1 EXPLORING INTERGENERATIONAL RESILIENCE IN ANTISOCIAL BEHAVIOR: BAD PARENTS WITH GOOD CHILDREN By BEIDI DONG A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DE GREE OF MASTER OF ARTS UNIVERSITY OF FLORIDA 2012
2 201 2 Beidi Dong
3 To my parents, for the unconditional love and encouragement with which they have always provided me
4 ACKNOWLEDGMENTS I would first like to thank my chair, Dr. Marvin D. Krohn, for his guidance and patience in writing this thesis. Dr. Krohn provided the inspiration for this project and the data that is analyzed. I would like to also thank my co chair, Dr. Chris L. Gibson and the additional member of my committee, Dr. Ronald L. Akers for their assistance in methodological and conceptual issues when I finish this project. I would also like to extend my gratitude to my parents, Guoqiang Dong and Ying Kang. Without their unco nditional love and encouragement, this project would not have been possible. They have been and will forever be my best friends and most loved ones.
5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 7 LIST OF FIGURES ................................ ................................ ................................ .......... 8 LIST OF ABBREVIATIONS ................................ ................................ ............................. 9 ABSTRACT ................................ ................................ ................................ ................... 10 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 12 2 LITERATU RE REVIEW ................................ ................................ .......................... 15 A Brief Overview of Risk and Protective Factors Approach ................................ .... 15 Intergenerational Continuity in Antisocial Behavior ................................ ................. 16 Theoretical Explanation of Intergenerational Resilience of Antisocial Behavior ...... 26 Significance of Current Resear ch ................................ ................................ ........... 30 Research Questions & Hypotheses ................................ ................................ ........ 31 3 METHODS AND ANALYTIC ISSUE ................................ ................................ ....... 33 Data ................................ ................................ ................................ ........................ 33 Measures ................................ ................................ ................................ ................ 36 Dependent Variable ................................ ................................ .......................... 36 Risk Factor ................................ ................................ ................................ ....... 37 Promotive/Protective Factors ................................ ................................ ........... 38 Control Variables ................................ ................................ .............................. 40 Analytic Issue ................................ ................................ ................................ .......... 43 Analytic Strategy ................................ ................................ ................................ ..... 50 4 RESULTS ................................ ................................ ................................ ............... 55 Bivariate Analysis ................................ ................................ ................................ ... 55 The Baseline Models ................................ ................................ .............................. 56 Individual Models ................................ ................................ ................................ .... 57 Comprehensive Models ................................ ................................ .......................... 60 Cumulative Promotive Effects ................................ ................................ ................. 61 5 DISCUSSION AND CON CL USION ................................ ................................ ........ 84 Summ ary of Findings ................................ ................................ .............................. 84
6 Limitations and Future Research ................................ ................................ ............ 90 Conclusion ................................ ................................ ................................ .............. 92 APPENDIX A LIST OF SC ALE ITEMS ................................ ................................ .......................... 94 B DESIDERATA FOR MULTIPLE REGRESSION ................................ ..................... 97 LIST OF REFERENCES ................................ ................................ ............................... 98 BIOGRAPHICAL S KETCH ................................ ................................ .......................... 110
7 LIST OF TABLES Table page 3 1 Descriptive statistics G3 childhood model ................................ ....................... 53 3 2 Descriptive statistics G3 adolescence model ................................ .................. 54 4 1 Bivariate correlations -childhood ................................ ................................ ....... 66 4 2 Bivariate correlations adolescence ................................ ................................ 68 4 3 Baseline model childhood ................................ ................................ ............... 70 4 4 Baseline model adolescence ................................ ................................ .......... 71 4 5 Individual model (G2 affective ties to G3 -childhood) ................................ ........ 72 4 6 Individual model (G2 affective ties to G3 adolescence) ................................ .. 73 4 7 Individual model (G2 consistency of discipline childhood) .............................. 74 4 8 Individual model (G2 consistency of discipline adolescence) ......................... 75 4 9 childhood) .................... 76 4 10 Individual mod adolescence) ............... 77 4 11 Individual model (G2 supervision of G3 childhood) ................................ ......... 78 4 12 Individual model (G2 supervision of G3 adolescence) ................................ .... 79 4 13 childhood) ..................... 80 4 14 adolescence) ................ 81 4 15 Effect of parenting measures on CBCL delinquency outcome childhood ........ 82 4 16 Effect of parenting measures on CBCL delinquency outcome adolescence ... 83
8 LIST OF FIGURES Figure page 3 1 G3 age distribution in RIGS ................................ ................................ ............... 52 4 1 Cumulative promotiv e effects G3 childhood model ................................ ......... 64 4 2 Cumulative promotive effects G3 adolescence model ................................ .... 65
9 LIST OF ABBREVIATION S CBCL CSDD Refers to the Cambridge Study in Delinquency Development DDS Refers to the Dynamic Developmental Systems Model G1 Refers to the first generation in the Rochester Youth Development caretaker, usually biological mother G2 Refers to the focal subject of the Rochester Youth D evelopment Study. G3 Refers to the third generation in the Rochester Youth Development biological child IPA MCMC Refers to the Markov Chain Monte Carlo method OCG Refers to the other primary caregiver of G3 subjects PPVT Refers to the Peabody Picture Vocabulary Test RYDS Refers to the Rochester Youth Development Study RIGS Refers to the Rochester Intergenerational Study SEM Refers to Structural Equation Modeling technique VIF Refers to Variance Inflation Factors
10 Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the R equirements for the Degree of Maste r of Arts EXPLORING INTERGENERATIONAL RESILIENCE IN ANTISO C I AL BEHAVIOR: BAD PARENTS WITH GOOD CHILDREN By Beidi Dong May 2012 Chair: Marvin D. Krohn Cochair: Chris L. Gibson Major: Criminology, Law, and Society Lifetime parental criminality ha s b een associated with offspring criminality An even more interesting finding from recent criminological studies is that parents with a history of involvement in delinquent behavior, drug use and other forms of antisocial behaviors are more likely to have ch ildren who also engage in those behaviors at similar developmental stage s. Existing stu dies, however, also demonstrate that a good portion of childre n born to delinquent parents do not themselves exhibit high levels of problem behaviors. It has been hypothesized that in addition to intergenerational continuity of delinquency, there exists a substantial degree of intergenerational discontinuity that needs to be examined. Despite the plausibility of intergenerational resilience in antisocial behavior, there is surprisingly little research that traces the life course of adjacent generations to examine possible protective factors and casual mechanisms that explain any observed level of intergenerational resilience. In reality, this line of research has great practical
11 implication because identify ing factors that mitigate and buffer the negative effects of having parents with delinquent histories could reduce the chance of occurrence of s behavior within families This thesis explores whether the quality of parent child r e l a t i onship protect s children from antisocial behaviors in the face of parental h istory of adolescent delinquen t behavior. In addition, the current study also examines whether such protection effects from good parent child relationship vary by the age of the child. Using dat a from the Rochester Youth Development Study and its follow up, the Rochester Intergenerational Study (RIGS), it is found that the correlation between parental (G2) adolescent delinquent history and G3 childhood and adolescent delinquent behavior is signif icant but not very large in magnitude. Results from multiple regression analysis indicate that parental (G2) affective ties to G3 is essential in intergenerational resilience of delinquency. G2 affective ties to G3 is statistically si gnificant as a promoti ve factor as well as protective factor, which means G2 affective ties to G3 will not only additively reduce the chance for a child with bad parents to become delinquent, but this protection effect will be more obvious for children with higher level of G2 r isk. In addition, protection from good parent child relationship seems not to vary by the age of G3 subjects. Theoretical and practical implications are discussed.
