Controlling for exposure or “street time” is n ecessary to reduce bias in modeling the semiparametric group trajectories since SGM would treat all parolee offending the same regardless of whether or not they were eligible to commit further crimes. Dependent Variables For each parolee, there are count data on crimin al arrests for violent, non-violent, and total crimes. The violent arrest measure is the sum of arrests for person crimes, which includes the following offenses: attempted murder, manslaughter (including vehicular) , robbery, aggravated assault, rape, and extortion/kidnapping. Non-viol ent arrest measures are the sum of property offenses, which includes the following offenses : attempted burglary/burglary, grand theft auto, receiving stolen property, and forgery. Total cr ime is simply the sum of both violent and nonviolent offenses. The operative de finition of crime in this stud y derives from official legal statutes that the parolee was determined to have violated. In this case, the crimes for which the parolee was arrested. Arrest data used for analysis do not include arrests for parole violations nor for any arrests for attempting an escape while confined to the CYA. Arrest measures were compiled from new arrests for the violent and non-violent crimes detailed above. Operationalization of the dependent measures occurred in a multi-step procedure. First, semi-parametric group (SGM) modeling was employe d to obtain offending trajectories for each of the twelve models in this study. SGM was then repeated for the entire sample for violent, nonviolent, and total crimes, as well as within race /ethnicity-specific samples for blacks, whites, and 82


Hispanics by each of th e three crime categories. 3 In addition to providi ng actual group-based trajectories, SGM also provides ot her critical information for th e second stage in operationalizing the dependent variable. It provi des separate group-based posteri or membership probabilities for every parolee belonging to each of the individual trajectories in each model. For instance, if there are four trajectories in a model, then SGM produces four pos terior membership probabilities. SGM also determines group placement based upon these probabilities. Typically, a probability of .7 or higher is considered suffici ent evidence of belonging to that group (Nagin, 2005). As such, based on this placement, group membership becomes the dependent variable in the second stage. Depending on the number of gr oup-based trajectories, th e dependent variable for each model is either a binary or a polytom ous categorical group-membership measure. Thus, the dependent variables for analyses are traject ory group memberships determined by individual posterior membership probabilities for each of the following crime categories: Count of violent crimes committe d between Time 1 and Time 7 Count of non-violent crimes committed between Time 1 and Time 7 Count of violent and non-violent crimes committed between Time 1 and Time 7 Independent Variables Crime theories addressed in this study propose certain factors that co rrelate to offending at various stages in the life-course. Childhood and adolescent factors such as age of onset, antisocial personality disorders, neuropsychologi cal impairments, aggression, low IQ, and drug or alcohol abuse. Many are though t or have been found to correla te to life-course persistent offending (Moffitt, 1993, Loeber, 1993, Patterson and Yoerger, 1993; Sampson and Laub, 1993). For example, Moffitt (1993) posits that the inabilit y to socialize properly an antisocial child as a 3 Semi-parametric group modeling is discussed in detail later in the chapter. 83


result of all or some of these risk factor s further exacerbates the child’s problems whereby antisocial behavior becomes entrenched or pathol ogical. Such pathologies link to initiation into offending at younger ages, typically higher rate and more seri ous offending throughout the lifecourse, even after the majority of other youthful offenders ha ve matured out of crime. On the other hand, Sampson and Laub (1993) also focus on criminality in adulthood, though they differ on how this occurs. They posit th at the quality of informal social controls exerted by community sanctions, marital bonds, and employment can function as a factor correlating to criminal behavior or as a turning point away from crime (or toward conformity/conventional behavior) at each age-graded step of the life-course. They also suggest that social control mechanisms differ at each age grade. For example, whereas young children are typically influenced more by parents and aut hority figures, during adol escence peers increase in importance. They, as well as Hirschi (1969), have argued that for adults, having bonds to others, especially functional marital bonds and social capital, are nece ssary mechanisms for conventional behavior. In other words, informal so cial control mechanisms link to critical stakes in conformity. The more meaningf ul bonds adults have the more they have a stake in conforming to social norms and societal laws. There are numerous studies that utili ze a wide range of both individual and familial/structural risk factors for studying criminal careers. For example, age of onset is thought to be a critical determinant in numerous studi es of the criminal career (Piquero and Chung, 2001). Earlier manifestations of antisocial behavi or have been linked to a higher likelihood of later antisocial/criminal behaviors (Blumste in et al., 1986; Piquero and Chung, 2001; Chung et al., 2002; Moffitt, 1994). The assumption is that those who are arrested earlier have higher 84


probabilities of even earlier antisocial or crimin al behavior and therefor e, are more likely to come to the attention of the criminal justice system at an earlier age. Other measures employed in this study are al so drawn from the criminology literature (see Loeber, 1993; Moffitt, 1993; Patterson and Yoerger, 1993; Sampson and Laub, 1993; and Blumstein et al., 1986). Key demographic and fam ily measures include ra ce/ethnicity, welfare reception, and family structure. Childhood risk f actors include juvenile alcohol and drug use, intelligence, and academic grade. Family criminal history includes mother, father, and sibling criminality. Ward’s incarceration history factors include the length of confinement in the CYA, number of escape attempts, vocationa l training, and educational training. There are also measures on the wa rd after release from the CYA. Data from this portion of their history includes whether they suffered from mental illness or some form of personality defect, and whether they rece ived drug/alcohol treatment, an d post-release drug and alcohol, which are measured at each time period. Finally, st ructural level stake in conformity measures that are employed include whether the ward was married or employed full time. The following is a description of each independent measure in the models: Race/Ethnicity was coded as (1) White, (2) Black, and (3) Hispanic. Since race was discussed at length above, it is only necessary to note that prio r research by Ge and colleagues (2002) and Elliott (1994) found that blacks and Hi spanics persisted longer than whites did. However, neither used the trajectory models pr oposed in this study. Furt hermore, Elliott’s study only examined race differences using broad ca tegories of white/non-white thereby excluding Hispanics (or subsuming them within the black/white dichotomy); the largest minority group in the United States at 38.8 million people (U.S. Cens us Bureau, 2006). As a predictor in the full models and as a complete sample in the race/eth nicity-specific models, race/ethnicity serves as a 85


key component in understanding persistence and de sistance in this study. Incidentally, race and ethnicity could not be further broken down to investigate intra-ethnic or racial differences. Therefore, whether any one parolee was biracial or of mixed race or ethnic heritage cannot be determined from this data. Age of Onset is a reliable continuous variable deno ting the age in which individuals first initiate delinquency or first come to the formal attention of criminal justice system. In this study, it is measured as the age at first arrest , although other studies employ a variety of constructs such as self-reported initiation or age at firs t contact with the crim inal justice system. Age of onset is perhaps the most widely reported correlate in the career criminal/developmental criminology literature. Most life-course and developmental theories are premised on the belief that fact ors inducing early chronic offending manifest earlier in life than factors leading to late onset in frequent offending and thus maki ng some individuals more prone to offending earlier in life (Chung et al., 2002; Sampson, Laub, and Nagin, 1998; Moffitt, 1993). Other theoretical perspectives su ch as Hirschi’s (1969) bonding theo ry and Akers’ (2000) social learning-social structure theory are premised on early onset of delinquency. For Hirschi, the lack of effective parenting wi ll lead to behavioral problems in unr estrained juveniles while for Akers, it leads to an imbalance of defin itions favorable toward delinquency rather than conformity (see also Simons, Chyi-In, Conger, and Lorenz, 1994). Empirical studies reflect somewhat consistent results for the early onset of offending for many of the most high-rate pers istent offenders (Tremblay et al., 2004; Sampson et al., 1998; Wolfgang et al., 1972) and the most serious offende rs (Chung et al., 2002; Loeber and LeBlanc, 1990). However, in studies that measured age of onset in general terms it becomes less clear as to whether it predicts general involvement in crime (Farrington, 2003). In the criminology 86


literature, age 14 seems to be the generally referenced cut point; however, such a determination is not unyielding (McGloin and Pratt, 2003; Tibbets and Piquero, 1999; Moffitt et al., 1994). Relative to other age of onset measures, ther e are both advantages and disadvantages in using age at first arrest as an indicator of age of onset. Eggles ton and Laub (2002) point out that using official reports poses a disadvantage in that there are many people that offend (even at very high rates) but do not come to the attention of the criminal justice system. Conversely, Weis (1986) and Eggleston and Laub (2002) suggest that the advantages of using official records is that they are typically more reliable than other forms of information on offenders and that they provide other important data that might not be available otherwise Juvenile risk factors comprise the following three variables: Intelligence (IQ) , Claimed Grade, and Psychological Diagnosis . Intelligence is a reliable construct measured as a continuous variable ranging from 64-135. It is thought to be ne gatively correlated with crime such that individuals with lower IQs are more lik ely to be criminal than those with higher IQs. However, while IQ tests are well established, gaps in our understanding remain as to whether psychometric tests of intelligence actually measure innate intelligence or something else such as achievement, reading ability or te st ability (Vold et al., 2002). Moreover, there are different types of intelligence that people po ssess including performance and ve rbal intelligence (Wright, 2002). Notwithstanding these issues, the current literat ure reveals a link between IQ and a variety of crime types. In 1977, for example, Hirschi an d Hindelang concluded that the “weight of the evidence” suggest that differences in intelligence were more important than both race and class, noting that measurements of intelligence are not racially or culturally biased. They further reported that having a low IQ increased the li kelihood of arrest and imprisonment through its effects on school failure. Wilson a nd Hernnstein (1985) take a sim ilar position but argue that the 87


relationship is indirect; that lo w intelligence leads to poor sch ool performance, which in turn, increases the probability of criminal behavior. Issues of intelligence and race are controversia l. Much of the controversy stems from the manifest effects of racism in the U.S. However, this should not preclude discussions of IQ, race, and crime, if only since it could lead to a better understanding of how structural and other factors might lead to such differences. Yet, those w ho oppose this line of res earch, aside from the history of people using purporte d genetic differences as a means of power and control over others, and for justifying racially discrimina tory laws, oppose it because many studies fail to properly account for structural, cult ural and historical factors that have entrenched a social hierarchy in the U.S. In their oft discussed but much criticized book, The Bell Curve , Hernnstein and Murray (1994) claim that there are direct differences in offending that corr elate to racial differences in intelligence. They also give primacy to the belief that intelligence is mostly an inherited trait, which suggests that inherited in tellectual ability is a source of all differentiation in offending. Similar to previous studies in th e first half of the twentieth ce ntury, where researchers such as Sutherland (1931) criticized the methodology, vali dity, and racist under pinnings of studies on intelligence and other biopsychological based res earch, their findings provoked an especially sharp outcry for the perceived or real racial bias in their rese arch (Cullen, Gendreau, Jarjoura, and Wright, 1997; Gould, 1996). 4 4 Criticism of Hernnstein and Murray’s research is well documented. There are no less than a dozen academic and non-academic critiques of their wo rk that raise serious challenges to both th e validity and reliability of their findings but, perhaps, the single most prominent criticism is with the conclusions they draw from their research. Their conclusions have led to serious charges of racism. Criticisms notwithstanding, their findings to some degree comport with findings in other research on intelligence even if th e authors draw very different conclusions about what they mean. 88


Yet, there is other evidence of racial differen ces in IQ scores and crime. Blacks and other minorities consistently score on average 15 points less on IQ tests than whites while offenders scored eight points less than nonoffenders (Moffitt, 1994), not to mention that blacks and to a lesser degree, Hispanics, have higher self-repor ted rates of criminality. There are compelling links between low verbal intelligen ce and antisocial behavioral differences in childhood (Gibson et al., 2001; Moffitt, 1994; Patterson, 1990). Patt erson (1990) suggests that it manifests in noncompliance and other antisocial be haviors that prohib it the development of proper social and cultural skills (see also Aguilar et al., 2000). Ka ndel and colleagues assert that high IQ can as buffer children from antisocial behavior duri ng childhood (Kandel et al., 1988). Research on IQ and adult offending, on the other hand, has not f ound any consistent links (Lynam, 1993) while, some researchers suggest that IQ does not predic t who will desist from crime (Laub et al., 1998). For whatever IQ is tapping into, it seems th at perhaps those with higher IQs are simply equipped to avoid better official detection by the criminal ju stice system than are those with lower IQs. To what extent heredity or groupbased factors have on purported differences in intelligence or that intelligence itself is an im mutable condition that directly correlates to variations in criminal propensity, and to what extent it predicts participation in any particular crime type still remains an opaque but hot button issue. Grade is a reliable continuous scale of the juveni le’s academic achievement ranging from 1.9 to 13.6. Academic achievement negatively correlate s to crime, which suggests that those with lower scores are more likely to participate in cr iminal behavior than t hose with higher scores. The research on education and crime reveals strong evidence of educational differences in offending, particularly in adolescence and early adulthood (Robins, 2006; Hong and Ho, 2005; 89


Jajoura, 2005; Sampson and Laub, 1993; Hirsch i, 1969). One of the strongest educational predictors of crime is high school graduation (W.T. Grant Foundation, 1988). Across a number of different educational indicators, however, the research reveals that students that perform poorly on te sts, are marginalized by others, are less motivated to succeed academically, and do not graduate from high school, are more likely to be criminal (W.T. Grant Foundation, 1988). The literature al so shows correlations betw een education and employment (Cernkovich and Giordano, 2001), effective parenting (Hay, 2001; Simo ns et al., 1988; Hirschi, 1969), and racial differences in offending (W.T. Grant Foundation, 1988). Psychological Diagnosis is a dichotomous variable measured as (1) presence or (0) absence of a psychological or personality conditi on. It is a rough indicato r of neuropsychological health in that it measures the presence or absence of diagnosed psychological problems in adolescence. This measure was collapsed from a multiple item index. Unfortunately, it is not a direct measure of parolee’s neuropsychological co ndition as a juvenile be cause it does not give us any information about the presence of many clinically assessed neuropsychological conditions or any information about the physical health of the brain (Moffitt, 1993). It might even be more appropriate to consider this measure simply a psychological condition since the assessment is more in line with traditional psychology th an neuropsychology. Regardless, proper brain functioning is an important corr elate of offending and developmental neuropsychology has been given considerable attenti on over the past century. Neuropsychological and psychological m easures offer predictive power beyond intelligence by providing information on the physi cal condition of the br ain (Moffitt, 1994: 278). Verbal ability, hyperactivity, impulsive personality, visual-spatial analys is, language processing, poor memory, and an array of other neurops ychological impairments link to criminality, 90


especially across individuals with similar IQ scor es (Piquero, 200; Bartusch et al., 1997; Moffitt, 1994; Caspi, Elder, and Bem, 1987). In addition, neuropsychology focuses on particul ar areas of the brain, including the frontal and temporal lobe impairment, by using a multitude of different tests including the Wechsler Intelligence Scales for Children (WISC-R), Rey Aud itory Verbal Learning Test, and many others (Brooker, 2005; Moffitt, 1994). Importantl y, this study is not purporting measuring actual neuropsychological deficits but rather it poses a simple proposition that differences in measured psychological diagnoses, along with IQ , and academic ability, signify potential neuropsychological differences in offending. In short, it seems that many types of psychological disruptions or physical br ain abnormalities that link to a variety of different antisocial behaviors. And though the variables employed in this study do not provide direct measures neuropsychological deficits, together they are better than not em ploying any measure. Drug Use and Alcohol Use comprises a total of four variab les in this study. The first two are simple dichotomous variables that determine whether or not the juvenile used any drugs or alcohol prior to incarceration. The second ones are post-release indices that indicate postincarceration drug or alcohol. The post-incarceration alcoho l and drug use indices were calculated by summing alcohol and drug use fo r each of the seven years post-release, respectively. Although illicit drug use itself is a criminal act, the research on drug use linking to other crime, including repeat and long-term associatio n with offending, is abundant. However, there are still many elements of the relationship that are unclear (White et al., 2002; Huang et al., 2000). Past research shows links between drug us e and other problem be haviors in adolescence including increasing the likelihood of quitting school early, earlier sexual participation, and other 91


delinquent behaviors (Wallace and Bachman, 199 1). In addition, studies on recidivism consistently report upwards of 80% of repeat o ffenders had prior drug or alcohol issues with about 25% using a substance in the month prior to the criminal act and over 30% committing a crime while on drugs or alcohol (Travis and Vish er, 2005; Lattimore et al., 2004;). For those who have committed five or more crimes, the num ber increases to over 80% (Travis and Visher, 2005). Findings such as these are quite robus t in criminology, irre spective of methodology employed. The link between drug use and crime, however, is not necessarily one-way (LaFree et al., 2000). White and Gorman (2000) discuss three different relationships between drug use and crime: 1) drug use is a cause of crime, 2) crime is a cause of drug use, and 3) the relationship is spurious; both are linked through other factors (LaFree, 2000). The numerous relationships are indicative of multiple pathways to drug use and crime, and for that matter, different types of crime. For example, drug use is linked a variet y of violent and propert y crimes and many forms of delinquency (Kerner, 2005; Thornberry, 2005; Wright et al., 2004; Jessor, Donovan, and Costa, 1991) and it is also closely related to crime through other factors such as gang involvement (Thornberry, 2005). Alcohol use, conversely, is not necessarily synonymous with drug use, although the two are correlated. Nearly 20% of state and federal prisoners self-report committing violent crimes on alcohol alone (Travis and Vi sher, 2005). Some criminologists suggest drug and alcohol use, especially heavy use, contributes to the early onset of antisocial behavi or and delinquency in adolescence (Thornberry, 2005; Sampson and Laub, 1993). There are other negative outcomes for those who use both drugs and alcohol. Numer ous studies, particular ly those published in 92


recent decades, show how drug and alcohol use impai rs critical brain functioning, which leads to addiction and to further compli cations such as chronic health problems (Blumstein, 1998). In short, drug and alcohol abuse is thought to affect not only the immediate lives of individuals who abuse them but that it might have long-term consequences on future employment and success. Furthermore, research has shown that there are racial/ethnic differences in drug and alcohol us e among teens; that black youths are less likely to abuse drugs and alcohol than their white counterp arts are (Wallace and Bachman, 1991). Family measures comprise a relate d grouping of six variables: Family Structure, Number of Siblings, Economic Well-being, and three Family Criminality measures (there are separate variables for father, mother, and sibling criminality). Family measures are employed to get a sense of the structure and influences of the family on juvenile and adult offending. Family structure is a dichotomous variable th at indicates whether the juvenile’s family was intact. In other words, whether or not both parents were in the same househol d during the parolees’ childhood. Number of siblings is a continuous variable employe d to determine the relationship between family size and crime. Family Economic Well-being is a rough indicator of socioeconomic status of the family prior to the juve nile’s incarceration. In this case, whether the family received public welfare subsidies or not. Th e three family criminality indicators are all separate dichotomous variables m easuring whether the juvenile’s fa ther, mother, or siblings had committed crimes. Family factors are among the most important and consistently invest igated correlates of juvenile offending (Gibson et al., 2001; Wells a nd Rankin, 1991), and they strike to the core of the theories framing this study. Take, for exampl e, Gottfredson and Hirschi’s general theory (1990), and Hirschi’s (1969) bonding theory, whic h links offending to the strength of bonds 93


between juveniles and parents or authority figures. Sampson and Laub’s age-graded theory (1993) similarly includes the streng th of juvenile social bonds with parents, but also addresses changes in adult pro-social bonds family bonds, most prominently, marital bonds. Moffitt’s (1993) typology suggests that the development of life-course pers istent offending is chiefly the result of a “transactional” pro cess of ineffectual parenting of a child with neuropsychological deficits, which then leads to path ological antisocial behavior. There are numerous studies that illustrate the importance of family in adolescent development and aggressive behavior (T remblay, 2005; McCord, 2001). Parent-child relationships are thought to be one of the most robust predicto rs of adolescent delinquency. Juvenile delinquency, for example, has been associated with large family size (Blumstein, 1986), economic disadvantage (Sampson and Wilson, 1995; Massey and Denton, 1993), and parental and sibling criminality (Blumstein, 1986). Fam ily factors operate in numerous ways during a child’s development. The conflu ence of many negative factors such as single-parent headed households, parental criminality, family disrupti on, and economic marginalization, correlate with a broad range of negative outcomes fo r juveniles (Eggleston and Laub, 2002). Single-parent families and those from broken homes have been associated with increased risk of offending (Gibson et al., 2001) and with economic disadvantage, wh ich also leads to an increased risk of offending (O’Brien, 2003; Slovanien, 2000; Loeber and Farrington, 1993). Family disruption is particularly stressing on lower-class minority families, where the trappings of poverty and residential segregation increase s the marginalization of single mothers (and fathers) and inhibits the proper monitoring and socializa tion of young children (Thornberry, 2005; Sampson and Morenoff, 2004). Individuals from larg e families are more at risk to offend, even 94


during adulthood (Eggleston and Laub; 2002, Farrington and Loeber, 1993) and those whose parents or siblings offend are also more likely to offend. Prison Impact variables measure the effects of impr isonment on the parolee’s post-release offending. Four variables are employed to measure prisonization: Length of Stay, Escape Attempts, Educational Training, and Vocational Training . Length of Stay is a continuous variable measured as the number of mont hs the parolee was incarcerated. Escape Attempt is a dichotomous variable indicating whether or not the parolee had attempted to escape from the CYA. Educational and Vocational Training are two separate dichot omous variables indicating whether the parolee received educat ional training, vocational traini ng or both while incarcerated. There is a belief among many criminologists that imprisonment negatively affects the lives of the prisoner upon their reentry into society. Walters (2003: 400) define s prisonization as the adoption of the “folkways, mores, customs, and general culture of the pe nitentiary” (see also Clemmer, 1970: 270). There is litt le doubt that imprisonment leads individuals to develop certain attributes and manners in order to get by and that they can also learn new criminal techniques (i.e. prisons as “schools of crime) (Maruna, 2005). On the other hand, imprisonment might make individuals realize the er ror of their ways. However, individuals come to experience prison, one thing is clear: prisonization is decidedly more co mplicated and individualis tic than it is often treated in the litera ture (Maruna, 2005). Similar to Sampson and Laub’s (1993) view of life-course transitions, individuals experience incarceration differently, even if th ere are certain shared experiences. Individual experiences not only include differences based up on age, personality, and external factors that shaped the individuals behavior patterns prio r to incarceration but th ey also vary across institutions, space and time (Maruna, 2005; Walte rs, 2003). There are two models that have 95