12 CHAPTER 1 INTRODUCTION There is a well established literature indicating that offspring criminality is associated with lifetime parental criminality (e.g. Farrington et al., 1998; Hagan & Palloni, 1990). An even more interesting findin g from recent criminological studies is that parents with a history of involvement in delinquent behaviors, drug use and other forms of antisocial behaviors are more likely to have children who also engage in those behaviors at similar developmental stage (e.g. Ehrensaft et al., 2003; Farrington, 1993; Farrington, 2009; Kaplan & Liu, 1999; Kim et al., 2009; Thornberry, 2005; Thornberry et al., 2003; Thornberry et al., 2009). Such intergenerational relationship goes beyond traditional concurrent relationship including research on the impact of lifetime parental criminality on offspring delinquency, by examining both or multiple generations at similar developmental stages. Existing studies, however, also demonstrate that estimates of the strength of the rela particularly large in magnitude (e.g. Cairns et al., 1998; Furstenberg et al., 1990). In other words, only a portion of children with parents who have delinquent history wi ll later become delinquents. This implies that, in addition to intergenerational continuity, there also exists a substantial degree of intergenerational discontinuity that needs to be examined. Broadly defined, the concept of intergenerational discontinuit y refers to two patterns of intergenerational transmission of antisocial behavior. That is, some parents who themselves were not antisocial have children committing antisocial behaviors and some antisocial parents have children who are not antisocial (Thor nberry, 2005). To date, criminologists, psychologists and researchers from a variety of disciplines have
13 made substantial progress in understanding how a child may become antisocial even if his or her parents were and are law abiding. For instance, peer de linquency and low self control have been widely recognized as risk factors for individual antisocial behavior and general delinquency after controlling the criminal or delinquent history of parents (Akers, 1998; Gottfredson & Hirschi, 1990). Alternatively there also exist good children with bad parents, a phenomenon that might be called intergenerational resilience (Thornberry, 2005). According to Luthar and adaptation within th e context of significant adversity (2001, p .543) Implied in this concept are two essential components. There must be a substantial degree of exposure increase the probability of a negative outcome, in this case prevalence of antisocial behavior and general delinquency (Reppucci et al., 2002; Richman & Fraser, 2001). The other component implied in this concept is that given such challenging or threatening circumstan ces, positive adaptation is achievable due to certain factors and mechanisms. This line of research has great practical implication because identifying factors that mitigate or buffer the negative effects of having parents with delinquent histories could r educe the chance of occurrence of similar behaviors in their offspring, plausibility of intergenerational resilience in antisocial behavior, there is surprisingly little research that traces the life course of adjacent generations to examine the casual mechanisms that explain any observed level of intergenerational resilience. Given the existing research gap, this thesis investigates theoretically and empirically driven
14 p romotive/protective factors and causal mechanisms of intergenerational resilience in antisocial behavior. Using prospective longitudinal data from the Rochester Youth Development Study (RYDS), and its follow up, the Rochester Intergenerational Study (RIGS ), this thesis addresses two main research questions. First and foremost, does the quality of parent child relationship protect children from delinquent behaviors in the face of parental history of adolescent delinquent behavior? Most parents with a delinq uent history in the age crime curve suggests, also gradually desist from delinquency when they become older (Hirschi & Gottfredson, 1983; Gottfredson & Hirschi, 199 0). Accordingly, it is entirely possible for at risk adolescents who become parents eventually to develop adulthood positive features to protect their offspring from antisocial behavior at corresponding stages of development. To be more explicit, children can obtain protection from a healthy relationship with their parents against antisocial behavior even if their parents used to be juvenile delinquents (Hirschi, 1969; Kubrin et al., 2009). This question is analyzed using multivariate ordinary least regress ion models. The second objective is to see whether such protection effects from good parent child relationship vary by the age of the child. Existing studies have suggested that the influence of family, particularly parents, decreases with the aging of the next generation (e.g. Thornberry, 1987; Thornberry & Krohn, 2001; Gottfredson & Hirschi, 1990; Farrington et al., 2008).
15 CHAPTER 2 LITERATURE REVIEW A Brief Overview of Risk and Protective Factors Approach The risk and protective factors approach to researching the origins of antisocial behavior and general delinquency has over the last couple of decades become a dominant discourse in juvenile justice and now exerts a significant influence over policy y application of this epidemiological approach in criminology dates back to the seminal Cambridge Study of Delinquent Development in the 1970s (West & Farrington, 1973). Focusing on identifying the correlates of antisocial behavior and determining how to r educe the risk of those correlates, this approach has been widely adopted in criminological studies and there have been elaborations on how the key concepts in this approach, risk, promotion and protection could be articulated One way in which protectiv e factors have been conceptualized is as the opposite end of the risk factor continuum, simply decreasing the probability of negative outcomes (Hawkins et al., 1992; Labouvie & McGee, 1986; Loeber & Farrington, 2000; Stouthamer Loeber et al., 1993; 2002). However, treating protective factors simply as positive elements negatively associated with antisocial behavior and general delinquency blur s the conceptual and empirical distinction between risk and protection (Rutter, 1985; Fraser et al., 2004) I f prote ction is plainly considered the opposite end of the risk factor continuum, th is concept is of little practical significance (Hawkins et al., 1992). T denote this conceptualization. Rutter (1985) suggests that we not equate promotive factors with protective factors as was the case in a number of early studies ( Hawkins et al., 1992; Loeber &
16 Farrington, 2000; Stouthamer Loeber et al., 1993; 2002 ). Protective factors instead are defined environmental hazard that predisposes to a maladaptive outcome (Rutter, 1985, p.600) The implication of this definition is that factors should be considered protective if they differentiate between persons who face a comparatively high risk of disorder, but, in one case become deviant, and in the other, remain healthy. An accumulation of risk places individuals in a position where antisocial behaviors and delinquency are probabl e. Once the risk is established, subsequent protective factors may interact with the accumulation of risk to reduce that probability. Although much of the early research on promotive/ protective factors focused only on one or two factors and investigated th e impact that they had on decreasing the probability of antisocial behavior among at risk individuals (Fraser et al., 2004), a growing body of studies ha ve shown that it is the cumulative effects of promotive/protective factors from multiple domains that w ork to increase resiliency ( e.g. Smith et al. 1995; Luthar et al., 2000) Intergenerational C ontinuity in Antisocial Behavior T he proposition that children often resemble their parents is not likely to be a surprise Biological, particularly genetic, infl uences of parents on their children are becoming better understood (e.g. Ferguson, 2009; Moffitt, 2005; Rhee & Waldman, 2002; Walters, 1992) The importance of social and educational factors such as socialization and parenting has also been gradually uncovered In addition, the share of common physical, social and cultural environments by parents and children who live together may further contribute to the of (Brooks Gunn et al., 1993; Bradley & Corwyn, 2001). Most importantly res earchers from a variety of fields have investigated the specific processes whereby particular patterns of behaviors directly or
17 indirectly are transmitted across generations (e.g. Thornberry et al., 2009; Smith & Farrington, 2004; Capaldi et al., 2003; Kap lan et al., 2000; Patterson 1998; Serbin et al., 1998). While this thesis focuses on intergenerational resilience of antisocial behaviors and promotive/ protective factors, it is important and necessary to first review existing literature on how and why an tisocial behaviors are transmitted across generations. Thornberry (2009) and researchers from related fields (e.g. Cairns et al., 1998; Patterson, 1998; Serbin & Karp, 2004; Smith & Farrington, 2004 ) suggest three core elements of a standard intergenera tional study. First and foremost, different generations included in an intergenerational study should be studied at roughly the same age or developmental stage Thornberry (2009) argued that this intergenerational relationship has greater potential to full y capture the underlying, cross generational mechanisms than traditional longitudinal, concurrent research. Second, the research design should be prospective. For a two generation study of intergenerational continuity, for instance, it is important to have prospective data on both lvement in antisocial behavior. Third to reduce bias associated with reporting, having multisource, multimeth od repeated measures is ideal. (G2) current behavior influences their views of the (G1) (G3) behaviors. For example, highly antisocial parents may project their anti behavior, inflating the reports Strictly speaking, very few of the existing studies on interg enerational continuity of antisocial behavior have met all three aforementioned criteria. Relying solely on generation 2 (G2) as the source of data is one fundamental methodological flaw o f