guided research on prisonization over the past few decades (Walters, 2003). The first is the importation model developed by Sykes (1971), which holds that prisoner’s develop a subculture in order to attenuate th e “pain and mortification of prison” (Walters, 2003: 400). The other, by Irwin and Cressey (1962) argues that prison subcultures develop as a hybrid of other subcultures, each reflecting the communities in which they came from (Maruna, 2005; Walters, 2003: 401). Yet, according to both Maruna (2005) and Walte rs (2003), the empirical findings only show moderate support for either of these theories. The imprisonment variables in this study were employed to provide at least some indication of how imprisonment correlates to adult persistence and desistance in adulthood. For example, it seems theoretically pl ausible that the longer one is imprisoned, the more likely they will suffer deleterious prisonization effects and thus, more they would be more likely to return to offending after incarceration. Lars and Nelson (1 984) offer support for such a hypothesis. They found that the length of time served positively corr elates to later offending and is more powerful than prior criminal behavior in predicting late r offending. Likewise, escape attempts might also indicate a negative impact of imprisonment. It mi ght also be indicative of low self-control such that that those who attempt escapes are more likely to be risk take rs or thrill seekers than those who do not (all else equal). Conversely, the educational a nd vocational training variable s could signif y the positive effects of imprisonment. In this respect, those who have received training would be more likely to desist from crime after release than those who did not. And, while these variables say nothing about how that training was conduc ted or the quality of the training—for instance—if the ward was trained to cut lettuce rather than being compre hensively trained in a marketable job skill, nor does it tell us how effective the training program was. The simple effect of receiving one or both 96


of the types of training could signify differences in post-release ability to acquire desist from crime. Stakes in conformity comprise two measures: marriag e and employment. They were calculated by summing whether the parolee was ma rried or employed at each year for the seven years after release. These two variables comp rise a cornerstone of Sampson and Laub’s lifecourse theory of offending. Explicit in their theory is that all individuals can change and that the adoption of conventional behavior s occurs because of the devel opment of different informal social controls such as marriag e or employment in adulthood. Employment is an essential predictor for ma ny empirical studies of crime. However, the effect of employment on crime is not fully clea r and the empirical evidence is mixed (Piehl, 1998). One explanation of the mixed results is that employment might affect adolescent delinquents differently than adults because adolescents cannot formally work prior to age 15. Even if they are, opportunities are often limited to part-time employment. On the other hand, adult employment, along with ot her adult responsibilities, ho lds much more significance, particularly when considering familial support responsibilities and because social pressure dictates legitimate employment as an expectation of full a dulthood (Ness and Dasori, 1998). In addition, the type of employ ment could hold importance as well. Jobs that barely pay living wages or do not provide enough employment (i.e. underemployment) might not have the desired effect on crime. Tradit ionally, employment has been t hought to affect crime through a rational choice mechanism by increasing the likelihood that individuals will tu rn toward criminal activity when fewer legitimate jobs for teen s are available, and it decreases crime when legitimate opportunities are av ailable Blumstein (1998). 97


From a developmental standpoint, Horney and colleagues (1995) suggest that in the short term, even temporary employment will reduce future criminal behavior. Some research has found support for the link between employment and crime. For ex ample, Farrington, Gallagher, Morley, St. Ledger, and West (1986) found that unemployment increas ed the probability of conviction for adults (Farri ngton, Loeber, Yin, and Anders on, 2002). Likewise, Blumstein (1998) found that increased employment opportunities played a role in re ducing crime during the crack epidemic in the 1980’s. In a review of the economic and crime literature, Piehl (1998) found that even when there is a significant relationship; decreases in unemployment only reduced crime by less than two percent. Marriage, on the other hand, is thought to br eak the cycle of crime and other behaviors such as binge drinking and drug use, which occurs much more amongst youths while also making former offenders more “risk averse” (Sampson and Laub, 2004). Similarly, meaningful employment can change one’s outlook and lead to other conventiona l opportunities and the development of social capital. In a recent symposium on life-c ourse and developmental theory, Sampson and Laub suggested that marriage could have an effect on one’s propensity to offend and on their opportunities (Sampson and Laub, 2005). Th ey also assert that marriage is more dynamic given that the “state” of marriage can be a causal mechanism that changes within individual’s propensity. However, the mechanism that activates such a change is still unknown. As such, stake in conformity measures can be viewed as turning points for prior offenders, which suggests that those who score higher on th e indices will more likely belong to groupedbased trajectories that indicate desistance or de-escal ation rather than persisting or escalating offending. 98


Age at release is a continuous variable that measures the age at which the parolee was released from the CYA and is conceptualized as being closely related to the stake in conformity measures. Its relevance for understanding post-rel ease offending patterns is based in part on the sample itself. All of the parolees were in their ve ry late teens to mid twen ties, or just past peak age of offending. This period of emerging ad ulthood is when many, if not, most young adults move past earlier youthful indiscretions and by taking on adult roles. Moffitt (1993) suggests this is due to the closing of the maturity gap while Sampson and Laub (2004) suggest it is because of the development of stakes in conformity. In fact, age or aging seems to have an effect beyond both of these claims. Sampson and Laub (2004) and Shover and Thompson (1992) both argue that this is because aging increases the change s in perceptions of risk of being caught and perhaps their experiences with being confined. It may also be that aging increases perceptions of risk of incarceration, lesseni ng of ability to commit some types of crime and, coupled with gainful employment, less rewards relativ e to the risks (Shover and Thompson, 1992). Research Questions The criminal career model emphasizes isolati ng the various dimensions of offending from onset to termination (Farrington, 2003; Chung et al., 2002; Blumstein et al., 1986). Similarly, the current crop of developmental and life-course th eories all impart the necessity of examining potentially differential offending patterns between active offenders. If th ese positions are valid, then risk factors predicting offending should not only differentiate between who participates in crime and who does not, but they should predict important differences between active offenders (Sampson and Laub, 2003; Chung et al., 2002; Mo ffitt, 1993; Patterson and Yoerger, 1993; Loeber and Le Blanc 1990). Sampson and Laub suggest that these clai ms require that the existence of unique offending groups as identified by similar group-based offending patterns and should therefore possess different causal mechanisms . For juvenile delinquents in California, the 99

PAGE 100

CYA is the “last ‘stop’ in the juvenile justice system,” and the Preston facility population from which the sample is drawn comes from a particul arly criminal/seriously delinquent element CYA wards (Lattimore et al., 2004). Accordingly, th e CYA adequate for investigating offending trajectories between previously active offenders and for isola ting distinctive etiologies for different types of offenders should such different etiologies exist. Distinguishing the offending trajec tories of more serious persistent offenders versus those who desist or fail to recidivate could have important theoretica l and policy implications. First, many developmental theories of crime like thos e noted above explicitly call for such an examination and second, if it turns out that there is a group of persistent offenders then policies can be formed that deal with offender types. Furt hermore, differences in types of crimes, not just patterns of offending, are important for policy and academic reasons. Thus, it is not only sufficient but also necessary to examine offendi ng trajectories across different crime types, as this study does. While it is possible to examine individual crimes such as robber or assault, because there are many zeroes Additionally, amongst the recent studies on persistence and desistance employing trajectory models in order to answer similar ques tions, few have attempted to isolate the effects within and between different races or ethnicities, while others have only investigated particular types of crime. Therefore, with this in mind, four separate, but related qu estions are addressed: 1. What do violent, non-violent, a nd total crime trajectories look like in the full sample, as well as across racial/ethnic-specific samples? 2. Do the models reveal patterns of persistence in or desistance from offending as predicted by the theories? 3. Do childhood, adolescent, and adult risk fact ors, including formal and informal social control mechanisms, which are regarded as important in predicting individual criminal offending predict trajectory differences am ongst a population of post-release juvenile offenders who were released as young adults? 100

PAGE 101

4. Do those same risk factors operate similarly or differently in predicting trajectories across and within racia l/ethnic groups? Method of Analysis The empirical foundation for this project rests on investigating the correlates of persistence in and desistance from offending with the underlyi ng assumption that there are discrete groups of offenders with heterogeneous tr ajectories. This raises an important question: do offending trajectories that reveal patterns of stable persis tent behavior or desisting behavior (or other potential types of offending such as escalating or intermittent) have different risk factors that predict group membership in these trajectories? With the advent of contemporary developm ental, life-course, and general theories, understanding the risk factors that cause some people to persist or desist and the dynamic relationships that these processes entail takes on added significance given that, in one respect, theoretical positions diverge on these very concep ts. Differences that also encompass what these theories say about other important factors such as race and gender and about the relationship between age and crime. The critical objective of this study, therefore, is to investigate persistence and desistance between blacks, whites, and Hispan ics during a period of the life-course that has come to be termed “emerging ad ulthood” (Piquero et al., 2001). Toward this end, there are three phases to this study. In the first phase, group-based trajectories were modeled using semi-parametri c group modeling design, which in turn provided the dependent variables for the analytic models. In the second phase, analysis of variance (ANOVA) was calculated to ensure variation of the demographic variables across the dependent variables. In the la st phase, using the group membership s identified through SGM as the dependent variables, binary and multinomial l ogistic regression models were employed to 101

PAGE 102

examine how the independent variables predict trajectory group membership. Given the unsettled nature of SGM, what follows below is a brief account of its methodological underpinnings. Semi-Parametric Group Modeling A fitting method to address theoretical questions posed above would be to employ a semiparametric group modeling design (SGM) that esti mates trajectories for discrete groups of offenders based upon similar offending behavior (Nagin, 2005, Jones et al., 2001; Nagin, 1999; Nagin and Land, 1993). SGM was developed by Nagin and colleagues (Nagin and Land, 1993; Nagin, 1999; Jones et al., 2001) and provides a flexible a nd simple means of obtaining trajectories for which similar offending clusters belong. SGM is therefor e a logical choice to answer the study’s questions for three reasons. There are other reasons why SGM is an appropriate analytic t ool for this study. First, SGM is quite versatile in handling diffe rent types of dependent variable s. For example, it can fit semiparametric mixtures of Poisson and Zero-inf lated Poisson, censored normal, and Bernoulli distributions to longitudinal data (Chung et al., 2002; Jones et al ., 2001). As with most crime, data, even amongst populations of known offenders, offending is a relatively rare event. Since the dependent variables are group memberships ba sed on offending counts, Zero-inflated Poisson (ZIP) distributions were employed. ZIP models are appropriate for modeling count dependent variables that contain many zeros and better fit this data since individual crime patterns that comprise each of the crime types contain an over-dispersion of zeroes. SGM also uses a mixture of defined probabilit ies to identify discrete trajectories based upon a priori conceptualizations in order to determ ine if such conceptualizations are present in the models (Chung et al., 2002; Jones et al., 2001 ; Nagin, 1999). Third, SGM also fully utilizes the data by modeling trajectories of all individuals with at least two data points and thus, is well suited for longitudinal data. Based on this pr ocedure, SGM produces posterior membership 102

PAGE 103

probabilities that determine group membership within a trajectory, which in turn, are used as the dependent variable in the analytic models. Posterior Membership Probabilities and Bayes Information Criteria SGM produces posterior membership probabilitie s, which indicate the probability of each individual being assigned to each of the groups identified from the SGM outcomes. Determining group membership using this procedure has been shown to be quite effective and a valid means of group selection because membership probabil ities are based upon objectiv e criteria that best fits the behaviors determini ng the trajectories (N agin, 2005; Jones et al., 2001; Nagin, 1999). However, SGM allows the researcher to choose, within software limitations, the number of trajectories for each model. Nagin and others have addressed optimal gr oup number selection and procedures have been developed for selecting the most favor able model (Chung et al., 2002; Nagin, 1999; D’Unger et al., 1998). In a detaile d account, Nagin reports that ther e are two possible choices for selecting groups. The first, a like lihood ratio test, is inappropriate for this study since it is only suitable for problems in model selection where the alternative models are nested. The second option utilizes Bayes Information Criterion for selecting the optimal model (Land, McCall, and Nagin, 1996). Kass and Rafferty (1995) and Rafferty (1995) suggest that BIC is suitable for comparing both nested and non-nested models. Bayes probability functions have a long history in scientific inquiry. Numerous studies in criminology demonstrate the utility of the BI C for optimizing trajectory selection in SGM models (Chung et al., 2002; Land et al., 1998). However, even though minimizing the absolute BIC is typically adhered to, it is not necessary that one select the lowest absolute BIC (Nagin, 2005; Chung et al, 2002). Parsimony and theoretical considerations factor into determining optimal number of trajectories. But, as with any innovative methodological development, there 103

PAGE 104

are going to be questions about its validity or utility. In such cases, caution must be taken interpreting and pr esenting results. The current literature has identified four logical groups based upon developmental theory. Two groups have been given the lion’s share of a ttention, early onsetters-l ate desisters and late onsetters-early desisters. Chung a nd colleagues, and Nagin and colle agues also suggest that there are two other logically distinct groups that can be assumed; t hose who initiate and terminate earlier and those who initiate and terminate la ter (Chung et al., 2002; D’U nger et al., 1998; Tolan and Gorman-Smith, 1998). These other groups have b een given scant attenti on in the literature (Chung et al., 2002). This study departs from th ese distinctions insofar as determining trajectories based upon offending initi ation and termination. Rather, the trajectories are modeled using crime patterns and street time. Offending initiation (age of onset) employed as an independent variable. There are twelve analytic models in this study: one model for each racial/ethnic group by each crime type and a full model by each crime t ype. Moreover, exposure time in each model is accounted for in SGM modeling by including the vari ables within the SAS procedures that identify the trajectories. Analyz ing individual trajectories enable d determining what theoretical constructs best predicted traj ectory group membership (i.e. pers istence or desistance or some other pattern). If there is heterogeneity between trajectories for different groups and the risk factors culled from the developm ental theories that were em ployed in this study differ in predicting different group membership (e.g. onset age predicting persistent offending for high rate offenders), than the findings should have important theore tical and methodological implications since they would lend support to the developmental perspective. However, if the risk factors do not hold in pr edicting group membership, but rather, operate similarly across 104

PAGE 105

groups even though there might be significant heterogeneity betw een offending trajectories, the findings could better support other theories. Current Focus There are two objectives for this study. First, it is to investigate how variables employed in this study predict offending group membership of a sample of serious juvenile offenders that were on paroled from the CYA. Second, it is to determine the optimal number of trajectory groups. As discussed above, SGM model selection occurred using Bayes Information Criteria, however, what this means for persistence and desistance within a lifecourse/developmental/career criminal framework is still an open question. Nevertheless, in this scheme, membership in such trajectories is the result of similar offending patterns after release from the CYA. One would expect that risk factors drawn from Moffitt’s developmental taxonomy to significantly predict membership in hi gh-rate persistent grou ps, net of all other factors. On the other hand, if there were trajectories th at show desisting or deescalating behavior, than one would expect factors indicative of Sampson and Laub’s theory to emerge as significant. Even further, if variables operate similarly acro ss different trajectories, even those that suggest markedly different patterns (or rates) of offending, one could interpret such a finding as perhaps supporting Gottfredson and Hirschi assertion that there are no group differences. There are threats of creating a false dichotomy with such a scheme; however, recall that a major claim that Gottfredson and Hirschi make is that there are no career criminals. They argue against the need to disentangle va rious indicators across the life-cour se since the only factor that causes variation in offending is low self-control. If this were true, than th e risk factors employed here should not operate differently for different trajectory types. This would not be conclusive 105

PAGE 106

evidence of Gottfredson and Hirschi’s position but it would undermine the developmental contention of different risk factors being important for different types of offenders. Turning toward specific pred ictions of each theory, Moffitt (1993) posits that there are only two groups of offenders: LCP and AL. Accord ingly, the strongest pred ictors of LCP would be measures of individual and family characteris tics, which include health, gender, temperament, cognitive abilities, school achievement bonds, child-rearing practices, parent and sibling deviance, and socioeconomic status, “but not age” (Moffitt, 1993: 132). This study includes direct and indirect measures of some of her pr edictions including psychological status, academic grade, family structure and economic status, and family deviance. Though these may not be precise measurements, particularly regarding psychological status, it is conceivable that their combined effects would emerge as significant predictors if Moffitt were correct, net of all other factors. Herein this study builds on prior research in four ways. Firs t, using carefully selected independent variables, risk factors were invest igated to determine how they operate across and within racial/ethnic groups. Studies by Elliott (1994) and Ge and colleagues (2001) show that blacks persist in offending longer than whites pe rsist but they did not model persistence as studied here, nor did they include Hispanics in th eir research. Second, even if similar results to earlier studies are found, it remain s that further investigation is needed as to how risk factors comport with persistence in and desistance from offending. Third, it is important to restate the point that these data co me from a particularly criminal/ser ious population of juvenile offenders. Finding significant differences amongst a group of high-risk offenders could help illuminate our understanding of differences in offending after incarceration. Finally, be cause the data have variables from competing theories of crime, it was possible to investigat e the tenets of such 106

PAGE 107

theories within the same statistical models th ereby adding to our coll ective knowledge about the validity or practicality in understanding of criminal offending over time. The academic value and potential policy implications in addressing su ch suggestions make this study an important contribution to the literature. 107

PAGE 108

CHAPTER 7 RESULTS Data Analysis Data analysis took place in two phases. The fi rst phase presents descriptive and bivariate analyses (descriptive statistics are shown below in Table 7.1) to get an overall sense of the composition of parolees across important measur es in the study. Attent ion was then turned toward the multivariate models where analysis of offending trajectories over the seven-year postincarceration period are presented. Descriptive Statistics Of the 524 parolees in the sample, white parolees comprised th e largest racial or ethnic group with 254 (49%), followed by black parolees with 174 (34%) a nd lastly, Hispanic parolees with 87 (16%). Although it is pr eferable to have a more even distribution across race and ethnicity, SGM identified over 3,600 observations for data analysis since it is a person by period data set (i.e. 517 parolees 7 tim e-periods = 3,619 observations). An examination of IQ scores revealed that the overall sample was skewed. IQ scores ranged from a low of 64 to a high of 135. The av erage IQ score was 97 while the median was 97.99. About one-third of the parolees had above aver age IQ scores and nearly 12% had an IQ of over 112, which was greater than one standard deviat ion from the mean. This stands in contrast to nearly 20% of the sample th at had IQ scores greater than one standard deviation below normal. About 4% of the population scored a 75 or less, which is greater than two standard deviations below normal and generally considered the cut point for mental retardation (Kanaya, Scullin, and Ceci, 2003). When examined across race and ethnicity, white parolees had the highest average IQ at 101.81 while Hispanic pa rolees had the second highest average IQ at 108

PAGE 109

97.43. Black parolees had the lowest average IQ at 93.79. Analysis of variance revealed that IQ differences between racial groups we re significant at the .05 level. Table pendent measures and race/ethnicity 7.1. Descriptives for parolees across inde Independent Measures All ParoleesWhite Black Hispanic RACE/ETHNICITY N= 517 N= 254 N= 173 N= 87 JUVENILE MEASURES IQ Score [64-135]*** 97.99 (12.44)101.81 (12.04)93.79 (12.03)97.43 (11.90) Grade [0-14]** 10.34 (1.87)10.29 (2.04)10.65 (1.60)9.86 (1.82) Psychological Status ** 47.10% 50.30% 48.50% 34.40% Juvenile Alcohol Use 35.60% 36.50% 31.90% 42.20% Juvenile Drug Use ** 62.20% 58.80% 68.00% 62.50% Age at First Arrest [6-19] 12.59 (2.39)12.73 (2.43)12.43 (2.47)12.60 (2.18) FAMILY MEASURES Family Intact 46.80% 51.80% 43.50% 40.60% Family Welfare** 18.60% 11.80% 21.70% 29.70% Number of Siblings [0-12]*** 4.5 (2.73)3.79 (4.00)5.09 (3.09)4.98 (2.62) Father Criminality** 15.20% 19.40% 8.70% 18.80% Mother Criminality*** 6.60% 5.90% 8.00% 6.30% Sibling Criminality 54.30% 45.30% 65.20% 53.10% CYA MEASURES Length of Incarceration [2-147]***30.38 (24.95)27.45 (21.50)36.87 (26.78)24.69 (19.93) Escape Attempts*** 52.50% 63.90% 37.30% 53.20% Job Training 37.50% 37.10% 34.80% 45.30% Educational Training* 43.10% 36.50% 50.00% 46.90% POST-RELEASE MEASURES Post Release Alcohol Use [1-7]**1.76 (2.74) 1.65 (2.65)1.46 (2.60)2.46 (3.03) Post Release Heroine Use [1-7]***2.12 (2.64) 2.02 (2.61)1.64 (2.34)3.47 (2.88) Post Release Marriage [1-7]** 1.62 (2.23) 1.85 (2.34)1.36 (2.11)1.37 (2.04) Post Release Employment [1-7]*1.10 (1.56) 1.26 (1.66).93 (1.36)1.03 (1.62) Age at Release [16-22] 18.82 (1.07) 18.90 (1.14)18.73 (.96)18.79 (1.05)Statistical signidicance: * p < .10, ** p < .05, *** p < .01 Standard deviations are parenthesized for continuous variables Percentages are shown for dummy variables Of the parolees that were diagnosed with a psychological disorder, 47% had a personality disorder while the rest of the sample spread out among seven other di agnostic categories. Psychosis was the second most diagnosed at 3.6% followed by substance abuse at 2.1%. The 109

PAGE 110

remaining categories had less than 2%. Within r ace, both whites and blacks were evenly split between being diagnosed with a disorder or no t. On the other hand, Hispanic parolees were almost twice as likely not to have been diagnosed with a psychological disorder. Regarding substance use, the parolees were mo re likely than not to have used drugs and alcohol while juveniles. Slightly over half used both alcohol and drugs prior to incarceration. Most parolees who had used drugs as a juvenile did between two and fi ve different types of drugs with the average and median being two. Hi spanic parolees were the most likely to use drugs compared to blacks and whites, respec tively. Differences between racial groups were significant. The parolees were quite young when they were fi rst arrested as juveniles. They ranged in age from 6 years old to 19 years old. The mean age of first arre st was 12.6 and the mode was 13. In other words, half the parolees were early onsetters. Across race, whit e parolees’ had a mean age of 12.73, which was slightly older than both Hi spanic parolees at 12.60 and black parolees at 12.43. Given that family and individual risk factors su ch as drug or alcohol use have been linked to early onset, an examination of how the juvenile risk factors predicted age at first arrest was conducted. In a model not shown here, the result s of OLS regression identified a number of relationships. All else equal, paro lees whose families were not inta ct prior to incarceration were more likely to begin offending at younger ages th an those whose families were intact. Parolees with more siblings were more likely to be arre sted at an earlier age than those with fewer siblings. In other words, for every additional si bling, there was a -.12 decrease in age at first arrest. Though this decrease is seemingly small, it was significant at the .05 level. 110

PAGE 111

Adolescent drug use also significan tly predicted age at first a rrest. Parolees who had used drugs as juveniles were more likely to be ar rested at younger ages th an those who had not. Lastly, parolee grade was significant but at the .10 level. Parolees with lower a grade were more likely to be arrested earlier than were thos e with a higher grade. Although race was not a significant predictor in this model, some of the factors thought to be important in the life-course and developmental literature were significant an d they moved in the predicted direction. Turning to imprisonment histor y, parolees averaged 30 months in the CYA with a median of two years and a range between two mont hs and 147 months. The positively skewed distribution was a result of near ly 19% of the parolees servin g more than 48 months and 5% serving more than 72 months. In other words, a qua rter of the sample was more than one and two standard deviations from the mean, respectively. In contrast, half of the parolees that served two years or less were all within one standard deviation. Bivariate anal ysis by racial and ethnic group indicates that black parolees averaged 31.83 mo nths in the CYA, which was over ten months more than white parolees who averaged 21.43 a nd over one year longer than the Hispanic parolee average of 17.5 months. Nearly half of the parolees attempted at leas t one escape from the CYA with a majority of them never attempting an escape, and one parolee attempting ten. Differences across race and ethnicity were not statistically significant. Parolees differed by educational and job training received while in the CYA. Of the 524, 26% of the parolees received vocational training and over 30% received educational training. Only 16.6% of the parolees re ceived both types of training. Differences by race were not significant for vocational training but were for educational training. Within respective racial/ethnic groups, bl ack parolees and to a lesser degree, Hispanic 111

PAGE 112

parolees were more likely than white parolees to have received some form of educational training. An examination of all five post-release variables showed that they were positively skewed. For example, 31.3% of parolees used alcohol at least once during the se ven-year follow up. The mean was less than two and the mode was zero. Le ss than 16% of the parolees used alcohol over all seven years with a quarter of the sample using alcohol for four or more years. Within race and ethnicity, Hisp anic parolees had the highest average of 2.46 followed by white parolees at 1.65 and black pa rolees at 1.46, respectively. Diffe rences in alcohol use by race were statistically significant. In contrast, slightly less than half of the paro lees (48%) used heroin at least once after release with a third of the parolees using two or more times. Similar to alcohol use, Hispanic parolees had a mean heroin use of 3.47, which made them more likely to use heroin than both white parolees at 2.02 a nd black parolees at 1.65. Differences between racial/ethnic groups in heroin us e were statistically significant. Two post-release stakes in conformity measur es were employed in the study: marriage and employment. Regarding marriage, 44% of the paro lees were married at some point during the seven year follow-up period (it was also the mo dal category). Parolee marriage had a mean of 1.62 and a standard deviation of 2.23. Less than a qua rter of the parolees (115) were more than one standard deviation from the mean. Although th e differences were statistically significant, white parolees, with an average of 1.85, were onl y slightly more likely to be married than Hispanic parolees or black parolees w ho had equal means, which were 1.37 and 1.36, respectively. The parolees were somewhat less likely to be employed during this period. Only 53% of the parolees were employed at some point during the seven years after release. Fewer than 50 112

PAGE 113

parolees (9%) were employed at times four through seven and only 5% of the parolees were employed the entire time under study. When ex amined across race, employment seemed to follow what has been reported in the literature . White parolees, with a mean of 1.26, were the most likely to be employed. Hispanic parole es had the second highest mean of 1.03. Black parolees were not likely to be employed even once during the seven ye ars. The black parolee employment mean was .093. Differences between racial and ethnic group s were statistically significant at the .10 level. Analysis of Variance Table 7.2 presents the one-way ANOVA and mean offending by race and ethnicity results for trajectory models for all crime types. Oneway ANOVA was employed to ensure that there was significant variability between race and ethnicity across the full sample of parolees for total, non-violent, and violent crime type s. Variability was statistically significant for violent and nonviolent offending by race at p < .10 but was not statistically significant for total offending. Table 7.2. Mean number of offenses by crime type and race Race White Parolees (n=253)14.281.3212.95 Black Parolees (n=172)14.262.9611.29 Hispanic Parolees (n=87)16.102.0713.91 All Parolees (n=512) 14.582.0212.55 F 1.2326.68**2.70*Non-violent Total Violent CRIME TYPE Multivariate Analysis Figures 7.1 through 7.12 show the results of SGM for all crime types by each racial and ethnic group. Tables 7.3 through 7.14 are a seri es of binomial and multinomial logistic regression analyses, each of which examines se parate outcomes for trajectory group membership 113

PAGE 114

for total, non-violent, and viol ent crime. For each crime type, there are four models: a full model that includes race/ethnicity as an independent measure and three separate models that examine trajectory group membership for the white , black, and Hispanic samples. Notably, in a similar study by Chung and coll eagues (2002), SGM trajec tory identification was constructed according to age of onset and de sistance. However, this study diverges from the Chung and colleagues study and other similar st udies by identifying trajectories based on the seven-year offending patterns. Age of onset was employed as an independent variable not as a factor in determining trajectories in the first place. Furthermore, an examination of the different trajectory models appears to show very similar trajectory patterns, however, this is not necessarily the case. Based upon group percentages, which indicate the percentage of the sample belonging to each group and are illustrated at the bottom of each trajectory figure, and the fact that simila r appearing trajectories occur at different rates of offe nding, leads to interpretations. This is an important distinction because the conclusions drawn from the offe nding trajectories must be sensitive to the percentage of those in the offending group and th e level of offending. With in trajectory models and across trajectory models in this study, ther e were similar trajector ies of offending, but as discusses below, a more thorough analysis reve als that the group percentages and levels of offending markedly differ. Moffitt’s theory addresses this issue. Her theo ry is predicated on the size of the offending group and the type of offending patterns different groups’ exhibit. A particular example of this can be found in the violent crime trajectory mode ls section below, which identified a two-group model for each of the samples. Both the per centages in each model differ by group and by the 114

PAGE 115

offending levels of the individual trajectories. In turn, differences in levels of offending led to different conclusions. Total Crime Figures 7.1 through 7.4 and Tables 7.3 through 7.6 present the results for total crime trajectories. Total crime is the sum of both non-violent and violent crimes employed in the later models. Offending patterns were identified usin g SGM for the full sample and for separate samples of white, black, and Hispanic parolees. Total Crime Full Sample Illustrated in Figure 7.1, SGM id entified a four-group trajectory model for the total crime full sample model. Although a five-group trajectory model had a marginally lower absolute BIC and because nearly identical pa tterns of two of the trajectories, a four-group trajectory was employed in this model. Figure 7.1. Total crime full sample trajectory model 115

PAGE 116

There is a second compelling st atistical reason for choosing th e four-group as well. Long (1997) points out that from a statistical standpoint, increasing th e categorical values in the dependent variable decreases the po wer of the model. This is a re sult of how multinomial logistic regression operates. Simply, a seri es of binary logistic regressi ons are regressed on one of the categorical groups. That is, if ther e are five values, then each valu e in the dependent variable is a 0/1 binary category. In the ensuing iterations, multinomial logistic regression uses one of the categories as a reference group fo r which the other categorical va lues are compared. Thus, the more categories in the dependent variable the more complex the underlying statistical process becomes. Thus, the less predictive power it has. Re stricting the groups for this reason is not only theoretically acceptable but statistically supported. As shown in Figure 7.1, SGM identified four trajectory groups. Group 1 is a chronic group that steadily committed at least one crime per year over the seven years. Group 2 is an escalation group. At Time 1, they committed about 1.7 crimes but by Time 7 they had increased to nearly three crimes. Group 3 was an initi al high rate group th at deescalated in offending. At Times 1 and 2, they were committing over four crimes pe r year, but by Time 7 their offending decreased to about one crime. Finally, Group 4 is a highchronic group. They star ted out at over four offenses at Time 1 and were at nearly six offe nses by Time 7. The groups are labeled as follows: 1) Low Chronic, 2) Escalation, 3) De-e scalation, and 4) High Chronic. Table 7.3 presents the results of multinomial l ogistic regression of total crime for the full sample. Eight variables emerged as significant. Race/Ethnicity was significant for the white parolees in Group 3. The coefficient was positive, indicating that white parolees were more likely to be in the De-escalation grou p compared to the High-Chronic group. 116

PAGE 117

117 Among the juvenile measures, IQ, Juvenile Alcohol Use, and Juvenile Drug Use were significant. The sign of the coefficient for IQ was significant and pos itive for Group 1. This indicates that parolees with higher IQs were more likely to be in the Low-Chronic group compared to the High-Chronic group. Juvenile Alcohol Use was positive and significant for Group 3. This suggests that parolees that were more likely to use alcohol when they were juveniles were more likely to be in the De -escalation group compared to the High-Chronic group. Conversely, Juvenile Drug Use was signi ficant but negative for Group 3. This would indicate that parolees that were less likely to use drugs as a juvenile were more likely to be in the De-escalation group than the High-Chronic group. Next, among the Family Measures only Number of Siblings was significant for the full sample. The positive sign of the coefficient for Group 1 indicates that pa rolees that had more siblings were more likely to be in the LowChronic group compared to the High-Chronic group. Among the CYA Measures, Escape Attempts was significant for Group 2. The coefficient was negative, which indicates that parolees that were less likely to attempt an escape while incarcerated were more likely to be in the Escalation group compared to the High-Chronic group. Finally, among the Post-release Measures, onl y two variables emerged as significant. Both Post-release Heroin Use and Marriage were si gnificant but for two different groups and in a different direction. The coeffici ent was negative for Heroin Use for Group 1, which indicates that parolees that were less likely to use heroin af ter release were more likely to be in the LowChronic group compared to the High-Chronic group. For Marriage, the sign of the coefficient for Group 2 was positive. This indicates that parolees that were more likely to be married after release from the CYA were more likely to be in the Escalation group compared to the HighChronic group.

PAGE 118

118 Table 7.3. Total crime full sample multinomial logistic regression model GROUP 1 GROUP 2 GROUP 3 Low Chronic Low Escalation De-escalation (n=143) (n=169) (n=151) BEx p (B)WaldBEx p (B)WaldBEx p (B)Wald RACE/ETHNICITY White 0.6121.8450.8060.7732.1671.483 1.1763.2403.339*Black -0.1920.8250.0650.3841.4680.3270.3271.3870.230 JUVENILE MEASURES IQ Score 0.0611.0636.936**0.0081.0080.1510.0311.0322.175 Grade 0.0051.0050.0010.1141.1210.736-0.0390.9610.093 Psychological Status 0.2241.2510.1640.7682.1562.2540.3811.4640.548 Juvenile Alcohol Use 0.5131.6690.7390.1661.1810.093 0.8922.4412.586*Juvenile Drug Use -0.7040.4951.098-0.9540.3852.211 -1.0640.3452.771*Age at First Arrest 0.1211.1291.2570.0721.0750.5250.0161.0170.028 FAMILY MEASURES Family Intact 0.4321.5400.690-0.1500.8610.1020.5391.7141.266 Family Welfare -0.7310.4811.081-0.4760.6210.514-0.4040.6670.364 Number of Siblings 0.3171.3738.781**0.1201.1281.4710.1111.1171.211 Father Criminality 0.3701.4480.2980.1311.1400.0460.8812.4142.019 Mother Criminality -3.1910.0411.305-3.7340.0241.846-3.3870.0341.508 Sibling Criminality 0.7572.1321.929-0.0230.9770.0020.4141.5120.686 CYA MEASURES Length of Incarceration -0.0070.9930.395-0.0110.9891.281-0.0060.9940.309 Escape Attempts -1.2070.2991.072 -1.1570.3155.593**-0.9280.3952.469 Job Training -0.8650.4212.504-0.4010.6690.646-0.3790.6850.554 Educational Training -0.0250.9760.002-0.5280.5901.232-0.5950.5511.527 POST RELEASE MEASURES Post Release Alcohol Use -0.0450.9560.240-0.0480.9530.3380.0671.0690.652 Post Release Heroin Use -0.3120.7327.878**-0.0670.9360.457-0.1980.8201.897 Post Release Marriage 0.1311.1391.052 0.2041.2263.047*-0.0200.9800.026 Post Release Employment 0.2861.3312.375-0.0870.9170.2240.0661.0680.132 Age at Release 0.1601.1730.459-0.2080.8120.9360.1941.2140.807Group 4 (High Chronic) is the reference group (n=54) Statistical signidicance: * p < .10, ** p < .05

PAGE 119

119 Total Crime White Sample Figure 7.2 presents the SGM results of th e total crime white parolee sample. SGM identified a four-trajectory group model as havi ng the lowest absolute BIC. Group 1 was a lowrate chronic group that committed about one crime per year across all seven years. Group 2 was a de-escalating group that initially committed about three crimes at Time 1 but by Time 7 they decreased in offending to about one crime. Gr oup 3 was a high-chronic de-escalating group that committed over six crimes at Time 1 but de-es calated to just under three crimes by Time 7. Group 4 was an escalating group that started out at about 2.5 crimes at Time 1 but by Time seven they increased to about 5 crimes. Based on the o ffending trajectories, the groups were labeled as follows: 1) Low Chronic, 2) De -escalation, 3) High Chronic De-e scalation, and 4) Escalation. Figure 7.2. Total crime white sample trajectory model Table 7.4 presents the results of multinomial logistic regression for total crime white parolee sample. Nine variables emerged as signif icant. Among Juvenile M easures, IQ and Age at

PAGE 120

120 efficient for IQ was positive for Groups 1 and 2, indicati t that g was significant for Group 1. The sign of the co efficient was negative for offending Group 1. This would indicate that white CYA parolees that were less likely to receive job training while incarcerated in the CYA were more likely to be in the Low-Chronic group compared to the Escalation group. Among the Post-release measures, three variab les emerged as significant. Heroin Use was significant and negative for Group 2, which i ndicates that white parolees that were less likely to use heroin were more likely to be in the De-escalation group compared to the Escalation group. Marriage was positive and significant for Group 1, indicating that white parolees that were more likely to be married were more lik ely to be in the Low-Chronic group than the Escalation group. The last signifi cant independent variable in th e white sample full model was Age at Release. The coefficient was negativ e for Group 3, which indicates that younger white to be in the High-Chronic Deescalati First Arrest were significant. The sign of the co ng that white parolees wi th higher IQ scores were more likely to be in the Low Chronic or De-escalation groups compared to the Escalati on group. The coefficient for Age at First Arres was positive for Group 1, indicating that white parolees that were older at first arrest were more likely to be in the Low-Chronic gr oup compared to the Escalation group. Among Family Measures, both Father Criminality and Sibling Criminality were significant and negative for white parolees. Th is indicates that white parolees whose father and siblings were less likely to be criminal were more likely to be in the High-Chronic De-escalation and Deescalation groups, respectively, comp ared to the Escalation group. Only one independent variable emerged as significant among CYA Measures. Job Trainin parolees at time of release fr om the CYA were more likely on group compared to the Escalation group.

PAGE 121

121TTable 4. al e itemp logisticren odel 7.Totcrimwh sale multinomial regssiom GOUP 1 Loroic (=90) Bxp(B ENIE MESURS ore 0.101107 0.188207 hologal Stus 0.024024 ile Aoholse 0.283328 ile Dug Us 0.01983 t Fi Arre 0.330391 ILY MEAURE ly Intct 0.12883 ly Wfare 0.27757 ber oSiblin 0.227255 r Criinalit 3.08046 er Cminaly 1.008740 inali 1.11327 SURS ncarcation 0.00998 mpts 0.203225 1.68185 Traing 1.178249 EEASURS e Alchol Ue 0.01988 e Herin Us 0.27762 e Maiage 0.346413 e Emoymet0.316372 ase 0.672958ferencGro=32) e: * R RP 2 w Chn -et hoela n n) () E)W E)W E)W JUVLAE IQ Sc 1.7.266** 0.71.6.** 4 0. Grade 1.0.909 01.2. 8 0. Psycicat 1.0.001 20.0. 89 0. Juvenlc U 1.0.090 20.0. 9 0. Juvenre -70.0.000 62.0. 56 1. Age arstst 1.3.016* 41.0. 8 1. FAMSS Famia -40.0.023 70.1. 6 0. Famiel -80.0.052 92.0. 44 0. Numf gs 1.1.551 31.0. 5 0. Fathemy -80.2.354 20.2. 83 3. * Mothriit 2.0.265 18.1. 2 0. Sibling Crimty -80.1.413 50.3.** 77 1. CYA MEAE Length of Ier -20.0.007 21.0. 48 1. Escape Atte 1.0.049 30.0. 29 0. Job Training -50.3.960** 70.1. 84 2. Educationalni 3.2.021 72.1. 7 0. POST RELASE ME Post Releasos -20.0.005 61.0. 8 0. Post Releasoe -10.1.916 00.5.** 25 0. Post Releasrr 1.3.363* 21.0. 95 0. Post Releaspln 1.1.267 41.1. 8 0. Age at Rele 1.2.084 71.2. 6 3. *Group 4 is Ree up (n Statistical signidicanc GROUP3 Chrnic D-escator n=25 xp(Bald 1.014091 1.019007 0.749086 2.157464 0.191073 1.269163 1.561218 0.641101 1.057074 0.034175 5.430584 0.279376 0.954875 0.972001 0.227031 1.563240 1.147576 0.883381 0.823655 1.059031 2.307440 or Hig aldB 828 0.01 395 0.01 671-0.2 1890.76 513-1.6 1700.23 6810.44 369-0.4 6390.05 351 -3.3 0781.69 910 -1.2 966-0.0 007-0.0 188-1.4 5900.44 2520.13 683 -0.1 859-0.1 2500.05 165 0.83 GOU Descala (=107 aldBxp(B 09101 0.28323 -0.70495 -0.40669 0.84329 0.07076 -1.02358 0.72072 0.14154 -2.75064 2.12336 -1.81163 0.02023 -0.07929 -0.89408 0.99710 0.07079 -0.44644 0.17188 0.31369 0.56762 < .05 < ** .10, p p

PAGE 122

122 Total Crime Black Sample Figure 7.3 and Table 7.5 present the results of the total crime blac k parolee sample. SGM identified a four trajectory group model base d on the lowest absolute BIC. Based on the individual patterns of offending trajectories, Group 1 was a low-chronic offending group that committed about one crime per year across all seven time periods. Group 2 was an escalating group that committed two crimes at Time 1 but by Time 7 they were committing more than three crimes. Group 3 was a de-escalating group that starte d out at a high rate of five crimes at Time one but by Time 7 they decreased to about one crime. Group 4 was a high-chronic escalation group that started at two crimes at Time 1 but by Time 7 they were committing over eight crimes. e black trajectory model Figure 7.3. Total crim

PAGE 123

123TT 7.5. Total crime black sample multinomial logistic regression model GROUP 1 GR OUP GROUP3 Descalaion n=50 aldBxp(BWald 2250.061.071733 0020.301.355533 006-0.100.895006 0540.952.594431 662-0.890.445353 483-0.380.735107 1341.163.201883 138-3.690.026508 0350.271.316087 0282.674.46362 286-1.490.230111 3040.211.244026 667-0.020.949038 795-0.760.484390 206-0.000.990000 033 -2.500.078138 158-0.060.994001 158-0.220.770783 218-0.200.779739 250-0.450.628757 2570.261.303172 2 Low Chronic Escalation -et (n=47) (n=67) () BExp(B)WaldBExp(B)W E) LE MEASURES Iore 0.0531.0550.9090.0231.0230. 8 1. 0.5711.7711.546-0.0180.9820. 4 0. Pological Status 0.1301.1390.007-0.1020.9030. 1 0. Jile Alcohol Use -2.0090.1341.141-0.3140.7310. 3 0. Jile Drug Use 0.4291.5360.077-1.0140.3630. 0 0. t First Arrest 0.3641.4400.909-0.1810.8340. 0 1. ILY MEASURES Fy Intact 2.64714.1133.013*0.4321.5400. 3 0. Fy Welfare -4.9520.0072.567*-3.1290.0441. 3 1. er of Siblings 0.3391.4041.4010.0471.0480. 5 1. Father Criminality 2.0577.8210.6150.2971.3460. 2161. Mother Criminality -1.1020.3320.053-2.3350.0970. 6 0. Sibling Criminality 1.7015.4771.188-0.7000.4970. 8 0. CYA MEASURES Length of Incarceration -0.0410.9602.835*-0.0290.9712.*5 5.**Escape Attempts -3.2680.0383.716*-0.9870.3730. 2 0. Job Training 1.5284.6100.8400.6001.8220. 1 0. Educational Training -2.0080.1341.441-1.9470.1432. 5 3.*POST RELEASE MEASURES Post Release Alcohol Use-0.1260.8810.364-0.1690.8451. 0 0. Post Release Heroin Use -1.1910.3046.458**-0.2970.7431. 6 0. Post Release Marriage -1.1050.3315.498**-0.1230.8840. 5 0. Post Release Employment 1.7125.5414.347**-0.5530.5751. 6 0. Age at Release -1.3260.2653.066*-0.2970.7430. 5 0.Group 4 is Reference Group (n=14) Statistical signidicance: * able JUVENI Q Sc Grade sych uven uven Age a FAM amil amil Numb < .05 < .10, ** p p