18 cross generational study (e.g. Zuravin et al., 1996). Important co behaviors are measured retrospectively and estimates of the magnitude of the relationship across generations may have been distorted by reliance on a single reporter (Kaufman & Zigler, 1992; Velleman, 1992 ). As a result, the last two core elements of a standard intergenerational study have been compromised Not having a large enough sample could be another major problem, especially associated with earlier cross generational studies (Thornberry et al., 2003). With regard to rigorous empiri cal studies of intergenerational continuity of antisocial behavior, two specific issues have been extensively investigated. One is to establish the level of intergenerational continuity in diversified forms of antisocial behavior. As Serbin and Karp (2004) mentioned, the increased risk that may be behavior pattern) and general (in terms of broad elevation of risk for numerous negative outcomes like school failure, early pare nthood and so on). Thus far, a majority of studies have reported some significant intergenerational link ages for antisocial behavior, while others do not (Thornberry et al., 2009). Huesmann with colleagues (1984; 2003) studied 600 children from age 8 to mid adulthood, plus their parents and children. They found that intergenerational continuity of aggression was even greater than individual continuity across the life course when successive g enerations were assessed at approximately the same age. Furstenberg et al. (1990) followed a sample of 404 African American women from disadvantaged Baltimore families and their children over 20 years. They found an elevated risk of teenage parenthood in t he next generation born to teenage mothers, compared to non teenage mothers. Nevertheless,
19 the majority of children of teenage mothers did not continue to become teenage parents. Offspring of teenage mothers who did become teenage parents were much more li kely to live in poverty and have poor academic performance in school than other offspring who did not repeat the adolescent parenthood of their mothers. Chassin et al. (1998) conducted a prospective multigenerational study of 214 families, including grandp arents (G1), parents (G2) and adolescent offspring (G3). They found that there existed intergenerational linkage of smoking behaviors across the three generations. Specifically, parenting behaviors were predictive of adolescent smoking, both directly and i ndirectly by raising the likelihood of affiliating with smoking peers Using data in three successive generations, Thornberry, Krohn and Freeman Gallant (2006) reported drug use by the child, but the effect is only for their so called supervisory parents who had contact with the child at least once a week. Similarly, Bailey et al. (2006) also examined three successive generations and found that substance use for grandpar ents problem behavior at ages 13 14. An important multigenerational study of antisocial behavior initiated in the Uni ted Kingdom worth detailed mentioning is the Cambridge Study in Delinquency Development (CSDD), a prospective longitudinal survey of the development of antisocial behavior and offending in 411 males led by David Farrington. Research subjects were first con tacted in 1961 1962 around the age of 8 9 from a wo rking class inner city area of S outh London. Thus far, Farrington with his research team has accumulated a wealth of information on the study males and their life courses, their
20 parents and also their offs pring (Bijleveld & Farrington, 2009). Early findings reported by Farrington et al. (1975) and West and Farrington (1977) showed that when the study males were aged 20 on average, 48% with convicted fathers were convicted compared with only 19% of those wit h unconvicted fathers. Meanwhile, 54% of the study males with convicted mothers were themselves convicted compared with only 23% of those with unconvicted mothers. A more recent replication study done by Farrington, Coid and Murray (2009) extended earlier analyses to the study males aged 50, investigating the intergenerational transmission of offending across three generations. Not surprisingly, they found that there was a similar degree of intergenerational transmission of convictions from G1 males to G2 m ales (odds ratio =3.5) and from G2 males to G3 males (odds ratio=3.2), whereas the degree of intergenerational transmission from G1 females to G2 males (odds ratio=2.3) and from G2 males to G3 females (odds ratio=2.0) was less. T here was little evidence of intergenerational transmission of offending from G1 to G3, except from grandmothers to granddaughters (odds ratio=2.5). Interestingly, they also found that the study males (G2) who have two criminal parents were not more likely to be convicted than those with only one criminal parent, a conclusion which needs further research to verify (see also Kim et al., 2009 for a different result). Similarly, using conviction data on five generations that span an even longer period of time, from the year 1882 to 2007 in the Netherlands, Bijleveld and Wijkman (2009) also found that parental convictions would increase the risk of offspring convictions, although the risk increase is not extremely high and parental delinquency before birth is found to be irrelevant (see a l so Farrington et al., 1975 for a different
21 result). The risk increases more, and more consistently, when a narrower subset of of fenses (more serious offens es) was included. Other research also provides evidence of a robust intergenerational relationship ( e.g. McCord, 1991; Kaplan & Liu, 1999 ; Farrington, 2009; Kim et al., 2009). In contrast, other studies have either found no significant correlation between and child s antisocial behavior or partial support for intergenerational continuity under s tringent conditions. For instance, Cairns et al. (1998) failed to find inter generation al transmission of aggressive behavior. T he correlations between aggressive behaviors of the mothers when they were children and aggressive behaviors of their children in early school years were unreliable and only modest. Hops et al. (2003) found that G2 s adolescent externalizing behavior was not significantly correlated with G3 s childhood externalizing behavior. The seemingly obvious fact that adjacent generations tend to behave in similar fashion with respect to antisocial behavior is not universally supported by empirical studies (Thornberry et al., 2003), which, in turn, has stimulated research interests in intergenerational discontinuity and potential protective fac tors that buffer existing parental risk against their child. The current study thus pursue d the discontinuity pathway. A n equally important if not more important, issue regarding intergenerational continuity of antisocial behavior is to discover the multi faceted processes involved in the transmission of risk. In effect, identifying causal mechanisms explaining intergenerational continuity will give important hints on possible protective factors that are central to intergenerational resilience study.
2 2 Of th e causal mechanisms considered, family processes, particularly parenting behaviors, are perhaps the most important ones. Through examining antisocial behaviors of parents and children at ages 9 11 and 12 14 respectively, Kaplan and Liu (1999) reported a st at ages 35 39 significantly mediated the estimate of intergenerational continuity. Ehrensaft with colleagues (2003) confirmed both intergenerational continuity of antisocial behavior a nd the mediating role of parenting. In a relatively short term longitudinal study (15 month) with 126 African American and Hispanic at risk boys who had an older brother with a recent juvenile court conviction, they found that early onset conduct disorder behavior. High level of parent child conflict and lack of parental involvement and monitor adolescent ant isocial behavior is both directly and indirectly, via poor parenting, linked While parenting is certainly a central mediator of intergenerational continuity of antisocial behavior as existing literature shows, relevant factors that directly or indirectly precede parenting are worth considering as well. From both a relatively narrow point of view rooted in interactional theory of delinquency (Thornberry, 1987; Thornberry & Krohn, 2001; 2005) and a wider life course perspective (Elder, 1997) or the dynamic developmental systems (DDS) model (Capaldi et al., 2008), involvement in antisocial behavior has detrimental consequences for individuals that would ultimately compromise life opportunities of h is or her offspring. Additionally, there is a well established literature showing that stressful life events, depression and other forms of
23 positive developmen t in child ren (e.g. Belsky, 1984, 1993 ; Hops et al., 1987; Patterson et al., 1989; Sny der 1991; Capaldi & Patterson, 1991; Downey & Coyne, 1993; Kotch et al., 1999; Pears & Capaldi, 2001; Tolan et al., 2006). However, relatively few empirical studies have considere d links between depression and other life s tressors, combined with gendered effects, as potential contributors to parenting behaviors and as potential mediators of intergenerational continuity of antisocial behavior. T he Rochester Youth Development Study ( RYDS) (Thornberry et al., 2003) with its follow up Rochester Intergenerational Study (RIGS) is one of the very few studies that satisfy all three defining criteria of a standard intergenerational study Using data from the RYGS and RIGS, Thornberry with c olleagues (2003) examined gendered effects of financial stress and parenting in the intergenerational transmission of antisocial behavior. Results from structural equation models show that and positively related adolescent involvement in delinquency is related to a less effective parenting style, behavior. Interestingly, g to family poverty increases the chances that that G2 will experience higher level of financial stress during their early 20s, which, in turn, decreases effective parenting styles. However, for G2 mothers, their adolescent involvement in delinquency is not ial behavior during childhood. Instead,
24 parenting of G3. Additionally, for female s, economic disadvantage is also central to intergenerational transmission of antisocial behavior. As Thornberry et al. concluded, plays a more powerful mediating ro le in the model for mothers (2003, p.181) However, it is worth noting that Moffitt et al. (2001) reported that the risk processes for childhood and adolescent antisocial behavior are quite similar fo r boys and girls ( see also Verona & Sachs Ericsson, 200 Accordingly, taking into account of b simultaneously would be of great research significance. Given the substantial sample size it requires, few empirical stu dies have been able to do so. Using the same data and statistical techniques for analysis, Thornberry (2005) reported that, for G2 mother and their families, G2 adolescent drug use is only indirectly related to G3 delinquency, reported by their teachers, through the effects of G2 positive parenting. However, adolescent drug use increases the likelihood of G2 experiencing stress during transition period, which, in turn, has a significant impact on G2 parenting. There is also a direct association between ado lescent drug use and positive parenting style. In addition, G1 positive parenting reduces G2 adolescent drug use and is positively related to G2 positive parenting. On the contrary, for G2 supervisory fathers, the negative, hostile aspects of parenting are significantly related to G3 delinquency. Stressful life events of G2 supervisory fathers are also directly linked to G3 delinquency
25 beyond an indirect path via G2 negative parenting. These parental characteristics are related to their earlier development, especially adolescent drug use. Thornberry, Freeman Gallant and Lovegrove (2009 b ) specifically examined the impact of two parental stressors on the intergenerational transmission of antisocial behavior using data from the RYDS Again, for G2 mothers, ther e is a significant correlation between their own engagement in drug use and delinquency during conclusion can be drawn for supervisory fathers, but not for non supervisory fa thers. More specifically, for both G2 mothers and G2 supervisory fathers, adolescent antisocial behavior and a younger age at first birth significantly increase the level of depressive symptoms, which, in turn, significantly reduce parental attachment to t he child and proposed parental stressor negative life event s is lower than expected by theoretical models. In a summary piece for the aforementioned series of studies on interg enerational continuity of antisocial behavior led by Thornberry and his research team, Thornberry antisocial behavior across adjacent generations. The social environment either directly or in interaction with genetic risk, plays a vital role in our understanding of the origins and course of antisocial behavior (p.219 320) Therefore, the presence of some non significant or mixed findings should not be taken to mean that current research of intergenerational transmission of antisocial behavior is invalid or meaningless. To a significant extent, the strength of the parent child relationship in antisocial behavior, and
26 the degree to which mediators explain the intergenerati onal link, depend on the particular outcome variables considered, the particular environment or family related variables used and the age ranges for the parent and the child at which measurements are taken and may also reflect period effects if the dataset covers a long time span (Thornberry, Freeman Gallant & Lovegrove, 2009 a ; Bijleveld & Farrington, 2009). T hese same words may be suitable for research on intergenerational resilience of antisocial behavior as well. Theoretical Explanation of Intergeneratio nal Resilience of Antisocial Behavior T he explanation of intergenerational resilience of antisocial behavior is guided by criminological theories. Since we are interested in how the quality of parent child relationship may promote and/or protect children from delinquency in the face of parental history of adolescent delinquent behavior, the broad category of social control theories fits our research purpose (e.g. Hirschi, 1969; Reckless, 1961). One well established developmental and life course theory of antisocial behavior within the general framework of social control perspective is the interactional theory of delinquency first proposed by Thornberr y (1987) and subsequently expanded by Thornberry and Krohn (2001; 2005). Thornberry first articulated his interactional theory of delinquency by pointing out three major limitations of extant theories of delinquency: 1) relying on unidirectional rather t han reciprocal causal structure, 2) failing to tackle the issue of delinquency developmentally, specifying causal models for only a narrow age range, and 3) assuming uniform casual effects through the social structure and ignoring the sources of initial va riation in both delinquency and its presumed causes. Thornberry offers his interactional model as an alternative attending to these limitations U nlike classical
27 control theories proposed by Hirschi and others (e.g. Nye, 1958; Reiss, 1951 ), Thornberry s in teractional model does not assume that the attenuation of controls directly leads to delinquency or the freedom resulting from weakened bonds to be channeled into delinquency, especially serious prolonged delinquency, requires an interactive setting in which delinquency is learned, performed and reinforced ( Thornberry, 1987, p.865). Adopting a theoretical elaboration approach Thornberry included three core measures from social control theory attachment to parents commitment to school and belief in con ventional values and two core measures from social learning theory association with delinquent peers and adopting delinquent values and one ultimate outcome variable engaging in delinquent behavior to illustrate his initial version of an interactional mod el. Thornberry with colleagues (1991; 1994) tested the interactional model using data from early waves of the RYDS and concluded that is part and parcel of a dynamic social process (1994, p.75), and causal influences vary at differen t developmental stages and many causal relationship s are reciprocal (1991, p.29). Thornberry and Krohn (2001; 2005) further elaborated on the interactional model of delinquency by emphasizing the notion of equifinality, meaning that many potential combinations can produce the same result (Cicchetti & Rogosch, 1996), and incorporating explanations of both early childhood antisocial behavior and criminal behaviors after the teenage year. According to them, for example, the mix of causes that generate delinquency during the preschool years differs, to some extent, from the mix of causes that generate delinquency during childhood, adolescence and late adolescence/early adulthood. Relatedly, the success or failure with which previous
28 developmental stages have been traversed also influences the occurrence of antisocial behavior. While there are multiple causes for antisocial behavior and general delinquency, not all causal factors need to be activated to produce the outcome. The causal force varies across p eople according to the number of casual factors and the of the total magnitude of all relevant causal forces (Osgood, 2005, p.199) Additionally, the presence or absence of offsetting variables is also crucial for the occurrence of delinquency. As the magnitude of the causal force increases and the magnitude of general delinquency becomes more likely, and the severity of antisocial behaviors will also increase (Thornberry & Krohn, 2005). In effect, this path of thinking is consistent with the risk and protective factors approach that is adopted for the current research project. Thornberry and Krohn (2005) explicitly discussed the initiation of antisocial behaviors in four broad developmental stages: the toddler/preschool years, childhood, adolescence and late adolescence/early adulthood. Rather than having sharp boundaries, these developmental stages are viewed as regions of gradual, continuous process of human development. Specifically, the early onset of a small portion of the population who initiates antisocial behaviors during toddlerhood and the preschool years can be explain ed by the combination and interaction of neuropsychological deficit and difficult temperament (e.g. poor emotion regulation, impulsiveness and negative emotionality), ineffective parenting (e.g. physical punishment and low affective ties), and position in the social structure (e.g. poverty, unemployment and disorganized