PAGE 124

As illustrated in Table 7.5, eleven variables emerged as significant. None of the Juvenile Measures achieved statistical si gnificance at either the .05 or .10 level, however, two Family Measures did; Family Intact and Family Welfar e. The coefficient for Family Intact was positive and significant for Group 1, which indicates that black parolees that were raised in households with both parents were more likely to be in the Low-Chronic group than the High Chronic Escalating group. Family welfare was significant but the sign of the coefficient was negative for Group 1. This indicates that black parolees that had a family more likely to not receive welfare when they were juveniles were more likely to be in the Low-Chronic group compared to the High-Chronic Escalating group. Among CYA Measures, three variables emerged as significant. Length of Incarceration was significant and negative for Groups 1 thr ough 3, which indicates parolees that were rcerated for shorter periods were more likely to be in the Low-Chronic, Escalation, or Delation group compared to the High-Chroni c Escalation group. Escape Attempts was also ificant for Group 1. The negative coefficient indica tes that parolees that were less likely to mpt an escape while in the CYA were more likely to be in the Low-Chronic group than the h-Chronic Escalation group. E ducational Training was signif icant and negative for Group 3, ch suggests that black parolees that were less likely to receive educational training were more ly to be in the De-escalation group compared to the High-Chronic Escalation group. Four variables were signifi cant among Post-release Measures. Heroin Use, had a negative fficient for Group 1. This would indicate that black parolees that we re less likely to use in after release were more likely to be in Low-Chronic group compared to the High-Chronic alation group. Likewise, Marriage was also negative and significant, which indicates that k parolees that were less li kely to be married were more likely to be in the Low-Chronic inca esca sign atte Hig whi like coe hero Esc blac 124

PAGE 125

group wtified by SGM based on the lowest absolute BIC. Group 1 was a low t Time p es trajec ile than the High-Chronic Escalation group. Empl oyment was also significant. The negative coefficient for Group 1 would suggest that black parolees that were less likely to be employed were more likely to be in the Low-Chronic group compared to the High-Chronic Escalation group. Finally, Age at Release was also signific ant and negative, which indicates that black parolees that were released from the CYA at a younger age were more likely to be in the Lo Chronic group compared to the High-Chronic Escalation group. Total Crime Hispanic Sample Figure 7.4 and Table 7.6 present the results of the total crime Hispanic parolee sample. Three trajectory groups were iden -chronic offending group that committed about one crime per year in each of the seven years after release. Group 2 was a de-escalating group that committed about four crimes a 1 but by Time 7 they decreased to about one crime. Group 3 was a high-chronic offending grou that initially committed about three crimes at Time 1 but by Time 5 they increased to nine crim and then decreased to about five crimes by Ti me 7. As such, the three groups are labeled as follows: 1) Low-Chronic, 2) De-e scalation, and 3) High-Chronic. As shown in Table 7.6 three variables emerged as significant in the total crime Hispanic parolee multinomial logistic regr ession model. This is perple xing given the amount of crime committed and the strong probabilities of Hispanic parolees belonging to their respective tory groups. On the other hand, as discusse d below in the violent and non-violent offending models, disaggregation of offending by crime type seems to illustrate stronger correlations between the independent variables and trajectory group membership. Among Juvenile Measures, only one independent va riable emerged as significant. Juven Drug Use was significant and negative for Gr oup 2. The negative coefficient for Group 2 125

PAGE 126

indicates that Hispanic parolees who were less lik ely to use drugs as juveniles were more likely to be in the De-escalation group co mpared to the High-Chronic group. Escape Attempts was the only independent variable that emerged as significant among CYA Measures. The negative sign on the coefficien t for Group 1 indicates that Hispanic paro who were less likely to attempt an escape were more likely to be in the Low-Chronic group tha the High Chronic group. the lees n ease Measures, onl y Heroin Use emerged as significant. The negat ly Finally, among the Post-rel ive sign on the coefficient for Group 1 suggests that Hispanic parolees that were less like to use heroin after release from the CYA were more likely to be in the Low-Chronic group compared to the High-Chronic group. Figure 7.4. Total crime Hispanic sample trajectory model 126

PAGE 127

Table 7.6. Total crime Hispanic sample multinomial logistic regression model GROUP 1 Low Chronic GROUP 2 De-escalation 2.1628.6881.0071.7055.4990.946 -2.4500.0860.775 -4.1940.0152.825*Age at First Arrest 0.4331.5420.8100.2531.2880.731 FAM Family Intact -2.4510.0860.758-0.2250.7980.015 Number of Siblings 0.2511.2850.332-0.3200.7260.656 Mother Criminality -4.6880.0090.663-5.0150.0070.762 0.3160.2321.0302.8010.337 CYA MEASURES Length of Incarceration 0.0441.0450.421-0.0500.9510.888 Escape Attempts -3.9170.0204.307**-0.9470.3880.413 Job Training -2.4860.0831.486-1.9760.1391.351 Educational Training -0.7270.4830.182-0.6670.5130.183 POST RELEASE MEASURES Post Release Alcohol Use-0.3280.7200.9110.2091.2330.593 Post Release Heroin Use -0.9090.4032.421 -0.5950.5521.686 Post Release Marriage 0.4731.6050.886-0.4660.6270.983 Post Release Employment0.9692.6351.626-0.3820.6820.482 Age at Release -0.8120.4440.4161.2523.4991.815Group 3 is the Reference Group (n=13) Statistical signidicance: * (n=38) (n=35) BExp(B)WaldBExp(B)Wald JUVENILE MEASURES IQ Score 0.0621.0640.5610.0711.0740.870 Grade 0.3061.3570.167-0.4670.6270.574 Psychological Status 2.51712.3900.969-1.4390.2370.446 Juvenile Alcohol Use Juvenile Drug Use ILY MEASURES Family Welfare 0.3411.4060.0240.0121.0120.000 Father Criminality 1.3663.9210.3432.42311.2761.399 Sibling Criminality -1.150 p < .10, ** p < .05 Non-violent Crime As illustrated in Figure 7.5, SG M identified a four-group trajectory model. However, noticeably different than earlier models was th e amount of variation of the actual trajectory around the expected trajectory for one of the groups. Unlike the firs t three groups, Group 4 deviated much more around the expected trajectory (t he dashed line). This is mostly like because Group 4 is a very small trajectory group in te rms of group size (see Table 7.3). Although this indicates that tr ajectory models are should not affect interpretation of the models, the literatu re 127

PAGE 128

stronger when there is le ss error between the expected and actual trajectories. In any event, this was not a problem in this study since there were very few individual tr ajectory groups that noticeably varied around thei r expected trajectories. Non-violent Crime Full Sample As illustrated in Figure 7.5, ther e are four trajectories. Group 1 was a low-rate chronic offending group that steadily committed about one non-violent crime per year for each of the seven years after release. Group 2 was an escala ting offending group that initially committed less than two non-violent crimes at Time 1 but by Time 7 they were committing more than three offenses per year. Group 3 was a deescalating o ffending group. Between Time 1 and Time 3 they committed more than three non-violent crimes per year but by Time 7, their offending decreased by more than half to about one per year. Gr oup 4 was a high-rate chr onic offending group. This group initially began offending at slightly more than four non-violent crimes at Time 1. They then increased to over five crimes by Time 2, de creased slightly by Time 3 but then increased sharply between Times 3 and 5 to nearly eight non-violent offenses per year. Between Times 5 and 7, they decreased back to about five crimes but were still the highest rate offenders in the model. As such, the four groups are labeled as follows: 1) Low Chronic, 2) Escalation, 3) Descalation, and 4) High Chronic. e results from multinom ial logistic regression for the non-violent crime full sample. Nine variables emerged as significant. Race was significant and positive for blacks in the Escalation group compared to the High Chronic group. Specifically, black parolees were more likely to be in the Escalation group compared to the High-Chronic group . Among the juvenile risk factors, two variables were significant. IQ was positive and significant for Group 1, which suggests that parolees with hi gher IQs were more likely to be in the LowChronic group than the High-Chronic group. Age at first arrest (age of onse t) was also significant e Table 7.7 summarizes th 128

PAGE 129

129 lings was significant at either the .05 or s parolees that had more sibli ngs were more likely to be in the Es ng Figure 7.5. Non-violent crime full sample trajectory model calation group compared to the High-Chronic group. Two incarceration variables were significant. Both Escape Attempts and Job Training were positive and significant. This indicates that parol ees that attempted more escapes from the CYA were more likely to be in the Low-Chronic group . Likewise, parolees that received job traini were more likely to be in the Low-Chroni c group compared to the High-Chronic group. .10 level. The positive sign indicate and positive, which indicates that parolees that were arrested for a later in adolescence were more likely to be in the Escalation group compared to the High Chronic group. Among the family measures, only Number of Sib

PAGE 130

130 Table 7.7. Non-violent crime full sample mu ltinomial logistic regression model Grou p 1G r o u 2G r o u * p p 3 Low Chronic Escalation Grou p Dee sconou alati Gr p (n=241) (n=96) 143) BEx (n= Ex p p (B)WaldBEx p (B)ldB Wa (Bd JUVENILEASURES White 0.3141.3690.1991.2563.51116.6682 Black 0.3391.4030.175 1.6895.41601 .0086 IQ Score 0.0501.0523.914** 0.0141.01457.0382 Grade 0.1061.1120.5970.1421.15273.0623 Psychologitatus 0.7342.0831.4710.5241.68956.2166 Juvenile Al Use -0.0220.9780.001-1.0220.36035.2480 Juvenilee -0.7040.4950.929-0.4620.63024.9971 Age at Firsest 0.1501.1621.791 0.2211.24899 .0733 FAMILY SURES Family Inta 0.0581.0590.0110.3351.39823.5859 Family We 0.1941.2140.0760.2781.32134.4870 Number ofngs 0.1821.1992.552* 0.0831.08662.0544 Father Crity -0.6640.5150.644-0.8120.44447.2889 Mother Crlity -3.2170.0401.370-1.8520.15730.5811 Sibling Criity 0.2241.2510.1390.5061.65929.5945 CYA MEURES Length of Incarceration 0.0081.0080.408-0.0190.98129.00.9894 Escape Attempts -1.5980.2026.587-1.0090.36587.90.3901 Job Training -1.0200.3612.774* -0.7350.48067.50.5629 Educational Training 0.3491.4170.3940.7282.07180.10.8273 POST RELEASE MEASURES Post-Release Alcohol Use0.0721.0750.4660.0151.01518.11.1182 Post-Release Heroin Use -0.2210.8023.439* -0.0710.93215 .20.8013 Post-Release Marriage 0.1901.2101.7520.1241.13273.01.0233 Post-Release Employment 0.4371.5483.106* 0.1381.14870.31.3983 Age at Release -0.0620.9400.061 -0.4620.63008 .11.1182Group 4 High Chronic (n=36) is the Reference Group Statistical signidican )Wal 1.90.85 2.81.58 1.02.09 1.00.18 1.30.19 1.20.11 0.31.48 1.00.38 1.71.03 1.50.40 1.00.20 0.70.07 0.00.82 1.70.93 0.63 2.21 0.83 0.11 1.08 3.34 0.02 1.79 0.19 2.5 077 3.7* 133 0.2 037 0.8 060 0.6 075 2.1 022 0.3-023 3.1* 070 0.3 080 0.1 062 0.4 053 0.8-039 0.4-210 0.6 085 1.7-011 2.3 -042 1.2-075 1.4-090 0.0 012 0.3 -022 0.6 023 0.2 035 2.9* 011 < .10, ** p p ce: * E M cal S lcoho Drug Us t Arr MEA ct lfare Sibli minali imina minal AS < .05

PAGE 131

Among the post-release measures, Heroin Use and Post-release Employment significantly predicted membership in the Low-Chronic grou p compared to the High-Chronic Group. Heroin use was negative, suggesting that pa rolees that were less likely to use heroin were more likely to be in the Low-Chronic group. Heroin Use was al so significant and negative for Group 3, which indicates that parolees that were less likely to us e heroin after release were more likely to be in the De-escalation group compared to the High-Ch ronic group. Employment was also negative. Parolees that were less likely to be fully empl oyed were more likely to be in the Low-Chronic group compared to the High-Chronic group. Non-violent Crime White Sample Figure 7.6 and Table 7.8 summarize the resu lts of SGM and multinomial logistic ession for the non-violent crime white pa rolee sample. SGM identified four offending ectory groups based on the lowest absolute BIC. Group 1 was a low-chronic offending group committed one crime per year for all seve n years. Group 2 was an escalating trajectory up that initially offended at one crime at Ti me 1 but by Time 6, increased to about four es. By Time 7, this group decreased to about three crimes. Group 3 was a de-escalating ectory group that initially committed between th ree and four crimes at Time 1 but by Time 7 decreased to about one crime. Group 4 was a high chronic offending group that committed een four and six crimes each year between Times 1 and 7, however, as Figure 7.6 illustrates, actually trajectory varied around the expect ed trajectory more noticeably than the other ups. This did not affect interpretation of the results or the trajectory group model. Further lysis revealed that nearly all of the pa rolees had a .7 or higher posterior membership bability. Nonetheless, based on the trajectory pa tterns illustrated in Figure 7.6, the four ectory groups as were labeled as follows: 1) Low Chronic 2) Escalation 3) De-escalation 4) h Chronic. regr traj that gro crim traj they betw the gro ana pro traj Hig 131

PAGE 132

132 Figure 7.6. Non-violent crime white sample trajectory model Table 7.8 presents the results of multinomial logistic regression for the non-violent c white sample. Thirteen variables emerged as signi ficant in the model. Am ong Juvenile Mea IQ and Juvenile Drug Use were both significant. IQ was positive, which indicates that white parolees that had higher IQ scores than lower IQ scores were more likely to be Low Chronic offending group than in the High-Chronic offendi ng group. Likewise, white parolees that were more likely to use alcohol as a juvenile were more likely to be in the Low-Chronic offending group than the High-Chronic offending group. Among the family measures, only Sibling Criminality was significant, but it was for bo the Low-Chronic group and the Escalation group. The positive sign for su ggests that whi parolees whose siblings were more likely to have a criminal record were more likely to be i Low-Chronic and Escalation offending groups than the High-Chronic offending groups. rime sures, th te n the

PAGE 133

133TTable 7.8. Non-violent crime white sampmu ltinomial logistic regression model le GROUP 1 Loronc (n102) BE(B)Wld 0871.914.80**.111950.29 6431.030.18 .720870.26 1698.472.86*0201.200.09 .244840.63 .979760.27 2541.891.18 .576071.60 .520190.99 7065.053.65*048492.28 .7680.713.86*.2370.074.68**37016945.61**.063660.18 .079240.86 5351.084.63**9082.793.28*.2260.970.87nce Gup (n6) GROUP 2 GROUP 3 w Chi Escalation De-escalation = (n=40) (n=86) xpa Exp(B)WaldBExp(B)Wald JUVEEAE IQ Sc 0.00 1.0050.0140.0201.0200.239 Grade -00.82-20.9980.000-0.1770.8380.690 Psychcatu 0.94 2.6570.918-0.0210.9790.001 JuvenicoUs -00.45-30.3980.763-0.2920.7470.097 Juveniue 2.75 1.0040.0000.6531.9220.254 Age att t 0.00-30.9770.0110.0011.0010.000 FAMMU Familyct -00.70 1.3180.0720.1431.1540.024 Familyfa -00.35-30.8750.0090.3081.3610.049 NumbSigs 0.25 1.0930.1590.0921.0960.217 Fatheri -10.23-60.1331.898-1.3530.2591.089 Mothemy -10.28-20.5650.0833.74742.4050.699 Sibliniminy 1.51 5.4132.673*0.7572.1310.734 CYA SES Lengtnra 0.1.04-70.9740.5830.0241.0250.670 Escape -110 -30.9390.003-1.2590.2841.648 Job Trg -218 -0.6810.128-1.0210.3601.178 Educatl in 2.0.7 1.2893.6291.5251.1563.1791.552 PO E SES Post-Rsehe01.01 0.3011.3522.592*0.2161.2411.613 Post-Rseoie-00.91 0.0161.0160.006-0.1320.8760.616 Post-Rseri 0.78 0.5481.7314.674**0.2291.2580.961 Post-Rselnt 0.46 0.8192.2692.875*0.9062.4753.889**Age ata -073-0.5820.5590.559-0.2080.3800.380Group ic dehe Referero=2 Statisticnce B 0.005 0.00 0.977 0.92 0.004 0.02 0.276 0.13 0.089 2.01 0.57 1.689 0.02 0.06 0.384 NIL ME ore ologil Sta le Alhol le Drg Us FirsArres ILY EAS Inta Welre er of blin Crimnality r Criinalit g Cralit MEAUR h of Icarce e Attmpts ainin ionaTrain ST RELEAS elea Alco elea Her elea Mar elea Emp Relese4 HighChron al sigidican SURS s e RES tion g MEAUR ol Us ne Us age oymeoffenng is t : * p < .05 < .* 10, * p

PAGE 134

Three variables emerged as significant am ong the CYA Measures. Escape Attempts was negative, which indicates that white parolees that were less likely to attempt an escape from the CYA were more likely to belong to the LowChronic offending group than the High-Chronic group. Both Job Training and Educational Traini ng were also signifi cant. The sign for Job Training was negative, whic h indicates that white parolees that did not receive job training were more likely to be in the Low-Chronic group than the High-Chronic group., On the other hand, white parolees that did receive educational training were also more likely to be in the LowChronic group than the High-Chronic group. Six variables were significant among the Po st-release Measures. Alcohol Use was positive significant for group two, whic h indicates that white parolees who were more likely to use hol after release were more likely to be in the Escalation group compared to the Highonic group. Post-release Marriage was pos itive and significant for Group 1 and Group 2. s indicates that white parolees that were more likely to be married after release from the CYA e more likely to be in the Low Chronic and Escalation groups compared to the High-Chronic up. Post-release Employment was significan t and positive for groups 1 through 3. This cates that white parolees that were more likely to be employed after release from the CYA e more likely to be in the Low-Chronic, Es calation, or De-escalation groups compared to the h-Chronic group. -violent Crime Black Sample Figure 7.7 and Table 7.9 present the results of SGM and multinomial logistic regression non-violent crime trajectory groups for the black parolee sample. SG M identified a three ectory group model based on the lowest abso lute BIC. Group 1 was a low-chronic offending up that committed about one crime per year in each of the seven years after release. Group 2 a de-escalating group that initially committe d four crimes at Time 1 but by Time 7 they and alco Chr Thi wer gro indi wer Hig Non for traj gro was 134

PAGE 135

decre Low e sign of the coefficient for both Group 1 and Group ased to one crime. Group 3 was an escalati ng group that initially committed one crime at Time 1 but increased to over five crimes by Time 4. They remained at about four crimes through Time 7. Based on the offending patterns the th ree groups were labeled as follows: 1) Low Chronic, 2) De-escalat ion, and 3) Escalation. Table 7.9 presents the results of multinomial logistic regression for the non-violent crime black parolee sample. In the entire model, six va riables in this model emerged as significant. Among Juvenile Measures, only Psychological Status was significant but it was for both the chronic and De-escalation groups. The positiv 2 indicates that black paro lees that were diagnosed with a psychological problem were more likely to be in both the Low Chronic and De-escalation groups compared to the Escalation group. Figure 7.7. Non-violent crime black sample trajectory model 135

PAGE 136

Table 7.9. Non-violent crime black sample mu ltinomial logistic regression model GROUP 1 GROUP 2 Low Chronic Deescalation n=103 n=50 BExp(B)WaldBExp(B)W JUVENILE MEASURES IQ Score 0.0041.004 ald 0.0100.0021.0020.003 Grade Psych*8 Age at First Arrest 0.1261.1340.425-0.0100.9900.002 Family Intact -0.0180.9820.0010.6881.9890.554 13 Number of Siblings 0.0981.1030.3320.1001.1050.299 Mother Criminality -3.6480.0260.949-3.6810.0250.940 CYA MEASURES f Incarceration -0.0160.9841.140 -0.0410.9603.414*Escape Attempts -0.3110.7330.144-0.1900.8270.045 Job Training -0.0530.9490.0030.1441.1540.019 Educational Training -1.6430.1932.780*-2.7050.0676.089**POST RELEASE MEASURES Post Release Alcohol Use0.0821.0860.3150.0441.0450.069 Post Release Heroine Use-0.0340.9660.0310.0331.0330.021 Post Release Marriage -0.2490.7801.605-0.2410.7861.049 Post Release Employment-0.0120.9880.0010.1041.1090.057 Age at Release -0.3960.6730.8420.2111.2350.188Group 3 Escalation is the Reference Group (n=20) Statistical signidicance: * -0.2250.7980.709-0.1390.8700.207 ological Status 2.83917.1016.371**2.90318.2355.610*Juvenile Alcohol Use 1.3693.9311.8701.3643.9131.547 Juvenile Drug Use -1.0830.3391.128-0.2500.7790.04 FAMILY MEASURES Family Welfare 0.2391.2700.0340.1531.1650.0 Father Criminality -0.4490.6380.0750.2661.3040.018 Sibling Criminality -0.1690.8450.036 1.5264.6002.407*Length o p p < .10, ** < .05 One Family Measure was also significant. Si bling Criminality was positive and significant for Group 2, which indicates that bl ack parolees that were more li kely to have siblings that committed crime were more likely to be in the De-escalation group than the Escalation group. Three CYA Measures were also significant in this model. Length of Incarceration was significant and negative for the De -escalation group, which indicates that black parolees that ere incarcerated in the CYA for a shorter period were more likely to be in the De-escalation was also significant and negative w group compared to the Escalation group. Educationa l Training 136