29 neighborhood). C s and likely to be causally interwoven. As prior research has shown, young children with negative temperamen tal qualities are more subject to parental hostility and physical punishment (e.g. Rutter & Quinton, 1984; Lee & Bates, 1985), which, in turn, leads to maladaptive and uncontrolled responses in the child (e.g. Moffitt, 1993; Belsky et al., 1996). Meanwhile empirical studies have also shown that structural adversity increase s parental stress and reduce their social capital, which, in turn, leads to ineffective parenting (Belsky et al., 1996; Conger et al., 1994; Patterson et al., 1992). Childhood onset of a ntisocial behavior is then particularly associated with family factors. Structural adversity increases stressors such as parental depression and financial worries, impeding effective parenting, which, in turn, has a strong impact on antisocial behavior (e. g. Jang & Smith, 1997). Moreover, both structural adversity and ineffective parenting opportunities. Accordingly, taking into account both empirical findings from research o n intergenerational continuities of antisocial behavior and theoretical explanation on how antisocial behavior initiates in pre school and childhood years the following parental (G2) factors are proposed to have promotive and/or protective effe antisocial behavior: parental attachment to child, parental supervision, parental consistency of school activities. A s the child ages, greater effects from school and peers should be taken into account. Adolescents who initiate antisocial behaviors when they are about age 12 to age 16 are unlikely to have been exposed to the more extreme and interwoven casual
30 forces described up to this point. As Conger (1991) sugge sted, one of the major developmental tasks of adolescence is to establish age appropriate autonomy, which relies on the processes of both separation and continued connectedness from parental authority. To some extent, peers replace parents, or at least are added to parents, as another major source of rewards and approval of behaviors (Gecas & Seff, 1990). Thus, youngsters who were buffered at earlier ages by strong pro social bonds are in a particularly precarious position during this stage of their lives There exist greater difficulties for their parents to provide alternative activities that could keep their children from problematic influences as in childhood years (Thornberry & Krohn, 2005). Thus, we expect that protection effects of good parent child r elationship would become weaker when the child grows up. Significance of Current Research The major goal of this study is to explore intergenerational resilience in antisocial behavior and general delinquency with a specific focus on identifying promotive and protective factors. While prior work on intergenerational transmission of antisocial behavior is meaningful and valuable, it lacks in several significant areas. In congruence with the primary objective of this thesis, no prior literature exists that addresses specifically the issue of intergenerational resilience of antisocial behavior and delinquency. This is quite surprising when we consider the accumulated evidence in the field for which estimates of the strength of the non concurrent relationship There are, thus, a significant number of good kids with parents with delinquent histories. hood protective factors against their own adolescent risk regarding antisocial behaviors of
31 empirical studies that have been done by Thornberry with colleagues mentioned earlier on intergenerational transmission of risk. Their main research focus is on mediating relationships using structural equation modeling (SEM) techniques, while this thesis research focuses on moderation. A preliminary list of promotive/ protective factors has been suggested for further research. Second, previous empirical studies and criminological theories, for instance, the interactional theory of delinquency, have indic ated that the impact of family, particularly parenting behaviors, on antisocial behavior would decrease as the child grows older. The same logic applies to the protection side. As a preliminary study, this thesis investigates whether the sets of promotive/ protective factors as well as the moderating mechanism differ for G3 childhood versus adolescent years. The present research is the first to take into account the protection side from both an intergenerational and developmental perspective. Cumulative effe cts have also been reported. Research Questions & Hypotheses R Q1 Does good parent child relationship protect children from delinquent behaviors in the face of parental history of adolescent delinquent behavior? Hypothesis Good parent child relatio nship ( parental attachment to child, parental supervision, parental consistency of discipline parental involvement in significant promotive and/or protective own adolescence risk.
32 R Q2 Does the group of promotive and/or protective factors ( as well as the moderating mechanism ) against parent al (G2) adolescent risk differ in G3 childhood versus adolescent years? Hypothesis It is hypothesized that the number of significant promotive and protective factors would decrease when the child grows up, net of effects of important control variables.
33 CHAPTER 3 METHODS AND ANALYTIC ISSUE Data Data for the current study were obtained from the Rochester Youth Development Study (RYDS), and its follow up, the Rochester Intergenerational Study (RIGS), which, together, constitute a comprehensive investigation of both pro social an d anti social development across the human life course (Thornberry et al., 2003). The RYDS has followed a panel of juveniles from their early teenager years (age of 14) till age 31, and 14 waves of interviews have been completed. The RYDS began in 1988 wi th an original sample of 1000 seventh and eighth grade students in the public schools of Rochester, New York, referred to as Generation 2 (G2). Since the base rates for serious delinquency and drug use are relatively low (Elliott, Huizinga, & Menard, 1989; Wolfgang, Thornberry, & Figlio, 1987), in creating the original sample, youth at high risk were oversampled in the Rochester study. Specifically, males were oversampled (75% versus 25%) because they are more likely than females to commit delinquent acts. Students from high crime rate areas of the city were also oversampled based on the assumption that living in such areas is a significant risk factor for juvenile delinquency. In order to identify those high crime rate areas, each census tract in Rochester was population arrested by the Rochester police in 1986. The initial sample was predominantly composed of minority (68% African American, 17% Hispanic, and 15% White) and males (77%). In Phase 1 of the RYDS, the students (G2) and their primary caretakers (most often the biological mother), referred to as Generation 1 (G1), were interviewed nine
34 and eight times respectively at 6 month intervals. Phase 1 covered the adolesce nt years of G2 subjects, ages 14 18. In Phase 2, G2 subjects with their parents were interviewed at three annual intervals, at ages 20 22. Multiple measurements and instruments have been adopted to study a wide range of topics including neighborhood charac teristics, family structure, family relationships, peer relationships, social networks and support systems and serious and minor problem behaviors. Two additional interviews of G2 al life trajectory such as education level, job and family formation was collected. In addition to interview data, the RYDS also collected child maltreatment information from the Department of Social Services and official arrest data from the Rochester pol ice department. The retention rates compare favorably to other panel studies of anti social behaviors. At Wave 12, 85% (846) of the initial 1,000 subjects were re interviewed, and the completion rate for parent interviews was 83%. Eighty percent of G2 subj ects were re interviewed at Waves 13 and 14. The Rochester Intergenerational Study (RIGS), which launched in 1999, focused on the oldest biological children of the original RYDS subjects. The children, referred to as Generation 3 (G3), enter the RIGS stud y as they turn 2 years old. Interviews are conducted annually with the original RYDS subjects, another primary caretaker of the child and all children who are 8 years old and above (n=371 in year 1, they ranged in age from 2 to 13, with an average age of 6 years and n=479 in year 10, they ranged in parenting behaviors and peer friend
35 relationship and antisocial behavior. Additional data have also been gathered from other agencies such as the schools and social services (Thornberry, 2009). The current an alysis uses data from Wave 2 9 of the RYDS, and Research Year 1 2 and Year 7 8 data of the RIGS on 336 G3 subjects. Data from multiple waves are used to assure the proper time order of variables, accordingly retaining the integrity of sk measures are taken at Wave 2 9 of the RYDS (corresponding year 1988 adolescence outcome measures are taken at Research Year 2 and Year 8 of the RIGS (corresponding year 2000, 2006). Since promotive/protective factors are to be measured after existing risk and before delinquency outcome, G2 adulthood promotive and protective measures are taken from Research Year 1 of the RIGS (corresponding year 1999) for G3 childhood outcome and Year 7 (corresponding year 2005) for G3 adolescent outcome. The analysis sample (N=336 for G3 childhood model and N=322 for G3 adolescence model) for the current investigation was almost evenly composed of male (51%) and female (49%) subjects. The sample was predominantly African American (77%) and had a mean age of 6.13 years old at Research Ye ar 1 of the RIGS. As shown in Figure 3 1 the youngest participant in the study sample was around 3 years of years old, with a minimum of 14.1 years of age and a maximum of 23.6 years of age. Approximately 29% had less than a high school diploma, the majority, 60% had a high school diploma and 11% had some education beyond a high school diploma.
36 Measures Depende nt Variable General delinquency Both G3 childhood and adolescence delinquent behaviors Checklist (CBCL), which is one of the most commonly used standardized measures for evaluat ing behavioral and emotional problems in children between the ages of 2 and 3, and between 4 and 18 (Achenbach et al., 1991) In the RIGS, the CBCL was administered to both G2 subjects and the other primary caregivers of G3 subjects. If G2 subjects were f emale (biological mothers), their own interview data were used, whereas for male G2 subjects (biological fathers), the OCG interview data have been used. Previous studies using the same datasets (Thornberry et al., 2009a; 2009b) have behavior and only a small proportion of G2 male subjects live with their oldest biological child and barely can they really know whether G3 subjects have committed delinquency. The primary careg iver was asked how often it is true that the child participates in different types of delinquent behaviors. Responses were indicated on a 3 point ordinal typically avera ged the responses for the include items and a score is calculated only if 80% or more of the scale items are non missing (e.g. Thornberry et al., 2003; Thornberry et al., 2009a; 2009b). Of central importance in this thesis study is the delinquency behavior component of the CBCL. An 18 item general delinquency scale after misbehaving, hangs around with others who get in trouble, vandalizes, physica lly attacks people and etc ( Appendix A for complete list of items) Most items in the scale
37 are comparable to moderate and minor delinquency items used in the RYDS Delinquency Indices, which have been included as risk factor for the current study. The scale yielded a reliability of 0.80 in G3 childhood model and a reliability of 0.86 in G3 adolescence model. Table 3 1 and 3 2 show the average CBCL Delinquency score for G3 childhood to be 0.184 (range 0 2), with a minimum of 0 and a maximum of 0.889 and for G3 adolescence to be 0.187 (range 0 2), with a minimum of 0 and a maximum Risk Factor Parental (G2) adolescent risk The measure of the parental (G2) adolescent risk is based on G2 self repor ted data collected in their adolescent interviews from phase 1 of the RYDS. At 6 month interval, G2 subjects responded to a self reported delinquenc y index containing 32 items ( Appendix A) covering a range of delinquent behaviors from minor offenses such a s status offenses and petty theft, to more serious crimes like robbery, burglary and aggravated assault. G2 subjects were asked if they had committed each offense since the last interview, in other words, during the past 6 months and, if they had, how many times. A cumulative incidence score of involvement in delinquent behaviors between Wave 2 and 9 of the RYDS covering the age span of 14 to 18 was used in the current investigation. Since the original scale is positively skewed (childhood: Skewness=2.51; a dolescence: Skewness=2.62), a log transformation has been applied. Table 3 2 shows for childhood model, the mean for G2 adolescent risk after log transformation is 2.89 with Skewness=0.09, and for adolescence model, the mean is 2.79 with Skewness=0.16. Pre vious studies have demonstrated that the same cumulative delinquency incidence score has strong reliability and validity (Thornberry & Krohn, 2003; Thornberry et al., 2009).
38 Promotive/Protective Factors tachment to G3 are measured by a 10 originally designed to measure the extent and severity of parent child relationship problems as perceived and reported by the parent and grand parent (Hudson, 1996). The IPA has been found to be reliable and valid in empirical studies assessing parent child relationship (e.g. Taylor et al., 2007; Thornberry et al., 2003). G2 was asked, for Appendix A for complete list of items). Responses were indicated on a 5 5). Items have been averaged to provide the mean score, and the reliability of the scale is 0.77 (Thornberry et al., 2009). The mean score for G3 childhood sample is 4.42 (range 1 5) and for adolescence sample is 4.48, with higher scores indicating higher G2 consistency of discipline A four consistency in disciplining G3. G2 was asked once they have decided on a punishment, discipline the child es were indicated on a 5 (5). Items have been averaged to provide the mean score, and the reliability of the scale is 0.69. The mean score for G3 childhood sample is 3.50 (range 1 5) and for
39 adolescence sample is 3.82. All four items were reverse coded, thus higher scores A seven item scale is used to measure asked how often they read to child, talk to child about what he/she did during the day ( Appendix A for complete list of items). A similar five involvement play on a 5 po (5). The mean score for G3 childhood sample is 3.79 (range 1 5) and for adolescence The reliability of the scale is 0.73. G2 supervision of G3. The supervision of G2 on G3 is measured by a 3 item scale. G2 was asked during the times when they were taking charge of the child, how point scale from G3 childhood sample is 4.77 (range 1 5) and for adolesc ence sample is 4.66, with scale is 0.70. The closeness of G2 as a parent towards was asked since last interview,
40 (1) (0) Items have been summed to create a scale between 0 and 4. The summed score for G3 childhood sample is 2.23 and for adolescence sample is 2.02. Th e reliability of the scale is 0.74. Control Variables G2 concurrent delinquency Empirical studies have shown that active delinquent parents have concurrent, detrimental impact on the well being of their offspring (e.g. Garnier & Stein, 2002; Loeber & St outhamer Loeber, 1986). Since our research focus here is on non crime and delinquency needs to be controlled. A 19 item self reported delinquency index similar to the one used to measure parental (G 2) adole scent risk is adopted here ( Appendix A for complete list of items). As mentioned earlier, data from Research Year 1 and Year 6 of RIGS will be used here. Because the original scale is positively skewed (childhood: Skewness=4.47; adolescence: Skewne ss=5.02), a log transformation has been applied. The mean for childhood model is 1.34 with Skewness=1.40, and for adolescence model, the mean is 0.81 with Skewness=2.11. G3 cognitive competence Numerous studies have demonstrated that cognitive skills and antisocial behavior are intertwined (e.g. Bellair & McNulty, 2010; Gordon, 1987; Hirschi & Hindelang, 1977; Lynam et al., 1993; Moffitt, 1990). Standardized scores from the Peabody Picture Vocab ulary Test PPVT III has been adopted as a proxy to measure cognitive competence for G3. The PPVT III has a median reliability
41 use with minorities such as African Ame rican people (Washington & Craig, 1999). The mean score for the childhood sample is 89.76 with a standard deviation of 12.27, and the mean score for the adolescence sample is 90.02 with a standard deviation of 12.08. Race/ethnicity of G2 Research has sho wn that child rearing practices are different among ethnic groups. For instance, corporal punishment is more common among African Americans than whites or Hispanics (e.g. Berlin et al., 2009). Information on the race/ethnicity of G2 was also recorded as p art of the demographic profile of the Rochester Intergenerational Study (RIGS) and several indicator variables were created to include race/ethnicity of G2 in the model. The indicator variables were African American (1 if yes, 0 if otherwise), Hispanic (1 if yes, 0 if otherwise) and white (1 if yes, 0 if otherwise). As mentioned above, the majority of the sample was African American in both childhood and adolescence analysis sample. Education level of G2 Low parental education level has been linked to bot h poor child rearing practices and antisocial behavior in children (e.g. Smith & Brooks Gunn, 1997; Cote et al., 2006). Information regarding the highest education level of G2 was obtained from G2 interviews at Research Year 1 (1999) of the RIGS. Indicator variables were created for the maximum level of education being less than a high school diploma (1 if yes, 0 if otherwise), a high school diploma (1 if yes, 0 if otherwise), and some education beyond high school (1 if yes, 0 if otherwise). The majority of G2 subjects had a high school diploma. Research has shown that younger parents tend to have poorer child rearing skills such as more frequent use of corporal punishment than parents who bear child at a later age ( Berlin et al., 20 09; Straus, 1994 ). In addition,
42 precocious life transition such as being a parent before the normatively expected age has also been demonstrated as risk factors for deviance of premature parents as well as for their offspring (Krohn, Lizotte & Perez, 1997) 19.04 years old with a standard deviation of 2.04. Gender of G3 Empirical Studies on intergenerational continuity of antisocial behavior mentioned earlier have outlined the importance of taking G3 gender into consideration. It is highly possible that gendered effects also exist throughout intergenerational resilience male (51.83%) and female (48.17%) subjects. Very similarly, the analysis sample for adolescence was evenly compo sed of male (51.41%) and female (48.59%) subjects. G2 as supervisory parent Previous research (Thornberry et al., 2009b) has indicated that only about 20 30% of the fathers (G2 males) live with the child (G3 subjects) and those who do not have varying de grees of contact with their child (Smith et al., 2005). Research (Thornberry, 2005; Thornberry et al., 2006) also shows that substantive differences exist, in the level of intergenerational continuity, between fathers who live with or have frequent contact with the child and those who do not. Accordingly, a dummy variable was created to indicate whether G2 included in the current
43 either lived with the child or, on average supervised the child at least once a month, supervisory parents. G3 temperament G3 temperament was included as a proxy to control potential genetic influence across generations. Specifically, two scales have been created to mea sure G3 anger/frustration and impulsivity. Questions like child gets angry when told he/she has to go to bed and child usually rushed into an activity without th inking about it were asked ( Appendix A for complete list of items). Responses were indicated on a 4 reliability of the anger/frustration scale is 0.77 and the rel iability of the impulsivity is 0.60. The mean score for G3 childhood sample is 2.14 (range 1 4) for anger/frustration and 2.87 (range 1 4) for impulsivity, and for adolescence sample is 2.12 for anger/frustration and 2.88 for impulsivity. A higher score in dicates higher level of anger/frustration and impulsivity. Analytic Issue The present study employs multiple regression (MR) analysis, which has been widely recognized as a general data analytic system (e.g. Cohen, 1968; McClendon, 1994), to assess the pro motive and protective effects of parenting variables on intergenerational resilience of delinquent behavior. The idea behind the MR model is to relate a set of independent variables to an outcome variable, for purposes of explanation and/or prediction, wit h an equation linear in its parameters. Kelley and Maxwell (2010) have listed specific desiderata for applied studies that utilized multiple regression technique. Desiderata that are relevant to the analys is part are addressed here ( Appendix B for complete list of the desiderata).