PAGE 137

for Group 1 and Group 2. This indicates that blac k parolees that were less likely to receive educational training while in th e CYA were more likely to be in the Low Chronic and Deescalation groups compared to the Escalation group. Non-violent Crime Hispanic Sample Figure 7.8 and Table 7.10 illustrate the results of SGM and multinomial logistic regression for the non-violent offending Hispanic parolee sample. A three trajectory group model was identified based on the lowest absolute BIC. Group 1 was a low-rate chronic offending group that committed between one and two crimes per year for across all seve n years after release. Group 2 was a de-escalating group that initially committed about four crimes at Time 1 but decreased to about one crime by Time 7. Gr oup 3 was a high chronic escalating group that initially committed about three crimes at Time 1, increased to about seven crimes by Time 4, but decreased to about four crimes by Time 7. Based upon offending patterns in the model, the three groups were labeled as follows: 1) Low Chroni c, 2) De-escalation a nd 3) High Chronic. Table 7.10 illustrates the results of multinomia l logistic regression for the Hispanic nonviolent offending sample. Ten variables emerged as significant. Among J uvenile Measures, IQ was significant for Group 2. This indicates that Hisp anic parolees that had higher IQ scores were compared to the High-Chronic group. Juvenile Alcoh Amon more likely to be in the De-escalation group ol Use was also significant and positive but for Group 1. This indicates that Hispanic parolees that were more likely to use alcohol af ter release were more likely to be in the Low Chronic group compared to the High Chronic group. g Family Measures, two variables emerged as significant. Family Welfare was negative, which indicates that Hispanic parolees whose families were less likely to receive welfare were more likely to be in the De-escalation group th an the High Chronic group. Father Criminality was also significant but positive for Group 2. This i ndicates that Hispanic parolees that were less 137

PAGE 138

likely to have a father that was criminal were more likely to be in the De-escalation group compared to the High Chronic group. Figur ic and De-escalation groups, which indicates e 7.8. Non-violent crime Hispanic sample trajectory model Among CYA Measures, only Length of Incarcerati on significant. The negative sign of the coefficient for Group 2 indicates that Hispanic pa rolees that were incarcerated in the CYA for a shorter period were more likely to be in the De -escalation group compared to the High Chron group. Among Post-release Measures, four variables were si gnificant. Post re lease Heroin Use was significant and negative for both the Low Chronic that Hispanic parolees that were less likely to use heroin after release from the CYA were more likely to be in the Low Chronic or De-escal ation groups compared to the High Chronic group. Postrelease Marriage was significant and nega tive for group two, indicating that Hispanic 138

PAGE 139

parolees that were less likely to be married we re more likely to be in the De-escalation grou compared to the High Chronic group. P p ost-release Employment was significant and positive for Group 2. This indicates that Hispanic parolees th at were more likely to be employed were more likely to be in the De-escalation group compar ed to the High-Chronic group. Likewise Age at Release was also significant and positive for Group 2. This indicates that older Hispanic parolees that were released from the CYA when they were older parolees were more likely to be in the De-escalation group compared to the High Chronic group. Table 7.10. Non-violent crime Hispanic samp le multinomial logistic regression model GROUP 1 GROUP 2 Low Chronic De-escalation (n=50) (n=20) BExp(B)WaldBExp(B)Wald JUVENILE MEASURES IQ Score 0.1891.2081.279 0.3861.4722.924*Grade 0.5891.8020.171-1.9480.1421.420 Psychological Status 6.405604.7382.058 -4.8820.0080.768 Juvenile Alcohol Use 6.0030.8242.908*-1.3620.2560.144 Juvenile Drug Use -1.7990.1660.105-2.3990.0910.099 Age at First Arrest 1.1233.0741.238 0.0061.0060.000 FAMILY MEASURES Famil Family Welfare 3.9000.3990.995 -11.7070.0005.372**Father Criminality -1.8090.1640.114 11.3751.9794.008**Sibling Criminality 1.7345.6650.1806.7440.0952.182 EASURES Length of Incarceration 0.0351.0360.070 -0.3640.6952.882*Escap Job Tr* 81*Statistical signidicance: * y Intact -2.4140.0890.135-8.5470.0001.828 Number of Siblings 0.4841.6220.402 -3.4620.0316.398**Mother Criminality -3.6730.0250.127-7.7410.0000.498 CYA M e Attempts -3.4250.0331.272-2.4060.0900.319 aining -5.9570.0032.267 -7.3740.0013.610 Educational Training -3.4990.0300.851-4.5850.0101.468 POST RELEASE MEASURES Post Release Alcohol Use-0.6950.4991.522-0.4150.6600.497 Post Release Heroin Use -1.8860.1522.748*-1.8670.1552.693*Post Release Marriage 0.5521.7360.551 -2.0900.1243.359 Post Release Employment-0.0590.9420.0030.3651.4400.033 Age at Release -2.8280.0591.842 3.7361.9302.7Group 3 High Chronic is the Reference Group (n=17) p p < .10, ** < .05 139

PAGE 140

Violent Crime Figures 7.9 through 7.12 present the results of the SGM trajectories . Tables 7.11 through 7.14 present the results of binary logistic regression for violent offending. For all four samples, a two-trajectory group mo del was identified. Violent Crime Full Sample As illustrated in Figure 7.5, SGM identified a two trajectory group model based on the lowest absolute BIC. Group 1 was a desist ing group. Group 2 was a low-rate offending group. idered a Low Chronic group. The two groups we re labeled as follows: 1) Desi ster, 2) Low-Chronic. Table 7.11 summarizes the results of violent offending fo r the full sample. Seven variables emerged as significant. Race/Ethnicity was significant and pos itive, which indicates that black and Hispanic parolees were more likely to be in the Lo w-Chronic group than in the Desister group. Figure 7.9. Violent crime full sample trajectory model Because it remained steady at around one crime per year for all seven years, it was cons 140

PAGE 141

Race was p ositive and significant for the violen t crime full model. This indicates that Hispa an t among the Juvenile Measures. The negative s that parolees that were less likely to use drugs were more likely to be Among the CYA Measures, only Length of Incar ceration was significant. The positive sign of the coefficient indicates that the longer the parolee was incarcerated in the CYA the more likely they were to belong to the Low-Chroni c violent offending group than the Desisting group. Lastly, Alcohol Use and Heroin Use were si gnificant among the Post-release Measures. The positive sign for alcohol use suggests that the mo re likely the offender was to consume alcohol after release, the more likely they were to belong to th e Low-Chronic violent offending group compared to the High Chronic group. Heroin use was also significant. However, the sign of the coefficient for post-release Heroin Use was negative. This indicates that parol ees that were less likely to use heroin at any point after release were more likely to be long to the Low-Chroni c violent offending group ompared to the Desister group. nic and black parolees were more likely to be in the Lo w-Chronic offending group compared to the Desister group. Notably, this findi ng requires more scrutiny in the disaggregated models. Only Juvenile Drug Use was signific sign of the coefficient indicate in the Low-Chronic group co mpared to the Desister group. The Family Measures had two variables emerge as significant in the full sample. Family Intact was positive, which indicates that those parolees whose families were more likely to be intact when they were juvenile s were more likely to be in th e Low-Chronic group compared to the Desister group. Likewise, Sibling Criminalit y was also positive and significant, indicating that parolees that had criminal siblings were more likely to be in the Low-Chronic group than the Desister group. c 141

PAGE 142

Table 7.11. Violent crime full samp le logistic regression model BExp(B)Wald RACE/ETHNICITY 0.5061.6596.949**JUVENILE MEASURES IQ Score -0.0090.9910.516 Grade 0.1271.1352.485 Psychological Status -0.4460.6402.066 Juvenile Alcohol Use 0.2981.3470.819 Juvenile Drug Use -0.0840.9203.706*Age at First Arrest -0.0590 FAMILY MEASURES .9431.056 Fami Fami ning -0.2320.7930.670 E MEASURES Post Rel Post ly Intact 0.4411.5542.643*ly Welfare -0.2880.7500.685 Number of Siblings -0.0610.9411.295 Father Criminality 0.2131.2370.337 Mother Criminality 0.1531.1660.089 Sibling Criminality 0.4771.6122.898*CYA MEASURES Length of Incarceration 0.0171.0178.981**Escape Attempts -0.1780.8370.453 Job Training -0.2610.7700.853 Educational Trai POST RELEAS ease Alcohol Use 0.1101.1175.169**Release Heroin Use -0.1130.8933.855**Post Release Marriage -0.0150.9850.052 Post Release Employment -0.0570.9440.298 Age at Release -0.0570.9450.199Statistical signidicance: * p < .10, ** p < .05 Violent Crime White Sample Figure 7.10 shows the violent crime trajectories and Table 7.12 illustrates the results of binary logistic regression for th e sample of white parolees. Model selection for the white parole sample, however, was not based on the e lowest abso lute BIC because there would not have been any v BIC o and a between .8 a nd 1.1 crimes per year across all seven years. ariation in the dependent variable with a one trajectory group model. Thus, because the for the two-group model was only marginally high er, the two-group model was selected. The tw groups identified by SGM were a desisting group th at averaged less than .2 crimes per year low-chronic group that fluctuated 142

PAGE 143

143 Figure 7.10. Violent crime white sample trajectory model Based on the patterns of offending identified by SG M, the two groups were labeled as followed: 1) Desister and 2) Low-Chronic. Table 7.12 presents the results of binary lo gistic regression of white parolee violent offending. Three variables emerged as significan t. Among the Juvenile Measures, Psychological Status was the only significant variable. The coefficient was positive, which indicates that white ith a psychological disorder were more likely to belong to the Low-Chronic group co Among the Juvenile Measures, only Mother Cr iminality as significant. The sign on the coefficient was positive, which indicates that white parolees whose mothers were more likely be a criminal when they were juveniles were more likely to belong to the Low-Chronic group compared to the Desister group. Of the rema ining variables, only Escape Attempts was parolees that were more likely to be diagnosed w mpared to the Desister group. significant. The negative sign of th e coefficient suggests that paro lees that were less likely to

PAGE 144

attempt an escape from the CYA were more likel y to be in the Low-Chronic violent offending group. Table 7.12. Violent crime white samp le logistic regression model BExp(B)Wald JUVENILE MEASURES IQ Score -0.0150.9850.332 Grade 0.0931.0970.458 Psychological Status 1.9777.2239.690**Juvenile Alcohol Use -0.1290.8790.042 Juvenile Drug Use 0.7322.0800.853 Age at First Arrest -0.0160.9840.019 FAMILY MEASURES Family Intact -0.0540.9480.008 Family Welfare 0.3041.3560.158 Number of Siblings 0.0171.0170.019 Father Criminality -1.1710.3102.042 Mother Criminality 2.90618.2846.308** Sibling Criminality -0.0800.923 0.020 CYA MEASURES Escape Attempts -1.1820.3074.065**0.2111.2360.151 Educational Training -0.3320.7170.386 POST R Post Release Alcohol Use 0.0421.0430.173 Post Release Marriage -0.0610.9400.279 Age at Release 0.2911.3381.392 Length of Incarceration 0.0211.0212.354 Job Training ELEASE MEASURES Post Release Heroin Use -0.0490.9520.146 Post Release Employment 0.0811.0850.230Statistical signidicance: * p < .10, ** p < .05 Violent Crime Black Sample Figure 7.11 illustrates the results of SGM for th e black parolee violent crime sample and Table 7.13 presents the results of binary logistic regression. Two trajectories were identified by SGM. Group 1 was a desisting group that averaged less than .2 crimes per year for all seven years. Group 2 was a low-rate chronic group that averaged be tween .78 and 1.0 crime pe for all seven years. They were labeled as follows: 1) Desister and 2) Low Chronic. r year 144

PAGE 145

Figure 7.11. Violent crime black sample trajectory model Table 7.13 illustrates the binary logistic regression result s for the violent crime black sample. Three variables were significant in this model. Among Juvenile Measures, only Grade was significant. The positive sign of the coefficien t indicates that black parolees who had higher reading and math scores were more likely to be in the Low Chronic group compared to the Desister group. only significan t variable among the Family Measures. The positive lack Father Criminality was the sign of the coefficient indicates that black parolees that had a father that was criminal when they were juveniles were more likely to belong to the Low-Chronic group compared to the Desister group. Although no CYA Measures were significant, one Post-release Measure was. Post-release Employment was significant and the coefficien t was negative, which indicates that b 145

PAGE 146

parolees that were less likely to be gainfully em ployed after release were more likely to be in the Low-Chronic group compared to the Desister group. Table 7.13. Violent crime black samp le logistic regression model BExp(B)Wald JUVENILE MEASURES IQ Score -0.0100.9900.220 Grade 0.4141.5133.393**Psychological Status 0.8572.3572.097 Juvenile Alcohol Use -0.1080.8970.758 Juvenile Drug Use -0.8960.4081.927 Age at First Arrest 0.7842.1911.243 FAMILY MEASURES Family Intact 0.3501.4190.445 Family Welfare 0.2821.3260.197 Number of Siblings 0.0131.0130.018 Father Criminality 1.6335.1202.619*Mother Criminality 0.4921.6350.347 Sibling Criminality -0.7460.4741.721 CYA MEASURES Length of Incarceration 0.001 Escape Attempts 0.350 1.0010.016 1.4190.463 Job T Educae: * raining -0.8160.4422.196 tional Training -0.2320.7930.151 POST RELEASE MEASURES Post Release Alcohol Use 0.1301.1381.985 Post Release Heroin Use 0.0541.0550.185 Post Release Marriage 0.0791.0830.328 Post Release Employment -0.4310.6502.843*Age at Release 0.0331.0340.016Statistical signidicanc p p < .10, ** < .05 Violent Crime Hispanic Sample The final violent crime model examines trajectory group membership for the Hispanic parolee sample. Figure 7.12 illustrates the tr ajectory groups identified by SGM. Table 7.14 shows the results of binary logistic regression for trajectory group me mbership. Based on the lowest absolute BIC, a two trajectory group mode l was selected. The trajectory patterns reveal that group one averaged between .2 and .4 crimes between Times 1 and 2, decreasing steadily until nearly reaching zero by Time 7. This appears to be a desisting group. On the other hand, at 146

PAGE 147

Time 1, Group 2 initial offending was nearly .45, a nd then it decreased slightly to .2 crimes at Time 4 than increased to 1.7 crimes by Time 7. Based on these patterns the two groups were As illustrated in Table 7.10, six variable s emerged as significant. Among Family Measures, three variables were significant. Family Intact was positive and significant, which indicates that parolees whose families were more likely to be intact when they were juveniles were more likely to be in the Escalation group compared to the Desister group. Family Welfare was also significant. The positive sign of the coefficient suggests that Hispanic parolees whose family received welfare when they were juvenile s were more likely to be in the Escalation group compared to the Desister group. Number of Siblin gs was also significant but the coefficient was negative, which indicates that Hispanic parolees w ith fewer siblings were more likely to be in the Escalation group than the Desister group. therefore labeled as follows: 1) Desister and 2) Escalation. Figure 7.12. Violent crime Hispanic sample trajectory model 147

PAGE 148

Table 7.14. Violent crime Hispanic sa mple logistic regression model BExp(B)Wald JUVENILE MEASURES IQ Score 0.0761. Grade -0.6250 0791.053 .5351.972 Psycho Juven Juvenile Drug Use -0.5600.5710.136 FAMILY MEASURES Family Welfare 3.57035.5183.086*Father Criminality 2.0067.4321.521 Sibling Criminality 1.0842.9580.365 Length of Incarceration 0.0141.0140.080 Job Training -2.8860.0562.668*0.3541.4241.561 -0.1780.8370.301 st Release Marriage 0.8722.3913.424*st Release Employment -0.2450.7830.231 ge at Release -0.0650.9370.013tatistical signidicance: * logical Status -1.5820.2061.057 ile Alcohol Use -1.2660.2820.691 Age at First Arrest -0.0120.9880.002 Family Intact 2.58213.2292.774*Number of Siblings -0.5530.5752.679*Mother Criminality 0.9292.5320.167 CYA MEASURES Escape Attempts -1.5020.2231.339 Educational Training 4.30073.6854.541**POST RELEASE MEASURES Post Release Alcohol Use Post Release Heroin Use Po Po A p p S < .10, ** < .05 Among CYA Measures, two variables were si gnificant. Job Training had a negative oefficient, which indicates that Hispanic parole es that were less likely to receive job training hile incarcerated in the CYA were more lik ely to belong to the Escalation group. Conversely, e coefficient for Educational Training was positiv e, indicating that parolees that were more kely to receive educational training while inca rcerated in the CYA were more likely to belong the Escalation group than the Desister group. Only Marriage was significant among Post-re lease Measures. The positive sign of the oefficient indicates that the more likely Hispanic parolees were to be married the more likely c w th li to c they were to be in the Escalation gr oup compared to the Desister group. 148

PAGE 149

CHAPTER 8 DISCUSSION AND CONCLUSION Discussion The objective of this study was to examine offending tr ajectories among a group of young adults who were previously incarcerated for committing serious violent and non-violent crimes as juveniles and tracked for se ven years after release from th e California Youth Authority. The data were adequate for investigating key correlates of viol ent and non-violent offending while also comprising an adequate demographic breakd own along racial and ethn ic lines, to include black, white, and Hispanic parolees. This made it possible to investig ate subgroup racial/ethnic comparisons on all the independent measures acr oss the non-violent, violent, and total crime types. Toward this end, the results produced some notable findings that bear import for current theoretical debates in life-course criminology. First, a broad consideration of the trajectory patterns across all twelve models revealed that—consistent with prior research—this sample of offenders were more likely to commit nonviolent crimes after rel ease rather than violent crime. Not one trajectory group in any of the four non-v e full model shows th at nearly 100% of the sample f ring the follow-up iolent models averaged less than one crime during the seven year follow-up period. Conversely, only the highest-rate violent o ffending groups in the full and black models consistently averaged slightly less than one violent crime over the se ven years. These groups comprised only 27% and 37% of all offenders in each model, respectively. Trajectory group modeling also revealed th at overwhelmingly this sample comprised persistent offenders. Examination of the total crim committed at least one crime each year for the seven years after release with only 8% o the sample averaging less than one crime per y ear. Only one group in th is model, only the deescalation group (29% of the sample) steadily de-escalated in offending du 149

PAGE 150

period. At Time 1 and Time 2, the de-esca as committing over four crimes per year, but by Time 7 they decreased to one crime. On the other hand, Group 2 and Group 4, which comprised about over 44% of offenders actually individuals in the sample committed at least fourteen crimes or more. offending models, not one trajectory group averaged less than one crime per year for any of the seven yea The non-violent offending models also showed de-escalation in o ffending that steadily in each model) started out offending at a very high rate at Time 1, between three and four crimes, but by Tim non-violent offending for over 60% of the full and white samples and over 40% of the black and lation group w escalated in offending from Time 1 to Time 7, while Group 1 persisted steadily at about one crime per year. I ndeed, only two offenders in the entire sample remained crime free the entire seven-year post follow-up while over half of the Examination of the non-violent mo dels elicited some important findings as well. First, for all four non-violent offending models, there were at least three heteroge neous trajectory groups. This would suggest that there are multiple type s of offenders generally and within race and ethnicity. Most notable about the non-violent o ffending patterns is that, similar to the total rs after release. For the full and white samples, over 53% and 60% of the samples, respectively, were committing ove r two non-violent crimes per year. For the black and Hispanic models, there were only three trajectory groups , though they too all committed over one crime per year. headed toward desistance. Each of the four di stinct non-violent de-es calation groups (one group e 7, they were down to one or less crimes. Alternately, there were also multiple highrate persistent and escalating non-violent offendi ng groups that illustrate the high frequency of Hispanic samples. 150

PAGE 151

Race/Ethnic Trajectory Differences The race/ethnic-specific group models showed somewhat similar patterned offending for all crime types but with important differences in violent offending rates. For the violent offending models, there were clearl y two different trajectories in each racial/ethnic samples and for the overall model. However, the black samp le contributed most to the overall model. Whereas both trajectory groups in the white model averaged less than one crime per year (less than . c year) but by Time 7 th ey were all down to about one crime per years. The white nonviolen the non-violent full sample model only but more importantly, the direction went against what the literature would predict. That is, all else equal, those that were arrested 6 crimes per year for four of the years), the persistent offending group in the black sample (37% of all black parolees) aver aged over one violent crime per year. The Hispanic violent crime model showed escalating offending for a sm all group of offenders; however, it was not significant in the overall model. For non-violent crime, the white parolee sample had four distinct offending trajectories, while the black and Hispanic models had three each. However, in each of the racial and ethni samples, the low-chronic persistent offending group offended at about one crime per year and they each had a de-escalation group that started offending at high rates in Time 1 (over three crimes per t offending model also had a high-chronic trajectory group who alternated between four and six crimes per year for the entire seven years. Risk Factor Effects The developmental criminology literature places an emphasis on a relatively small cadre of causal mechanisms. Most prominent among them is age of onset or ag e of initiation into offending. In this study, as in others, age of onset was measured by the paro lee’s first arrest. It was significant for 151