44 Missing data All variables that were included for the analysis were examined for missing data. Missing values in the dataset appeared to be missing at random (MAR). There was no sign that the missingness depends on an outside v ariable not in the model or depends on the variable itself, though it is practically impossible to test whether the MAR condition is satisfied (Allison, 2002; Little & Rubin, 2002). For the childhood model, control variables showed relatively low level s of missing responses. Out of the 11 control variables, the range of missing responses was parent or not, g2 education dummy variables, g3 gender) and 10.11% (g3 anger/frust ration and impulsivity). The dependent variable also showed a low level of missingness 2.38%. Parenting variables, however, showed a higher level of missing responses. The range of missing responses for independent variables was between 2.97% (G2 adolescen Given that the study is intergenerational as well as longitudinal in nature, the level of missing responses is reasonably higher for adolescence model. Out of the 11 control variables, the range o f missing responses was between 11.44% (g2 race not) and 20.80% (g3 anger/frustration and impulsivity). The dependent variable showed a modest level of missingness 7.14%. The range of missing responses for independent variables for adolescence model was between 12.11% (g2 adolescence risk) and 26.39% (g2 consistency of discipline) As suggested by Harrell (2001), the data have been interrogated for patterns of missingness. It did not appear to contain any sort of systematic patterns. According to
45 Allison (2002), listwise or case deletion is the method that is most robust to violations of pro bability of missing data on any of the independent variables does not depend on the values of the dependent variable, then regression estimates using listwise deletion will be unbiased (Allison, 2002, p.6 7) which is exactly the case for our analytic mod els. However, because our analysis sample size is relatively small (n=336 for childhood model and n=322 for adolescence model) and a listwise deletion leads to a substantial decrease in our sample size (eliminating 137 subjects in childhood model and 120 s ubjects in adolescence model), we could not adopt the listwise deletion method. In addition, methods like mean substitution and pairwise deletion should not be adopted in general (Allison, 2002; Kelley & Maxwell, 2010; Schafer & Graham, 2002). Furthermore, the pattern of missing data is not monotonic in our analysis sample. Accordingly, multiple imputation is desirable for dealing with missing data in our analysis sample. 9.2 1 w ith Markov Chain Monte Carlo (MCMC) method 2 As Allison (2002) suggested, dependent variable has been included in the imputation process. Even though multiple imputation is the desirable way for dealing with missing data in our current investigation, it is still necessary to mention that imputing v alues is still an educated guess as to what a respondent would have answered (Allison, 2002). 1 SAS and all other SAS Institute Inc. product or ser vice name s are registered trademarks or trademarks of SAS Institute Inc. in the USA and other countries. indicates USA registration. 2 Our imputation model includes all the variables in the analysis T he number of times of imputation equals to 5. In short, three steps were involved in multiple imputation: 1) a series of imputed data sets were created based on chosen variables and an imputation method; 2) statistical analysis were carried out on each of the imputed data set; 3) parameter estimated from each of the imputed data sets were combined to get a final single set of parameter estimates.
46 Addressing assumptions Standard approaches to multiple regression (MR) rely on ordinary least squares (OLS) to estimate model parameters. Although no distribut ional assumptions were made about independent variables and estimation of the regression coefficients does not require any parametric assumptions, inferences from coefficient such as p values and confidence intervals for regression estimates do depend on a ssumptions (Agresti & Finlay, 2009; Kelley & Maxwell, 2010). All models were checked for outliers, heteroskedasticity and collinearity. Although some literature (e.g. Rose & Fraser, 2008; Quintano et al., 2010) suggests that regression diagnostics should b e done before data imputation, I ran regression diagnostics both before and after imputing the data. Outliers were checked in three different ways. First, studentized residuals were created for all independent variables. Attention should be paid to stude ntized residuals that exceed 2 in absolute value. Second, leverage statistics were examined for values greater than (2k+2)/n, where k is the number of predictors and n is the number of cut off point n on and after the imputation), it was removed. Thus, different regression models may have slightly different analysis sample size depending on the severity of the outliers in the model. Another important assumption underlying multiple regression techniqu e is that error variance is homogeneous across all values of the regressors. In other words, homoscedasticity is required (Agresti & Finlay, 2009; White, 1980). If the variance of the
47 residuals is non test stat istic and p value become unreliable because of the violation of the Gauss Markov assumptions in that the distance between the error terms do not have a systematic deviation from the regression line (White, 1980). When heteroskedasticity is detected, robust standard errors (RSE) were used in the model 3 Collinearity was also examined for those multiple regression models that contained more than one regressor. Variance inflation factors (VIF) were checked for all SAS program has also been used to check multicollinearity. Multicollinearity did not seem to be a major concern for all models. Tests on normality of residuals, on nonlinearity and issues of independence have also been conducted. Models examining modera tion As we mentioned earlier, one major research adolescence risk regarding general delinquency. Essentially, this buffering effect can be depicted as a statisti cal interaction term. The standard way of allowing for the possibility of an interaction (or equivalently, a moderator) is to add cross product terms in multiple regression models (Kelley & Maxwell, 2010). For example, with two regressors, the models becom es Y i = 0 1 X 1 2 X 2 3 X 1 X 2 i 3 P covariance matrix of the estimates under the hypothesis of heteroskedasticity, which provides a more accurate t test statistic and associated p value.