PAGE 152

later were more likely to persist in offe ndi ng—in this case—escalate in non-violent offending after i ncarceration than those th at were arrested at younger ages. Table 8.1. Summary of trajectory groups for each models and signi ficant risk factors Number of Trajectories Total Crime Non-violent Crime Violent Crime Full Sample 4 4 2 Black Sample 4 3 2 Significant Risk Factors Total Crime Non-violent Crime Violent Crime Full SampleJ.Alcohol Use, Number of SiblingsAttempts, Job Training, Heroin Use,Intact, Sibling Criminali Escape Attempts, Heroin Use, Employment of Incarceration, Heroin Marriage White Sample 4 4 2 Hispanic Sample 3 3 2White, Black, IQ, J.Drug Use, IQ, Age at First Arrest, Escape Race/Ethnicity, J.Drug Use, Family ty, Length Use, Alcohol Use Wh i te S amp l eIQ, Age at First Arrest, Father IQ, J.Drug Use, Sibling Criminality,Psychological Status, Mother Heroin Use, Marriage Marriage, Employment Criminality, Sibling Criminality, Escape Attempts, Educational Criminality, Escape Attempts Job Training, Age at Release, Training, Job Training, Alcohol Use Bl ac k S amp l eFamily Intact, Family Welfare, Psychologica l Status, Sibling Grade, Father Criminality, Attempts, Education Training, Educational Training Heroin Use, Marriage, Employment Age at Release Length of Incarceration, Escape Criminality, Length of IncarcerationEmployment Hi span i c S amp l eJ.Drug Use, Escape Attempts, IQ, J.Alcohol Use, Family WelfareFamily Intact, Family Welfare, Heroin Use Number of Siblings, Father Number of Siblings, Job Training, Criminality, Length of IncarcartionEducational Training, Marriage Heroin Use, Age at Release The lack of significant correlation in other offending models is at first impression a bit surprising given the importance that has been pl aced upon it in earlier st udies. One explanation for the weak correlation across the m odels is that age of arrest was not tapping into differences in unexplained childhood risk factors that might lead to antisocial or aggressive behavior for some offenders or to differences in age of onset in th e first place. However, the best explanation seems to be that in this sample the parolees were ove rwhelmingly similar regarding early onset, at least as measured by age at first arrest. Over 80% had been arrested prior to age 15, which is the low end of what is generally considered the p eak age range in offending (see Farrington, 1994; 152

PAGE 153

Steffensmeier and Allan, 2000). It appears that other factors such as IQ and psychological diagnosis played a more important role than the early onset of del inquency. Subsequent analysis Race/ethnicity was significant in all three fu ll sample models. It predicted persistent violent offending for white and black parolees but not for Hispanics. Compared to whites, blacks were much more likely to persist steadily in vi olent offending at over one crime per year for all seven years, while whites did not average even one violent crime per year and were at .20 by Time 7. Among the rest of the juvenile measures, IQ score and psychologica l diagnosis were both significant in five of the twelve models, whic h reflect some of the strongest effects on postrelease offending. For example, IQ predicted persistent offending for the non-violent crime full, white and Hispanic samples, and for the total crime full and white samples. The link between IQ and offending was negative in that parolees who had higher IQs tended to display lower persistent offending or de-escala ting trends relative to the high er rate or escalating trends. On the other hand, psychological diagnosis c onsistently predicted persistent non violent offending for black and Hispanic parolees and deescalating violent offending for the full and white parolee models. Taken together the violen t black model and total crime white model, it appears that there is some s upport for underlying psychological fact ors that link to persistent offending. I would add here that this does not de finitively suggest neur opsychological deficits per Moffitt’s theory, nor immutable differences in IQ as Hernnstein and Murray claim. Rather, my interpretation of the results reflects what I see as the overall pattern among these three variables, which is supported by the c onsistent, though not overwhelming findings. of variance revealed only significant effects at Time 4. 153

PAGE 154

Turning to the family measures, the results were somewhat consistent although fewer example, number of siblings correlated to non-viol ent full, total full and violent Hispanic models. Having more siblings predicted overall non-violent and . For total offending, but having less siblings for H ld ted ly father’s and sibling’s criminality. The best that could be inte rpreted here is that it was g at the olent offen g ed. his ns on raining increased the likelihood of pers istent non-violent offending and escalation of violent offending ispanics linked to de-esc alation in non-violent offending. However, de-escalation shou not be confused with termination because the traj ectory group had a high initial rate of offending at Time 1 and though it steadily declined by Time 7, the parolees were still being arrested of one non-violent crime per year. Finally, family criminalit y also played a small, but relatively isola role; main enerally steady in predicting persistent or deescalating behavior, but not by racial/ethnic group or crime type. Next, a further examination of the incarcer ation measures revealed some noteworthy findings. Length of incarceration, esca pe attempts, and training (bot h job and educational) played significant roles in predicting persistent and es calating offending. The findings suggest th longer the parolee was incarcerated, the more persiste nt likely they were to persist in non-vi ding. Likewise, the more escapes the parole es attempted linked to persistent non-violent offending and total offending; of which non-violen t offending shapes part of the total offendin trajectories. Job training also re lated to non-violent offending in th e way that would be predict Those who did not receive the training correlated to overall persistent no nviolent offending. T held for both the white and Hispanic samples. Not receiving training also predicted an escala tion in violent crime in the black parolee model. Although educational traini ng was significant across four models, the conflicting sig the coefficients make it difficult to interpret, especially when receiving educational t 154

PAGE 155

for H gs might seem to point to the relationship between property crimes and drug y up. ispanics. As a result of limitations on the data , the effects of the type or extent of job or educational training that the parolees received wa s not measurable. In this regard, we are left with more distal measures about the significance of training on future offending. It could very well be that they really did not receive the type of portable training that can be employed in the marketplace or that there were other factors in the parolee’s lives that undermined such training. For the post-release measures, the results hi ghlight the significance of heroin use in offending. Parolees who were less likel y to use heroin were more like ly to be in a low chronic or low persistent non-violent and to tal crime offending trajectories . It did not correlate to any violent crime models but was significant for al l total crime models in predicting persistent offending. Such findin use where one commits crime to support a habit. The marriage effect was significant for four mo dels, but went against what the literature would predict (Sampson and Laub, 2004, 1993). Genera lly speaking, those who were more likel to be married across the seven time periods we re more likely to actually escalate in offending. This was true for white parolees’ non-violent offending, and for all the total offending models. Finally, age at release was a fa irly strong predictor of non-violent and total offending. Overall, those who were younger when they were released from the CYA were more likely to belong to a persisting or escalating non-violent offending group while those who were older were more likely to belong to a de-escalating offending gro Lastly, I turn now toward investigating in a bit further detail the findings within the different racial/ethnic-specific samples. Three mode ls stand out. First, the significant predictors in the white parolee non-violen t crime model largely predicte d persistent and escalating offending. For instance, the significant incarcer ation measures correla ted with low chronic 155

PAGE 156

offending for whites and the significant post-releas e measures more or less predicted escalatin offending. On the other hand further examination of the violent Hispanic models shows that th significant predictors—in this case family m easures and incarceration measures—predicted escalation in offending for Hispanic parolees. All of the independe nt measures, except marriage, predicted in the direction that the literature would suggest. Fina lly, the total crim e black parolee g e mode ic th e difference between “photography” and “ h ere are only two types of offenders, lifecours at this ll independent measures used to approximate Moffitt’s theory, did not perform exceptionally well. l revealed a correla tion between the post-rel ease and family predictors and low chron offending. They too went in the directi on that the literatur e would predict. Conclusion In a recent response to Sampson and Laub (2005) in Criminology , Nagin (2005: 875) asserted the difficulty in explaining behavioral change s over time using group-based developmental trajectories; that the difference is akin to cinema.” This is, to be sure, only one of numerous issues revolving around developmental studies of crime and the methodological tools used to investigate behavior over time. Setting aside for a moment some other problems with trajectory group modeling, which Sampson and Laub (2005: 905) termed misconceptions (that indi viduals actually belong to the trajectory groups, that the number of trajectory groups is immutable, and that “trajectories of group members follow the group-level trajectory in lo ck-step) these findings revealed only mixed support for any of the theories employed in this study. For example, Moffitt (1994, 1993) posits that t e persistent and ad olescent-limited. Following her premise it is clear from this data th might not be the case, at least with respect to non-violent and overall offending. In both the nonviolent and total crime trajectory models, there were multiple distinct trajectories across a models, which convincingly showed different levels of offending. Moreover, the rough 156

PAGE 157

That is, although together, Psychological status, IQ, Grade, and Age at First Arrest predicted membership in persistent offending groups, they were mainly in the non-violent and total models, and they mostly predicted low-level chronic behavior, not high-chronic behavior. On the other hand, the violent crime trajectori es clearly showed two groups of offenders for all models; one group that persisted in o ffending for the entire seven years and another group that showed very little or intermittent offending. The overa ll violent crime model was significantly influenced by black violent offending, and to a much lesser ex tent, white violent ack parolees were more likely to show persistent behavior than whites. Yet, only one in lent s. Again, uld lute violent oes not necessarily mean support for self-c se gly ory offending. Bl dicator, psychological status, used to pred ict Moffitt’s theory was significant in the vio crime model, which was for the persistent o ffending trajectory group for black parolee the lack of empirical support fo r the risk factors could be the result of them not tapping into actual neuropsychological deficits or the transacti onal nature of family problems in dealing with children with neuropsychological defi cits that makes having such defi cits pathological or it co be that there are other unmeasur ed factors that are important to offending. There is support for Moffitt in terms of groupings. The best violent crim e models (i.e. those with the lowest abso BIC) were two-group models. There is a strong indi cation that there are diffe rent types of offenders. Yet, mixed support for Moffitt’s dual taxonomy d ontrol or age-graded theories. If anything can be extrapolated from the findings is that the theories did not fair too well eith er, at least with respect to th e independent predictors employed in this study. Marriage, for instance, while signif icant in five of the models, did not predict any real change in offending, such as desistance or de-escalati on. Employment was overwhelmin insignificant and in the one model it was signifi cant, the sign predicted opposite what the 157

PAGE 158

would predict. In addition, Gottfredson and Hirs chi’s contention that there are no groups is undermined somewhat, at least with respect that th ese were serious juvenile delinquents th been incarcerated, all of them still offended, and there were still differences in their patterns at had , if not et ak age in in offending, particul arly for violent crime. That marriage, for example, decreased violen or iologies. Nevertheless, this was not neces sarily a study of self-c ontrol and others might examine this same data and come to different conclusions. With respect to the age/crime relationship, given that all of them were past the pe offending while showing few signs of slowing dow n, it seems less likely that it is simply lowself control or age-specific di fferences in propensity that ge nerated their offending behavior. These offenders were all in their twenties, most in their late-twenties and thirties, thus they were past aggregate peak age in offe nding. Moreover, the trajectories only revealed a small percentage that seemed to be heading toward desistance and only two that actually desisted, thus also making it less likely that, at least with this sample , declines in crime is a matter of changes in frequency. These offenders showed pretty solid pe rsistent offending into their late twenties and while it is unknowable from this data what they did after the seven year s of tracking, there is a strong possibility given their prior histories that most continued to offend after Time 7. Findings from this study point to the need for further investigation of race and ethnic differences t offending for Hispanics bu t not for blacks or whites or th at drug and alcohol use were significant in violent offending are important findings. With the latter, it is unclear why this would be. Perhaps cultural expectations placed on Hispanic/Latino families have a more pronounced effect when indivi duals marry. Conversely, the ravages of inner-city black communities that have rendered many black men unsuitable for marriage would be less likely to have an effect even if they do marry. Likewise , drug and alcohol use was an important predict 158

PAGE 159

in the full model. This should come as no surp rise given the well-docum ented link between and alcohol use and offending. As discussed in Chapter 6, drug and alcohol use links to other criminal behavior in numerous ways, that is, both direct ly and indirectly. Inasmuch as these parolees were as likely be using some sort of intoxicant further substa ntiates the cor drug to relation. However, what cannot be inferr ering specif ut SGM egree to ed from the data is whether drug or alc ohol use actually caused vi olent offending or their subsequent arrests after release from the CYA. Non-violent trajectory patterns also revealed similar offendi ng behaviors such that there were at least three groups of offenders across race/ethnici ty. Although there was not much consistency with respect to how the risk f actors operated, the findings did show strong similarities across the dependent variables in the models. On the other hand, while SGM identified two trajectory groups across race and et hnicity, it also revealed that black parolee violent behavior was largely the driving force such that further examination of these differences is necessary. However, mine is not the last word. Further analysis that takes into account residential, familial and other exogenous fact ors that can go a long way in further uncov ic variation in offending that differe ntiates racial differences in arrests. To be sure, there are a number of limitations in this study. Recall that Sampson and Laub (2005) recently identified three misconceptions about trajectory group modeling. In carrying o this study, these misconceptions were certainly a concern. Take the first one, that individuals actually belong to the trajectory group they are assigned (Sampson and Laub, 2005: 907). is based on probabilities, which is not a problem in quantitative re search. However, the d which individual trajectories varied from group trajectories is unknown. With any recent methodological developments, more empirical studies need to be conducted that seek to better 159

PAGE 160

understand how individual trajectories comport with group trajectories. Second, group selection has been called into question. As with this study, other trajectory studi es have employed Bayes Inform his as a st, the data comprised all males. The fact that femal f r to robbery ender r. . nters in ation Criteria to optimize group selection (C hung et al., 2001). Yet, some still see t problem because relative differences in BIC can change the outcomes in the model, thus group numbers are not “immutable” (Sampson and Laub, 2005: 907). Nevertheless, this study adhered relatively closely to the BIC, except where pars imony made logical sense to collapse groups. There were also limitations in the data. Fir es have been increasingly represented in official arrest statistics poin ts to the importance o investigating their trajectory patterns. This take s on even more importance since they appea be committing crimes that have been traditionally thought of as male dominated such as and assault. Large-scale structur al changes that have occurred during a time of increased g equality have led to more convergence of ma le/female behaviors. Gender roles have been changing and more women are now in the workforce and out of their homes than at any point in history. It would follow that we would also witness changes in gendered criminal behavio Unfortunately, this study was not suited to inve stigate hypotheses about female persistence and desistance. This study was also limited by the fact that information on offending was somewhat dated That is, these offenders were released in the 197 0s and 1980s, thus there might be period-specific effects that could have potentially affected the outcomes. To fully realize the vitality of recent life-course and developmental theories, data th at cover an extended period of time that goes deeper into adulthood is needed. Fi ndings such as these should thus be historically placed. This was a period of notably increasing crime rates in the U.S. in general and in many urban ce California. For example, total index crimes increased by over 600,000 crimes between 1970 and 160

PAGE 161

1980 (violent and property crime rates increased as just as dramatically) (Bureau of Justice Statistics, 2006). Racial and economic inequality, gang activity and new drug markets such as the rise of the turf wars over cr ack cocaine lines also flourished. Arguably, these findings could t hus be understood better as a reflection of the problems throughout California, particularly regarding racial and ethnic di fferences. That black parolee violent crime drove the full samp le violent crime trajectory mode l should not be surprising. As noted e was hese hapter 3 for a more detailed discu ssion). While change in arrest patterns has affect to in Chapter 3, the crime rates for blacks and Hispanics have been consistently higher than whites, especially for serious crimes such as r obbery and homicide. In ra cially and ethnically segregated places such as Los Angeles, this has been linked to notable racial differences in offending and arrests and therefore, one would exp ect that trajectories fo r blacks and to some degree, Hispanics would be more likely to reflect the overall racial and ethnic differences in offending. Yet, this was not necessarily the cas e for the non-violent crime model where rac insignificant for in the full sample. Another way to place these findings in context would be to look at arrest rates, which t models are based. The “get tough” era that bega n in the 1970s marked a dramatic increase in arrests (see C ed whites, blacks, and Hispanics, they have had an especially pernicious effect on blacks and Hispanics (Petersilia, 2005; Tonry, 1995). Ma ndatory three-strikes laws and other such punitive legislation that has targeted serious repe at violent, property, and drug offenders seems have played a partial, if not, substantial role in racial variations in arrest. In addition, if this sample had been drawn from a contemporary population of offenders it is plausible that differences in offending traj ectories would be even starker. There would probably be many more blacks and Hispanics in the sample than whites and their arrest rates 161

PAGE 162

would be higher thus showing mo re persistent offending for both violent and non-violent cri If anything, the data for this study might undere stimate offending even though overall crime rates have decreased over the past thirteen years. The structural ch mes. anges discussed in Chapter 3 sugge ect ise s. o at d ore alcohol. Most likely, it is th o st that the black and Hispan ic population in the inner city has been left worse off even while other blacks and Hisp anics have benefited. The independent measures in this data were also only rough indicators of the theories that framed this study and as the analysis revealed, they did not particularly fit the theories all that well. I recognize that there are i nherent problems in using only pr oxy measures rather than dir measures. For example, an ideal investigation of Moffitt’s theory would be to use more prec indicators of verbal and executive deficits in indi viduals as they interact with familial processe This study was limited by the use of psychometric measures of IQ and pe rsonality disorders t infer about potential neuropsychological deficits and as such, linkages between them are still rather unclear in the literature. One potential issue in that has been raised in these findings concerns juvenile alcohol use compared to juvenile drug use. Only about 36% of the sample had documented alcohol use, which was far less than the drug use of 62%. As ide from being curious, these numbers are odds with the large body of literature on drug and alcohol use and crime. Pure speculation woul lead one to possibly conclude th at there is a stronger link between drug use and crime and m serious juvenile offending. Also, it could be that there were reside ntial issues that lead to many of the parolees to choose perhaps more easily ac cessible drugs compared to at the juvenile drug use was a compos ite measure of numerous types of drugs, thus allowing for a higher probability of offenders having used drugs (of any kind) compared t alcohol. In any case, from this da ta a strong conclusion is unknowable. 162

PAGE 163

Furthermore, family discord and other fam ily measures are important components of Moffitt, Gottfredson and Hirschi, and Sampson and Laub’s theories. For example, family criminality links in important ways to individual offending (Sampson and Laub, 1993). As such, both Sampson and Laub, and Moffitt recognize the im portance of family criminality as being potential indicator of family di scord. There were fe a w measures that link to low self-control beyon ew substantially from Moffitt’s developmental taxonomy, the fact th ive ch P. nd developmental theories d the notion that all criminality is an indica tor of low self-control, a point that Aker’s criticized as being tautological. Th ere were also numerous indicator s such as the family variables that each of the three theoretical positions encompass. Thus, teasi ng out which position was supported by such variables seems to indicate that a broad conclusion supports Gottfredson and Hirschi’s contention that there is no systematic li nkage of such variables to offending. That it is about ones ability to control their behavior in the presence of multiple and changing opportunities. Lastly, since this study dr at all but two offenders continued offending at relatively high rates after re lease from the CYA indicates that they were all potential life-c ourse persistent offenders. Despite inconclus evidence of Moffitt’s predictions about what cau ses LCP vs. AL offending, it would be a stret of logic to conclude that any of the parolees except the two co mplete desisters were LCP. Put another way, this study did not have a compar ison group of AL offenders in which to draw stronger conclusions. Moffitt’s AL typology is ce rtainly important in understanding criminal careers, particularly in making distinctions—should they exist—between different types of offenders, however it could be that Moffitt is wrong and that all offenders are potential LC This investigation, however, was not only abou t such indicators sinc e the theoretical a methodological disputes currently underway concerning life-course and 163

PAGE 164

are al , Nagin e d also found iolent offending, but it was only significant for th e black and white samples, and even r the ain unans so over the existence (and number) of group differences based on patterns of offending change, in addition to risk factors associated with them. As Sampson and Laub (2005) and (2005) debated, arguably the SGM methodology employed here might be well suited for testing Moffitt’s developmental taxonomy. Whatever the merits of this methodology relative to developmental theory, it is obvious that this study is not going to settle this issue. Much mor empirical work needs to occur before SGM becomes a well-entrenched, more widely accepte (and understood) methodological tool similar to th at of OLS regression or other conventional techniques. Toward this end, this study set out to exam ine persistence and desi stance under a criminal career and developmental/life-course framework. The findings certainly revealed persistent offending, particularly for non-violent and total crim e for nearly the entire sample. It persistence in v further, the black sample, which is arguabl y the strongest racial/ethnic difference in the entire study. At the same time, a substantial pe rcentage of the sample exhibited less violent behavior patterns. This could be interpreted as proceeding toward desistance; that is, assuming they were violent before hand. Conversely, non-vio lent offending showed substantial persistence and a small, but notable group of de -escalators that seemed to be in the process of desisting. As mentioned above, there were only two offenders in the entire sample who did not offend afte release. In sum, this study was not meant to convinc ingly answer the many issues surrounding recent developments in life-course and developmental theory nor the many questions that rem wered regarding the age/crime relationship. Future research will want to head in that direction by examining how exogenous factors in fluence persistence criminal behavior and, 164

PAGE 165

likewise, how factors influence desistance or termination from offending. In keeping a narrow focus, the goal was to better understand the processe s of persistent offending and desistance from offending and how extant theory predicted such processes. Notwithstanding the limitations, the findings produced only mixed results. Whether this was an effect of usi ng only rough the indicators, unsettled problems with SGM, or probl ems with extant theory is a matter to be take up in future endeavors. oretical n 165

PAGE 166

LIST OF REFERENCES Agnew, Robert. 1999. Strain and subcultura l theories. In Joseph F. Sheley (ed.), Criminology: A Criminological Theories . Los Angeles: Roxbury. Aguilar, Benjamin, L., Alan Sroufe, Byron Eg eland and Elizabeth Carlson. 2000. Distinguishing the early-onset/persistent and adolescence-onset antisocial behavior types: From birth to 16 years. Develo Contemporary Handbook . Belmont, CA: Wadsworth. Agnew, Robert. 2000. An overview of general strain theory. In Raymond Paternoster (ed.), pment and Psychopathology 12:109-132. Akers, Ronald L. 1977. Deviant Behavior: A Social Learning Approach . Belmont, CA: Wadsworth. Akers, Ronald L. 2000. Criminological Theories: Introdu ction, Evaluation, and Application. Belmont, CA: Wadsworth. Avi-Itzhak, Benjamin, and Reul Shinnar. 1973. Quantitative models in crime control. Journal of Criminal Justice 1:185-217. Barnett, Arnold, Alfred Blumstein and David P. Farrington. 1989. A pros pective test of a criminal career model. Criminology 27:373. Bartusch, Dawn Jeglum, Donald R. Lynam, Te rrie E. Moffitt and Phil A. Silva. 1997. Is age important? Testing a general versus a deve lopmental theory of antisocial behavior. Criminology 35:13-48. Beck, Allan J., and Paige M. Harrison. 2002. Prisoners in 2000. Washington, DC: Bureau of Justice Statistics. Benda, Brent B., Nancy J. Toombs and Mark Peacock. 2003. Discriminators of types of recidivism among boot camp graduates in a five-year follow-up study. Journal of Criminal Justice 31:539-551. Birbeck, Christopher, and Gary LaFree. 1993. The s ituational analysis of crime and deviance. Annual Review of Sociology 19:113-137. Blau, Judith R., and Peter M. Blau. 1982. The co st of inequality: Metr opolitan structure and violent crime. American Sociological Review 147:114-129. Blumstein, Alfred. 1998. Crime decline-why? National Institute of Justice Journal 237:7-20. Blumstein, Alfred., Jacqueline Cohen and Davi d P. Farrington. 1988. Longitudinal and criminal career research: Further classifications. Criminology 26:57-74. Blumstein, Alfred, and Jacqueline Cohe n. 1987. Characterizing criminal careers. Science 237:985-991. 166

PAGE 167

Blumstein, Alfred, Jacqueline Cohe isty A. Visher (eds.). 1986. Criminal Careers and “Career Criminals, ” vol. I. Washington, D.C.: National Academy Blumstein, Alfred, David P. Farrington a nd Soumyo Moitra. 1985. Delinquency careers: e : University of Chicago Press. itment to e 91inology 19:129-153. course haiken, Jan M., and Marcia R. Chaiken. 1982. Varieties of Criminal Behavior . Santa Monica, Cheurprakobkit, Sutham. 2000. Police-citizen co ntact and police performance: Attitudinal Chung, st and Daniel S. Nagin. 2002. Childhood predictors of offense trajectories. Journal of Research in Crime and n, Jeffrey A. Roth and Chr Press. Innocents, desisters, and persisters. In Michael Tonry and Norval Morris (eds.), Crim and Justice: A Review of Research, vol. 6. Chicago Bobo, Lawrence, and James R. Kluegal. 1993. Oppos ition to race-targeting : self-interest, stratification ideology, or racial attitudes? American Sociological Review 58:443-464. Briar, Scott, and Irving Piliavin. 1965. Delinquenc y, situational inducements, and comm conformity. Social Problems 13:35-45. Bulmer, Martin. 1984. The Chicago School of Sociology: In stitutionalization, Diversity, and th Rise of Sociological Research . Chicago: University of Chicago Press. Bureau of Justice Statistics. 2005. (last accessed on December 2, 2006). Bursik, Robert J. Jr. 1988. Social disorganiz ation and theories of crime and delinquency: . Problems and prospects. Criminology 26:519-551 Bushway, Shawn, D., Alex R. Piquero, Lisa M. Br oidy, Elizabeth Cauffman and Paul Mazerolle. 2001. An empirical framework for studying desistance as a process. Criminology 39:4 515. Bushway, Shawn D., Terence P. Thornberry a nd Marvin D. Krohn. 2003. Desistance as a developmental process: A comparison of static and dynamic approaches. Journal of Quantitative Crim Caspi, Avshalom, Glen H. Elder and Daryl J. Bem. 1987. Moving against th e world: Lifepatterns of explosive children. Developmental Psychology 23:308-313. C ernkovich, Stephen A., and Peggy C. Giordano. 2001. Stability and change in antisocial behavior: The transition from adolescence to early adulthood. Criminology 39:371-410. C CA: Rand. differences between Hispanics and non-Hispanics. Journal of Criminal Justice 28:325336. Ick-Joong, Karl G. Hill, J. David Hawkins, Lewayne D. Gilchri Delinquency 39:60-90. 167

PAGE 168

Clarke, Ronald V., and Derek B. Cornish. 1985. Modeling offenders’ decisions: A framew for research and policy. In Michae l Tonry and Norval Morris (eds.), Crime and Justice: A Review ork of Research , vol. 6. Chicago: University of Chicago Press. nstitute Clear, Todd R., Elin Waring and Kristen Scull y. 2005. Communities and reentry. In Jeremy ge, y of Punishment and Correction. New York: Wiley. Clowar eory of Delinquent Gangs . New York: Free Press. Cohen, Lawrence, and Marcus Felson. 1979. Social change and crime rate trends: A routine activities approach. American Sociological Review 44:588-608. Cohn, S an. 1991. Punitive attitudes toward criminals: Racial consen sus or racial conflict? Social Problems , 38:287-296. Colvin, Palgrave Press. ullen, Francis T., Paul Gendreau, G. Roger Jarj oura and John Paul Wright. 1997. Crime and the Dean, Charles W., Robert Brame and Alex R. Piquero. 1996. Criminal propensities, discrete Delisi, Matt. 2001. Extreme criminal careers. American Journal of Criminal Justice 25:239Denno, Deborah W. 1990. Biology and Violence: From Birth to Adulthood . Cambridge, UK: Du Boi The Souls of Black Folks . Chicago: A.C. McClurg & Co. nal of Sociology 103:1593-1630. Clear, Todd R. 1988. Statistical Predictions in Corrections . Washington, D.C.: National I of Corrections. Travis and Christy Visher (eds.), Prisoner Reentry and Crime in America . Cambrid UK: Cambridge University Press. Clemmer, Donald. 1970. Prisonization. In Norman Bruce Johnston, Leonard D. Savitz and Marvin E Wolfgang (eds.), The Sociolog d, Richard, and Lloyd Ohlin. 1960. Delinquency and Opportunity : A Th teven F., Steven E. Barkan and William A. Haltem Mark. 2000. Crime and Coercion: An Integrated Theory of Chronic Criminality . New York: Cullen, Francis T. and Robert Agnew. 2003. Criminological Theory: Past to Present . Los Angeles: Roxbury. C bell curve: Lessons from intelligent criminology. Crime and Delinquency 43:387-411. groups of offenders and persistence in crime. Criminology 34:547-574. 252. Cambridge University Press. s, W.E.B. 1903. D’Unger, Amy V., Kenneth C. Land, Patricia L. McCall and Daniel S. Nagin. 1998. How many latent classes of delinquen t/criminal careers? Results fr om mixed Poisson regression analyses. American Jour 168

PAGE 169

Eggers, Mitchell L., and Douglas S. Massey. A long itudinal analysis of urban poverty: Blacks i U.S. metropolitan areas between 1970 and 1980. Social Science Research 21:175-203. n Eggleston, Elaine P., and John H. Laub. 2002. Th e onset of adult offending: A neglected ggleston, Elaine P., John H. Laub and Robert J. Sampson. 2004. Methodological sensitivities to Elliott, Delbert S. 1994. Serious violent offenders: Onset, developmental course and termination-22. Farrington, David, P. 1986. Age and crime. In Michael Tonry and Norval Morris (eds.), Crime and Justice: A Review of Research , vol. 7. Chicago: University of Chicago Press. Farring evention of juvenile delinquency. Criminal Behaviour and Mental Health 4:209-227. Farring l behavior from childhood to adulthood. In Peter Cord ella and Larry Siegel (eds.), Readings in arrington, David P. 2000. Explaining and preventi ng crime: The globalization of knowledge Farrington, David P. 2003. Developmental and lif e-course criminology: Key theoretical and Farrington, David P., and J. David Hawkins. 1991. Predicting participation, early onset and later . f Criminology, 26:335-356. Farrington, David P., Rolf Loeber, Yanming Yin and Stewart J. Anderson. 2002. Are withinindividual causes of delinquency th e same as between-individual causes? Criminal Feagin, Joe R. 2001. Realities, and Future Reparations . New dimension of the criminal career. Journal of Criminal Justice 30:603– 622. E latent class analysis of long-term criminal trajectories. Journal of Quantitative Criminology 20:1-26. The American Society of Crimi nology 1993 Presidential Address. Criminology 32:1 ton, David P. 1994. Early developmen tal pr ton, David P. 1996. The development of offending and anti-s ocia Contemporary Criminological Theory . Boston, MA: Northeastern University Press. F The American Society of Crimi nology 1999 Presidential address. Criminology 38:1-24. empirical issues-The 2002 Sutherland Award Address. Criminology 41:221-256. persistence in officially recorded offending. Criminal Behaviour and Mental Health 1:133. Farrington, David P., Bernard Ga llagher, Lynda Morley, Raymond J. St. Ledger and Donald J West. 1986. Unemployment, sc hool leaving, and crime. British Journal o Behaviour and Mental Health 12: 53-68. Racist America: Roots, Current York: Routledge. Fergusson, David, Nicola Sawin-Campbe ll and John Horwood. 2004. How does childhood economic disadvantage lead to crime? Journal of Child Psychology and Psychiatry 45:956-966. 169

PAGE 170

Ge, Xiaojia, M. Brent Donnellan and Ernst Wenk. 2001. The development of persistent criminal offending in males. Criminal Justice and Behavior 28: 731-755. Geis, Gilbert. 2000. On the absence of self-control as the basis for a genera l theory of crim e: A Therapy inology 45: 574-592. e studies, and related topics. Criminology 24:213-34. Gottfre ngitudinal : Grasmi pirical im plications of Gottfredson and Hirschi’s general theory of Greenwood, Peter W fenders are Hard to Predict . Santa Monica, CA: Rand. y of Brunswick, NJ: Transaction Publishers. five critique. Theoretical Criminology 4: 35-53. Gibson, Chris L., Alex R. Piquero and Stephe n G. Tibbetts. 2001. The contribution of family adversity and verbal IQ to criminal behavior . International Journal of Offender and Comparative Crim Glueck, Sheldon, and Elea nor Touroff Glueck. 1950. Unraveling Juvenile Delinquency . New York: Commonwealth Fund. Goring, Charles. 1913. The English Convict: A Statistical Study . Montclair, NJ: Petterson Smith, 1972. Gottfredson, Michael R., and Travis Hirschi. 1986. The true value of Lambda would appear to b zero: An essay on career criminals, crimin al careers, selective incapacitation, cohort dson, Michael R., and Travis Hirschi. 1987. The methodological adequacy of lo research on crime. Criminology, 25: 581-614. Gottfredson, Michael R., and Travis Hirschi.198 8. Science, public policy, and the career paradigm. Criminology 26:37-55. Gottfredson, Michael R. a nd Travis Hirschi. 1990. A General Theory of Crime . Stanford, CA Stanford University Press. Gould, Stephen J.1996. The Mismeasure of Man . New York: W.W. Norton. ck, Harold G., Charles R. Tittle, Robert J. Bursik, Jr. and Br uce J. Arneklev. 1993. Testing the core em crime. Journal of Research in Crime and Delinquency 30:5-29. ., and Susan Turner. 1987. Selective Incapacitation Revisited: Why the High-Rate Of Haapanen, Rudy. 1990. Selective Incapacitation and the Serious Offender: Longitudinal Stud Criminal Career Patterns . New York: Springer-Verlag. Hagan, John. 1997. Crime and Capitalization: Toward a Developmental Theory of Street Crime rence P. Thornberry (ed.), Developmental Theories of Crime and in America. In Ter Delinquency . New Haney, Craig, and Philip Zimbardo 1998. The past and future of U.S. prison policy: Twenty years after the Stanford Prison experiment. American Psychologist 53 : 709-727. 170

PAGE 171

Hankins, Frank H. 1968. Adolphe Quetelet as Statistician. New York: AMS Press. Harer, Miles D., and Darrell J. Steffensmeier. 1992. The different effects of economic inequality Hawkins, J. David, and Ri chard F. Catalano, Jr. 1992. Communities That Care: Action for Drug Hay, C uency: A test of self-control theory. Criminology 39:707– 736. Heilbru is White, Ayonda Lanier, Martin Schmid, Gretchen White, Michelle Keeney, Melissa Westendorf, Luis Buinavert and Michael Shumate. 2000. Risk e offenders. Criminal Justice and Behavior 27:275-291. e Press. American e. In Michael R. ew Brunswick, NJ: Hong, S modeling across ethnic Horney erican Society of Horney 5. Criminal careers in the shortHuang, s and Richard F. Catalano. on black and white rates of violence. Social Forces 70:1035-1054. Abuse Prevention . San Francisco: Jossey-Bass. arter. 2001. Parenting, self-c ontrol, and delinq n, Kirk, William Brock, Denn factors for juvenile criminal recidivism : The postrelease community adjustment of juvenil Herbst, Leigh, and Samuel Walker. 2001. Language barr iers in the delivery of police services: A study of police and Hispanic inter actions in a midwestern city. Journal of Criminal Justice 29:329-340. Hernnstein, Richard J., and Charles A. Murray. 1994. The Bell Curve: Intelligence and Class Structure in American Life . New York: The Fre Hirschi, Travis. 1969. Causes of Delinquency. Berkeley: University of California Press. Hirschi, Travis, and Gottfredson, Michael . 1983. Age and the explanation of crime. Journal of Sociology 89:552-584. Hirschi, Travis, and Michael R. Gottfredson. 1994. The generality of devianc Gottfredson and Travis Hirschi (eds.), The Generality of Deviance . N Transaction Publishers. ehee, and Hsiu-Zu Ho. 2005. Direct and i ndirect longitudinal e ffects of parental involvement on student achievement: Second-or der latent growth groups. Journal of Educational Psychology 97:32-42. , Julie. 2006. An alternative psychology of cr iminal behavior-The Am Criminology 2005 Presidential Address. Criminology 44:1-16. , Julie, D. Wayne Osgood, and Ineke Haen Marshall. 199 term: Intra-individual variability in crime a nd its relation to local life circumstances. American Sociological Revie w 60: 655-673. Bu, Helen Raskin White, Rick Kosterman, J. David Hawkin 2000. Developmental associations between alco hol and interpersonal aggression during adolescence. Journal of Research in Crime and Delinquency 38: 64-83. 171

PAGE 172

Irwin, John, and Donald R. Cressey. 1962. Thie ves, convicts, and the inmate culture . Social Problems 10:142-155. Jajoura nvolvement? Results from a large-scale national probability sample. Criminology 31:149-172. Jessor, Richard, John Edward Donovan and Frances Marie Costa. 1991. Beyond Adolescence: Problem Behavior and Young Adult Development . Cambridge, UK: Cambridge nes, Bobby L., Daniel S. Nagin and Kathryn Roeder. 2001. A SAS procedure based on mixture Kanaya lynn effect and U.S. policies: The impact of rising IQ scores on American society via mental retardation Kandel, Elizabeth, Sarnoff A. Mednick, Lise Ki rkegaard-Sorensen, Barry Hutchings, Joachim Knop, J. Robert Rosenberg, and Fini Schulsinge r. 1988. IQ as a protective factor for tistical Association 90:773-795. Kenned nd the Law. New York: Vintage. Kluegal, James R. 1990. Trends in whites’ explanations of the black-white Gap in Kratzer enders: A test of Moffitt’s theory among males and females from childhood to age 30. Criminal Behaviour and Mental rivo, Lauren J., and Ruth D. Petersen. 2000. The structural context of homicide: Accounting for Krivo, John R. Reynolds. 1998. Race, segregation, and the concentration of disadvantage: 1980-1990. Social Problems 45:61-80. LaFree rime: Continuity and Change, vol. 1. Washington D.C.: U.S. Department of Justice. , G. Roger. 1993. Does dropping out of school enhance delinquent i University Press. Jo models for estimating developmental trajectories. Sociological Research & Methods 29:374-393. , Tomoe, Scullin, Matthew H., and Stephen J. Ceci. 2003. The F diagnoses. American Psychologist 58:778-790. subjects at high risk for antisocial behavior. Journal of Consulting and Clinical Psychology 56:224-226. Kass, Robert E., and Adrian E. Raferty. 1995. Bayes factors. Journal of the American Sta y. Michael R. 1997. Race, Crime, a Klein, Stephen, Joan Petersilia, and Susan Turn er. 1990. Race and imprisonment decisions in California. Science 247:812-816 . socioeconomic status, 1977-1989. American Sociological Review 55:512-525. , Lynn, and Sheilagh Hodgins. 1999. A typology of off Health 9:57-73. K racial differences in process. American Sociological Review 65:547-559. Lauren J., Ruth D. Peterson, Helen Rizzo and , Gary, Robert J. Burs ik, Sr., James Short and Ralph B. Taylor. 2000. The Changing Nature of Crime in America. In Criminal Justice, 2000 : The Nature of C 172

PAGE 173

Land, Kenneth C., Patricia L. McCall and Daniel S. Nagin. 1996. A comparison of Poisson, negative binomial and s emiparametric mixed Poisson regression models: With empirical applications to criminal careers data. Sociological Methods of Research 24:387. Larson rison. Journal of Criminal Justice 12: 601-615. Lattimo risty A. ying the characteristics of arrest frequency among paroled youthful Laub, J y ical knowledge. American Journal of Sociology 96:1402-1440. el f Laub, John H., Daniel S. Nagin and Robert J. Sampson. 1998. Trajectories of change in criminal offending: Good marriag es and the desistance process. American Sociological Le Blanc, Marc, and Rolf Loeber. 1993. Causality and the development of offending. In Dale F. t and Le Blanc, Marc, and Rolf Loeber. 1998. Devel opmental criminology updated. In Michael Tonry (ed.), Crime and Justice: A Review of Research, vol. 23. Chicago: University of Chicago Le Blanc, Marc, and Marcel Frchette. 1989. Male Criminal Activity from Childhood through Loeber, Rolf, and David P. Farrington. 1998. Serious and Violent Juvenile Offenders . Thousand Oaks, CA: Sage. Loeber . Farrington. 1991. Initiation, escalation, and desistance in juvenile offending and their correlates. Loeber, Rolf, and Marc Le Blanc. 1990. Toward a developmental criminology. In Michael Tonry : ogan, John R. 2003. How race counts for Hispanic Americans . Lewis Mumford Center for , James H., and Joey Nelson. 1984. Wome n, friendship, and adaptation in p re, Pamela K., John M. MacDonald, Alex R. Piquero, Richard L. Li nster, and Ch Visher. 2004. Stud offenders. Journal of Research in Crime and Delinquency 41:37-57. ohn H., and Robert J. Sampson. 1991. The Sutherland-Glueck debate: On the sociolog of criminolog Laub, John H., and Robert J. Sampson. 2001. Understanding desistance from crime. In Micha Tonry (ed.), Crime and Justice: A Review of Research , vol. 28. Chicago: University o Chicago Press. Laub, John H., and Robert J. Sampson. 2003. Shared Beginnings, Diverge nt Lives: Delinquent Boys to Age 70. Cambridge, MA: Harvard University Press. Review 63:225. Hay and Adrian Angold (eds.), Precursors and Causes in Developmen Psychopathology. New York: Wiley Press. Press. Youth: Multilevel and Developmental Perspectives. New York: Springer-Verlag. , Rolf, Magda Stouthamer-Loeber, Welm oet B. Van Kammen and David P Journal of Criminal Law and Criminology 82:36-82. and Norval Morris (eds.), Crime and Justice: A Review of Research, vol. 12. Chicago University of Chicago Press. L 173

PAGE 174

Comparative Urban and Regional Re search. University at Albany. ackLatinoReport/BlackLatinoReport.pdf (last accessed on December 2, 2006). Lombroso-Ferro, Gina. 1972. Criminal Man: According to the Clarification of Cesare Lombroso. Montclair, NJ: Patterso n Smith Publishing Group. Longshore, Douglas, Eunice Chang, Shih-chao Hs ieh and Nena Messina. 2004. Self-control and rmal Psychology 102:187-196. Matza, David. 1964. Delinquency and Drift . New York: Wiley. Mauer, . Washington, D.C.: The Sentencing Project. on D.C.: National Academy Press. Comparative C riminology Merton Long, J. Scott. 1997. Regression Models for Categorical and Limited Dependent Variables . Thousand Oaks, CA: Sage Publications. social bonds: A combined control perspective on deviance. Crime & Delinquency 50:542-564. Lynam, Donald, Terrie E. Moffitt and Magda St outhamer-Loeber. 1993. Explaining the relation between IQ and delinquency: Clas s, race, test motivation, school failure, or self-control? Journal of Abno Maruna, Shadd. 2001. Making Good: How Ex-Convicts Reform and Rebuild Their Lives. Washington, D.C.: American Psychological Association. Maruna, Shadd, and Russ Immarigeon. 2004. After Crime and Punishment . London: Will an Publishing. M aruna, Shadd, and Hans Toch. 2005. The impact of imprisonment on the desistance process. In Jeremy Travis and Christy Visher (eds.), Prisoner Reentry and Crime in America. New York: Cambridge University Press. Marc. 1990. More Young Black Males Under Corre ctional Control in U.S. than in College Mazzerolle, Paul, Robert Brame, Raymond Paternoster, Alex R. Piquero and Charles Dean. 2000. Criminology 38:1143-1172. McCord, Joan, Cathy Spatz Wido m and Nancy A. Crowell. 2001. Juvenile Crime, Juvenile Justice. Washingt McGloin, Jean Marie, and Travis C. Pratt. 2003. Cognitive ability and delinquent behavior among inner-city youth: A life-c ourse analysis of main, mediating, and interaction effects. International Journal of Of fender Therapy and 47:253-271. , Robert King. 1949. Social Theory and Social Structure . New York: The Free Press. Mitchell, Ojmarrh, and Doris Layton MacKenzi e. 2006. The stability and resiliency of selfcontrol in a sample of incarcerated offenders. Crime & Delinquency 52:432-449. 174

PAGE 175

Moffitt : A logical Review 100:674-701. ts 0. P agin, 9. Analyzing developmental trajectories: A semi-parametric, group-based Nagin, es of offenders. Criminology 33:111-140. nship of past and future Nagin, es of boys’ physical aggression, ong urnal of Behavioral Sciences O’Brie ournal e Criminology 19:1-32. ns: , Terrie E. 1993. Adolescence-limited and life-course persistent antisocial behavior developmental taxonomy. Psycho Moffitt, Terrie E., Donald R. Lynam and Phil A. Silva . 1994. Neuropsychological tes predicting persistent male delinquency. Criminology 32:277-30 Montagu, M. F. Ashley. The biologist looks at crime. Annals of the American Academy of olitical and Social Science , 217:46-57. Morris, Norval. 1972. Forward. In Marvin E. Wolf gang, Robert M. Figlio and Thorsten Sellin, Delinquency in a Birth Cohort . Chicago: University of Chicago Press. Daniel S. 199 N approach. Psychological Methods 4:139-177. Daniel S. 2005. Group-Based Modeling of Deve University Press. lopment. Cambridge, MA: Harvard Nagin, Daniel S., and David P. Farrington. 1992. The onset and persistence of offending. Criminology 30:501-23. Nagin, Daniel S., David P. Farri ngton, and Terrie E. Moffitt. 1995. Life course trajectories of different typ Nagin, Daniel S., and Kenneth C. Land. 1993. Age, criminal careers and population heterogeneity: Specification and estimation of a nonparametric, mixed Poisson model. Criminology 31:327-363. Nagin, Daniel S., and Raymond Paternoster. 1991 . On the relatio participation in delinquency. Criminology 29:163-90. Daniel. S., and Richard E. Tremblay. 1999. Trajectori opposition, and hyperactivity on the path to phys ically violent and nonviolent juvenile :1181-1196. delinquency. Child Development 70 N eff, James Alan, and Albana M. Dassori. 1998. Age and maturing out of heavy drinking am Anglo and minority male drinkers: A co mparison of cross-sectional data and retrospective drinking history techniques. Hispanic Jo 20:225-240. n, Robert M., and Jean Stockard. 2003. The cohort-size sample size conundrum: An empirical analysis and assessment using homicide arrest data from to 1960-1999. J of Quantitativ Parker, Karen F., and McCall, Patricia. 1991. Struct ural conditions and racial homicide patter A look at the multiple disadvantages in urban areas. Criminology 37:447-477. 175

PAGE 176

P aternoster, Raymond. 1989. Decision s to participate in and desist from four types of common an er and Kenneth H. Rubin (eds.), The Development and Treatment of Childhood Aggression . Hillsdale, NJ: Erlbaum. Patterso nquent behavior. Patterson, Gerald R., and Karen Yoerger. 1997. A developmental model for late on-set ), s. Piquero life-course persistent offending. Psychology, Crime & Law 7:193-215. 16:54-74. delinquency: Deterrence and the rational choice perspective. Law and Society Review 23:7-40. Patterson, Gerald. R. (1990). Some comments about cognitions as causal variables. Americ Psychologist 45:984-985. Patterson, Gerald. R., Deborah M. Capaldi and Lew Bank. 1991. An early starter model for predicting delinquency. In Debra J. Pe pl n, Gerald R., and Karen Yoerger. 1993. A m odel for early onset of deli In Sheilagh Hodgins (ed.), Mental Disorder and Crime . Newbury Park, CA: Sage. delinquency. In D. Wayne Osgood (ed.) Nebraska Symposium on Motivation Vol. 44: Motivation and Delinquency. Lincoln, NE: University of Nebraska Press. Petersilia, Joan. 1983. Racial Disparities in th e Criminal Justice System . Santa Monica, CA: Rand Corporation. Petersilia, Joan. 1999. Parole and prisoner reentry in the United States. In Michael Tonry and Joan Petersilia (eds.), Prisons: Crime and Justice: A Review of Research , vol. 26. Chicago: University of Chicago Press. Petersilia, Joan. 2005. From cell to society: Wh o is returning home? In Jeremy Travis and Christy Visher (eds.) Prisoner Reentry and Crime in America . New York: Cambridge University Press. Petersilia, Joan, Peter Greenwood and Marvin Lavin. 1977. Criminal Careers of Habitual Felons . Santa Monica, CA: Rand. Petersilia, Joan and Susan Turn er. 1993. Intensive probation and pa role. In Michael Tonry (ed. Crime and Justice: A Review of Research , vol. 17. Chicago: University of Chicago Pres Piehl, Anne Morrison. 1998. Economic conditions, work, and crime. In Michael Tonry (ed.), The Handbook of Crime and Punishment . New York: Oxford University Press. , Alex R. 2001. Testing Moffitt’s neur opsychological variation hypothesis for the prediction of Piquero, Alex R. and Jeff A. Bouffard. Fo rthcoming. Something Ol d, Something New: A Preliminary Investigation of Hirschi’s Redefined Self-Control. Piquero, Alex R., Alfred Blumstein, Robert Brame, Rudy Haap anen, Edward P. Mulvey and Daniel S. Nagin. 2001. Assessing the impact of exposure time and incapacitation on longitudinal trajectories of criminal offending. Journal of Adolescent Research 176

PAGE 177

Piquero, Alex R., Robert Brame, and Donald L ynam. 2004. Do the factors associated with lifecourse-persistent offending relate to career length? Crime & Delinquency 50:412-435. Piquero, Alex R., Robert Brame, Paul Mazero lle, and Rudy Haapanen. 2002. Crime in emerging iquero, Alex R., Robert Brame, Paul Mazerolle and Rudy Haapanen. 2001. Crime in emerging iquero, Alex R., and He Len Chung. 2001. On the relationships between gender, early onset, Piquero, Alex R., David P. Farrington and icago: Piquero, Alex R., John M. McDonald and Karen F. Parker. 2002. Race, local life circumstances, iquero an A. White. 2003. On the relationship between cognitive abilities nal Pope, Carl E Disproportionate minority confinement: A ffice Pratt, T empirical status of Gottfredson and Hirschi’s uetelet, M. Adolphe. 1968. . New affert Reed, G ssical -382. Reiss, Albert J., Jr. 1986. Co-offende and my Press. : aw adulthood. Criminology 40:137-70. P adulthood: Continuity and change in criminal offending. Washington D.C.: National Institute of Justice. P and the seriousness of offending. Journal of Criminal Justice 29:189-206. Alfred Blumstein. 2003. The criminal career paradigm. In Michael Tonry (ed.), Crime and Justice: A Review of Research. Ch University of Chicago Press. and criminal activity. Social Science Quarterly 3:654-670. P , Alex R., and Norm and lifecourse-persistent offending among a samp le of African Americans: A longitudi test of Moffitt’s hypothesis. Journal of Criminal Justice 31:399-409. ., Rick Lovell, and Heidi M. Hsia. 2002. review of the research li terature from 1989 through 2001 . Washington D.C.: U.S. O of Juvenile Justice and Delinquency Prevention. ravis C. and Francis T. Cullen. 2000. The general theory of crime: A meta-analysis. Criminology 38:931-964. A Treatise on Man and the Development of his Faculties Q York: Burt Franklin. y, Adrian E. 1995. Bayesian Model Selecti on in Social Research. Sociological R Methodology 25:111-64. ary E., and Peter Cleary Yeager. 1996. Organizational offending and neocla criminology: Challenging the reach of general theory of crime. Criminology 34:352 r influenc es on criminal careers. In Alfred Blumstein, Jacqueline Cohen, Jeffrey A. Rot h, and Christy A. Visher (eds.), Criminal Careers “Career Criminals,” vol. II. Washington, D.C.: National Acade Reiss, Albert J., Jr., and Da vid P. Farrington. 1991. Advancing knowledge about co-offending results from a prospective longi tudinal survey of London males. Journal of Criminal L and Criminology 82:360-95. 177

PAGE 178

Reitzel, John D., Stephen K. Rice, and Alex R. Piquero. 2004. Lines and shadows: Perceptions of racial profiling and the Hispanic experience. Journal of Criminal Justice 32:607-616 . 1-183. Robins, Lee N. 2006. Explaining wh en arrests end for serious juvenile offenders: Comments on n Rowe, David C., D. W . Nicewander. 1990. A latent trait approach to Rowe, Allan, and Charles R. Tittle. 1977. Life cycle changes and criminal propensity. Sampso fields: A revised theory of community-level ra cial inequality. In Ruth Pe terson, Lauren Krivo and John Sampson, Robert J., and John H. Laub. 1992. Crime and deviance in the life course. Annual Sampson, Robert J., and John H. Laub. 1993. Crime in the Making: Pathways and Turning Sampson, Robert J., and John H. Laub. 1995. Unders tanding variab ility in lives through time: 4:143ge Theories of Crime and Delinquency: Volume 7 of Advances in Criminological Theory . Sampso delinquent boys followed to age 70. Criminology 41:555-592. Reitzel, John D. and Alex R. Piquero. 2006. Does it exist? Studying citizen s’ attitudes toward racial profiling. Police Quarterly 9:16 Rice, Stephen K., John D. Reitzel, and Alex R. Piquero. 2005. Shades of brown: Racial profiling and the intra-ethnic differential. Journal of Ethnicity in Criminal Justice 3:47-70. the Sampson and Laub study. The Annals of the American Academy of Political and Social Science 602:57-72. Robinson, W. S. 1950. Ecological Correlations and the behavior of individuals. America Sociological Review 15:351-357. ayne Osgood and Alan W unifying criminal careers. Criminology 28:237-70. Sociological Quarterly 18:223-236. n, Robert J., and Lydia Bean. 2006. Cultural mechanisms and killing Hagan (eds.) The Many Colors of Crime: Inequaliti es of Race, Ethnicity and Crime in America. New York: New York University Press. Review of Sociology 18:63-84. Points Through Life. Cambridge, MA: Harvard University Press. Contributions of life-course criminology. Studies on Crime & Crime Prevention 158. Sampson, Robert J., and John H. Laub. 1997. A lifecourse theory of cumulative disadvanta and the stability of delinquency. In Terrence P. Thornberry (ed.), Developmental New Brunswick, N.J.: Transaction Publishing. n, Robert J., and John H. Laub. 2003. Life-course desisters? Trajectories of crime among Sampson, Robert J., and John H. Laub. 2004. Shared Beginnings, Diverge nt Lives: Delinquent Boys at Age 70. Cambridge, MA: Harvard University Press. 178

PAGE 179

Sampson, Robert J., and John H. Laub. 2005a. Seduc tions of method: Rejoinder to Nagin and Tremblay’s Developmental Trajectory Groups: Fact or Fiction? Criminology 43:905-915 . . ical and Social Science 602:12-45. ting Sampson, Robert J. and Janet L. Lauritsen. 1997. Racial and ethnic disp arities in crime and tute. Shaw, C s. approach. Law & Society Review 9:581-611. CO: e. Simons, Ronald L., Christine Johnson, Rand D. Conger and Glen Elder Jr. 1998. A test Simons redrick O. Lorenz. 1994. Two routes to Sampson, Robert J., and John H. Laub. 2005b. A life-c ourse view of the development of crime The Annals of the American Academy of Polit Sampson, Robert J., and John H. Laub. 2005c. A gene ral age-graded theory of crime: Lessons learned and the future of life-course criminology. In David P. Farrington (ed.), Tes Integrated Developmental/Life Course Theories of Offending. New Brunswick, NJ: Transaction. criminal justice in the United States. In Michael Tonry (ed.), Crime and Justice: A Review of Research . vol. 23. Chicago: University of Chicago Press. Sampson, Robert J. and William Julius Wilson. 1995. To ward a theory of race, crime, and urban s.), Crime and Inequality . Stanford, inequality. In John Hagan and Ruth D. Peterson (ed CA: Stanford University Press. Sellin, T outh . Philadelphia, PA: American Law Insti horsten. 1940. The Criminality of Y S haw, Clifford Robe. 1930. The Jack-Roller: A Delinquent Boy’s Own Story . Chicago: University of Chicago Press. lifford Robe, and Henry Donald McKay. 1942. Juvenile Delinquency and Urban Area Chicago: University of Chicago Press. Shinnar, Shlomo, and Reuel Shinnar. 1975. The eff ects of the criminal justice system on the control of crime: A quantitative Shover, Neal. 1996. Great Pretenders: Pursuits and Careers of Persistent Thieves . Boulder, Westview Press. Shover, Neal and Carol Y. Thompson. 1992. Age diffe rential expectations a nd crime desistanc Criminology 30:89-104. of latent trait versus life-c ourse perspectives on the stability of adol escent anti-social behavior. Criminology 36:217. , Ronald L, Wu Chyi-In, Rand D. Conger, and F delinquency: differences in the impact of pare nting and deviant peers for early versus late starters. Criminology 32:247-275. Simpson, Sally S., and Nicole Leeper Piquero. 2002. Low self-control, or ganizational theory, and corporate crime. Law and Society Review 36: 509-548. 179

PAGE 180

Slovanien, Jukka. 2000. Relative cohort size and age-specific arrest rates: A conditional interpretation of the Easterlin affect. Criminology 38:117-136. Steffensmeier, Darrell J., Emilie Andersen Allan, Miles D. Harer, and Cathy Streifel. 1989. Age and the distribution of crime. American Journal of Sociology 94:803-831. Sutherland, E. H. 1931. Mental deficiency and crime. In Kimball Young (ed.), Social Attitudes. New York: Henry Holt. Gresham. 1971. Society of Captives: A Study of a Maximum Security Prison. Princeto J: Princeton University Sykes, n, Press. on, Thornberry, Terence P. 1987. Toward an interactional theory of delinquency. Criminology course ed.), cy: Advances in Criminology Theory Thornberry, Terence P. 2 tterns of offending across the life course ironment in predicting early onset of offending: A test of Moffitt’s interactional hypothesis. Criminology 37:843-78. Tittle, C ties maybe in search of an explanation: Commentary on the age/crime debate. Criminology 26:75-85. Tittle, C O: olan, P it h. 1998. Development of serious and violent N Thernstrom, Stephan, and Abigail Thernstrom. 1997. America in Black and White: One Nati Indivisible. New York: Simon and Shuster. Thomas, William I., and Florian Znaniecki. 1958. The Polish Peasant in Europe and America. New York: Dover Publications. 25:863-891. Thornberry, Terence P. 1997. Introduction: Some advantages of developmental and lifeperspectives for the study of crime and delinquency. In Terrence P. Thornberry ( Developmental Theories of Crime and Delinquen vol. 7. New Brunswick, N.J.: Transaction. 005. Explaining multiple pa and across generations. The Annals of the American Academy of Political and Social Science 602:157-195. Tibbetts, Stephen G., and Alex R. Piquero. 1999. Th e influence of gender, low birth weight, and disadvantaged env harles R. 1988. Two empirical regulari harles R. 1995. Control Balance: Toward a General Theory of Deviance . Boulder, C Westview Press. Tittle, Charles R., and Harold G. Grasmick. 1998. Criminal behavior and ag e: A test of three provocative hypotheses. Criminology 88:309-342. atrick H., and Deborah Gorman-Sm T offending careers. In Rolf Loeber and David P. Farrington (eds.), Serious & Violent ns . Thousand Oaks, CA: Juvenile Offenders: Risk Factors and Successful Interventio Sage. 180

PAGE 181

Tonry, Michael. 1995. Malign Neglect : Race, Crime and Punishment in America. New Y Oxford University Press. ork: d. 1992. Parent and child training to prevent early onset of delinquency: The Montreal longitudinal-experimental study. In Joan McCord and Press. nd Christa Japel. 2004. Physical a ggression during early Turner, Michael, and Alex R. Piquero. 2002. The stability of self-control. Journal of Criminal Uggen, Uggen, Christopher, and Irving Piliavin. 1998. Asym etrical causation and criminal desistance. U.S. Ce nd Poverty, 2000. Washington D.C.: U.S. Department of Economics and Statisti cs Administration. Vold, G data and 3-95. rica. Belmont, CA: Wadsworth. lifestyle. Social Problems 38:333357. Walters 21. Warr, Mark. 1993. Age, peers, and delinquency. Criminology 31:17-40. Travis, Jeremy, and Christy Visher. 2005. In Je remy Travis and Christy Visher (eds.), Prisoner Reentry and Crime in America. New York: Cambridge University Press. Tremblay, Richard E. 2005. Tracking the orig ins of behaviour: Back to the future. The Criminologist 31:1-7. Tremblay, Richard E., Frank Vitaro, Lucie Be rtrand, Marc Le Blanc, Hlne Beauchesne, Hlne Boileau and Lucille Davi Richard E. Tremblay (eds.), Preventing Antisocial Behavior: Interventions from Birth Through Adolescence. New York: Guilford Tremblay, Richard E., Daniel S. Nagin, Jean R. Seguin, Mark Zoccolillo, Philip D. Zelazo, Michel Boivin, Daniel Perusse, a childhood: Trajectories and predictors. Pediatrics 114:e43-e50. Justice 30:457-472. Christopher, and Jeremy Staff. 2001. Work as a turning point for criminal offenders. Corrections Management Quarterly 5:1-16. National Institute of Justice Journal 237:20. nsus Bureau. Income a eorge B., Thomas J. Bernard and Jeffrey B. Snipes. 2002. Theoretical Criminology. Oxford, UK: Oxford University Press. Walker, Samuel. 2001. Searching for the denominator : Problems with police traffic stop early warning system solution. Justice Research and Policy 3:6 Walker, Samuel, Cassia Spohn and Miriam DeLone. 2004. The Color of Justice: Race, Ethnicity, and Crime in Ame Wallace, John M., and Jerald G. Bachman. 1991. Explaining racial/ethni c differences in adolescent drug use: The imp act of background and , Glenn, D. 2003. Changes in criminal th inking and identity change in novice and experienced inmates: Prisonization revisited. Criminal Justice and Behavior 30:399-4 181

PAGE 182

Warr, Mark. 1998. Life-course transitions and desistance from crime. Criminology 36:183-216. Careers and “Career Criminals,” vol. 2. Washington, D.C.: National Academy Press. Weitze uch. 1999. Race, class, and perceptions of police. Crime & Delinquency 45: 494-507. Wells, White, e drug-crime relationship. In Gary LaFree (ed.), Criminal Justice 2000, Volume 1: The Nature of Crime: Continuity and White, r-Loeber. 2002. Illegal acts committed by adolescents under the influence of alcohol and drugs. Journal of illiam T. Grant Foundation. 1988. The Forgotten Half: Pathways to Success for America’s Wilson, James Q., and Richard J. Herrnstein. 1985. Crime and Human Nature . New York: Wilson, William Julius. 1978. The Declining Significance of Race: Blacks and Changing ilson, William Julius. 1987. The Truly Disadvantaged: The Inner City, the Underclass, and Wolfgang, Marvin E., and Franco Ferracuti. 1967. The subculture of violence: Towards an Wolfgang, Marvin E., Robert M. Figlio and Thorsten Sellin. 1972. Delinquency in a Birth Wolfgang, Marvin E., Terence P. Thor nberry and Robert M. Figlio. 1987. From Boy to Manright, Bradley R. E. 2002. Intelligence and Crime. In Joshua Dressler (ed.) Encyclopedia of Weis, Joseph G. 1986. Issues in the measurement of criminal careers.” In Alfred Blumstein, Jacqueline Cohen, Jeffrey A. Roth, and Christy A. Visher (eds.), Criminal r, Ronald, and Steven T L. Edward, and Rankin, Joseph H. 1991. Fam ily and delinquency: A meta-analysis of the impact of broken homes. Social Problems 38:71-93. Helen R., and D. M. Gorman. 2000. Dynamics of th Change. Washington D.C.: U.S. Department of Justice. Helen Raskin, Peter C. Tice, Rolf Loeb er, and Magda Stouthame Research in Crime and Delinquency 39:131-152. W Youth and Young Families. New York: William T. Grant Foundation Simon and Schuster. American Institutions . Chicago: University of Chicago Press. W Public Policy . Chicago: University of Chicago Press. Integrated Theory in Criminology. London: Tavistock Publications. Cohort . Chicago: University of Chicago Press. From Delinquency to Crime: Follow-up to the Philadelphia Birth Cohort of 1945 . Chicago: University of Chicago Press. W Crime and Justice. New York: Macmillan. 182

PAGE 183

W right, Bradley R. E., Avshalom Caspi, Terr ie Moffitt and Ray Paternoster. 2004. Does the control, and crime. Journal of Research in Crime and Delinquency 41:180-213. perceived risk of punishment deter criminally -prone individuals? Rational choice, self183

PAGE 184

BIOGRAPHICAL SKETCH John David Reitzel is a native New Yorker a nd was born in Astoria, Queens, New York ity. He grew up in Astoria and in the beauti ful Hudson Valley region just north of New York City. After graduating from Mini sink Valley Central High School in 1987, John enlisted in the United States Navy and served aboard two ships: Commander Sixth Fleet flagship, USS Belknap (CG-26), home-ported in Gaeta, Italy and the USS Fulton (AS-11), home-ported in New London, CT. While aboard the Belknap, he was fortunate to visit numerous countries and historical sights and to have taken part in the historic Malta Summit between Mikhail Gorbachev and George H. W. Bush. John earned a number of medals and aw ards while in the Navy including a National Defense Service Medal, Navy Expeditionary Me dal, Navy/Marine Corps Overseas Deployment Ribbon, Captain USS Fulton Lett er of Commendation, and a Commander, Submarine Group Two Letter of Commendation. After his stint in the Navy, John headed to SUNY Cortland to pursue a degree in physical education but was hooked by his first sociology c ourse. Following some uneven years of college where he left school twice, John returned to Cor tland in 1999 to finally earn his Bachelor of Arts in Sociology in May 2001. Encouraged by his pr ofessors at Cortland, John entered graduate school in May 2001 to study sociology/criminology at the University of Florida. He completed his Master of Arts in Sociology in May 2003 and his Doctorate of Philosophy in Criminology, Law & Society in December 2006. He is the aut hor of several publishe d journal articles and book chapters, and is currently Assistant Profe ssor of Criminal Justice at Illinois State University. In his spare time John enjoys spendi ng time with his wife, Kuniko, and four cats, playing sports, biking, trave ling, and watching Florida Gators football and New York Yankee baseball. C 184