48 Because of the inclusion of the cross product terms in the model, two additional issues need to be addressed. The first issue is how to interpret 1 2 coefficients when there also exist cross product terms in the mode l. Some researchers have interpreted those coefficients as if they corresponded to main effects, but this is not generally correct (Aiken & West, 1991). Main effects are typically defined as the constant effects of one variable across all values of another 1 2 coefficients never represent constant effects of the predictors in the presence of an interaction (Cramer & 1 and X 2 are centered by subtracting the sample mean fom all scores (mean centering), yielding a new coding with a mean of 0 (Kelley & Maxwell, 2010, p.290) Thus, the X 1 and X 2 coefficients from mean centered equations represent the effects of the predictors at the mean of the other predictors. In addition equation, they may also be considered as the weighted average effect of each predictor coefficient across all observed values of the other predictor (p.38 39) The other issue is the introduction of mu lticollinearity when we include cross product, in other words, higher order terms in the model. More specifically, as Aiken and highest order term produces large standard errors for the regression coefficients of the lower orders terms, though the standard error of the highest order term is unaffected (p.35) Cohen (1978) and Pedhazur (1982) also acknowledged the multicollinearity problem arising from higher order terms. As a matter of fact, mean centering technique is also suitable for dealing with this multicollinearity problem. The cross product term
49 with centered variables does not overlap with first order terms in the model, and the estimates and their standard errors are similar as in the no interaction model (Agresti & Finlay, 2009). The only remaining correlation between cross product and first order terms is due to non normality of the variables (Aiken & West, 1991; See also Marquardt, 1980; Smith & Sasaki, 1979; T ate, 1984). Variable selection techniques are justified Another desideratum listed by up approach to examining regression equations that include cross product or other higher order terms. At each step of the hierarchical step up approach, the variables previously par t of the reported results that shows the estimated model improvement when comparing the richer model to the simpler models (Kelley & Maxwell, 2010, p.287) The improvement is often gauged in terms of the change in overall effect sizes or R 2 Although the step up approach is quite common among social scientists, it could lead to the interpretational problem of considering main effects before we determine whether interactions exist, which is one central issue in the present study. In reality Aiken and We st (1991), Finney et al. (1984) and many other researchers emphasized the distinction between the theoretical versus exploratory basis of the interaction. They contended expectation of an interaction, step down procedures should be used 1991, p.105) For the present study, we have relatively strong theoretical and empirical support for promotive factors, but the theoretical ground for the cross product terms or
50 protective fa ctors is not very strong since this is the first exploratory study to examine this moderation effect. As a result, the step down approach recommended by Aiken and West (1991) was adopted here. Analytic Strategy 9.2 st atistical package is used to estimate a series of multiple regression models. The analysis for the present study occurs in four involvement in delinquent behaviors and G3 outcome is estimated for both childhood and adolescence models. A statistically significant and positive association between adolescent delinquent acts indeed influence th e healthy development of their offspring (G3). Bivariate associations between theoretically and empirically derived parental measures and G3 delinquency have also been estimated. 2) A baseline model was estimated including G2 adolescent risk as the predict or variable and all control variables in the model. In line with the first hypothesis, G2 adolescent risk should be positively related to G3 childhood and adolescence delinquency net of the control variables. 3) Since this is an exploratory study, a multip le regression model was r un for each theoretically and empirically driven parental variable, examining their possible promotive and protective factors when other proposed parental variables were not simultaneously included in the model. Through this proces s, we are able to see whether certain parental variables are more important than others when we only have risk and control variables in the model. 4 ) Following the suggestion given by Aiken and West (1991), a comprehensive multiple regression model was est imated by the step down approach. A
51 were significant, then each cross product terms become candidates for follow up tests. Because all our proposed promotive facto rs are theoretically and empirically driven, we keep all of them in the model. Doing so informs the literature and leads to the accumulation of knowledge about the issue of intergenerational resilience of delinquent behaviors. A cumulative effect of promo tive factors has also been graphed.
52 Figure 3 1. G3 age d istribution in RIGS
53 Table 3 1 Descriptive statistics G3 childhood m odel Variables Mean S.D. Min Max G2 concurrent delinquency 1.341 1.909 0 7.010 19.082 2.046 14.100 23.600 G2 as supervisory parent 0.848 0.360 0 1 G2 HS diploma 0.601 0.491 0 1 G2 beyond HS 0.110 0.313 0 1 G2 Black 0.771 0.421 0 1 G2 Hispanic 0.162 0.369 0 1 G3 cognitive competence 89.760 12.276 45.000 126.000 G3 gender 1.482 0.500 1 2 G3 anger/frustration 2.139 0.714 1.000 4.000 G3 impulsivity 2.874 0.358 1.833 4.000 G2 adolescent risk 2.889 2.151 0 6.831 G2 affective ties to G3 4.423 0.407 2.700 5.000 G2 consistency of discipline 3.501 0.808 1.250 5.000 G2 involvement in activities 3.788 0.637 1.833 5.000 G2 supervision of G3 4.777 0.375 3.000 5.000 life 2.236 1.405 0 4.000 G3 childhood delinquency 0.184 0.187 0 0.889 N=336
54 Tab le 3 2 Descriptive statistics G3 adolescence model Variables Mean S.D. Min Max G2 concurrent delinquency 0.804 1.574 0 6.613 19.044 2.045 14.100 23.600 G2 as supervisory parent 0.849 0.359 0 1 G2 HS diploma 0.606 0.489 0 1 G2 beyond HS 0.116 0.320 0 1 G2 Black 0.782 0.414 0 1 G2 Hispanic 0.148 0.356 0 1 G3 cognitive competence 90.025 12.081 45.000 121.000 G3 gender 1.486 0.501 1 2 G3 anger/frustration 2.127 0.723 1.000 4.000 G3 impulsivity 2.886 0.356 1.833 4.000 G2 adolescent risk 2.796 2.153 0 6.831 G2 affective ties to G3 4.487 0.464 2.600 5.000 G2 consistency of discipline 3.823 0.690 1.500 5.000 activities 3.881 0.671 2.000 5.000 G2 supervision of G3 4.665 0.466 3.333 5.000 life 2.027 1.429 0 5.000 G3 adolescent delinquency 0.187 0.230 0 1.500 N=322
55 CHAPTER 4 RESULTS Bivariate Analysis A correlation analysis ( Table 4 1 for childhood model and Table 4 2 for adolescence model) was first conducted to determine the nature of the relationship between the variables in the study. As expected, G2 adolescence risk was significantly correlated with both G3 childhood deli nquent behaviors (r=0.23, p<0.01) and G3 adolescence delinquent behaviors (r=0.19, p<0.01). In regards to parental measures, for the childhood model, G2 affective ties to G3 (r= 0.34, p<0.01), G2 consistency of discipline (r= 0.20, p<0.01) and G2 involveme 0.16, p<0.01) were significantly and negatively correlated G3 childhood delinquent behavior. However, G2 supervision (r= (r= 0.08, p=0.19) were not significantly relate d to G3 childhood delinquent behavior. For the adolescence model, G2 affective ties to G3 (r= 0.43, p<0.01), G2 consistency of discipline (r= 0.20, p<0.01) and G2 supervision of G3 (r= 0.13, p=0.049) were significantly correlated with G3 delinquent p=0.55) was not significant. It is also worth mentioning that the highest correlation among parenting measures is around 0.40. To some extent, it means different parental variable indeed measured different concepts. Several significant relationships were also observed between G3 delinquent concurrent delinquency (r=0.15, p<0.01) was positively and significantly correlated with G3 delinquent behavior. In addition, both of the two G3 temperament control variable
56 were significantly correlated with G3 delinquent behavior (anger/frustration: r=0. 45, delinquency (0.14, p=0.02) was positively and significantly correlated with G3 delinquent 0.15, p=0.01) is negativ ely and significantly related to G3 delinquent behavior, which indicates that there might exist a lagged effects for the negative effects of precocious life transition such as being a parent before the normatively expected age to show up. Moreover, both of the two G3 temperament control variable are significantly correlated with G3 delinquent behavior (anger/frustration: r=0.31, p<0.01; impulsivity: r=0.18, p<0.01). The Baseline Models Results from the first multivariate analysis can be found in Table 4 3 for childhood model and in Table 4 4 for adolescence model. The baseline model for G3 childhood reveals that after taking important control variables into consideration, G2 adolescence risk (b=0.014, p<0.01) is still significantly related to G3 childhood d elinquent behaviors. 0.011, p=0.026) was also significant, which means not having a precocious parent reduces the chance for G3 to commit delinquent behaviors. In addition, G3 anger/frustration (b=0.121, p<0.01) is significant. A s suggested by Kelley and Maxwell (2010), the omnibus effect size of the model should also be reported. The most widely used omnibus effect size in social science research is the squared multiple proportion of variance in Y that can be accounted for by the K regressor variables (Kelley & Maxwell, 2010, p.284) biased. Thus, the adjusted value of R2, denoted RA2, should be reported and used as the be adjusted R2 for this model is around 0.267,
57 suggesting that variables included in the model accounted for about 26.7% of the explained variance in G3 childhood delinquent behavior among the samp le. The baseline model for G3 adolescence reveals that after considering important controls, G2 adolescence risk (b=0.013, p=0.053) is marginally significantly correlated 0.023, p<0.01) was also significant. Additionally, G3 anger/frustration (b=0.102, p<0.01) was significant. The average adjusted R2 for this model was about 0.181, suggesting that variables included in the model accounted for about 18.1% of the explained variance in G3 adolescenc e delinquent behavior among the sample Individual Models G2 affective ties to G3 Table 4 5 and 4 6 show the results from OLS regressions that assess the promotive and/or protective effects of G2 affective ties to G3 on G3 childhood and adolescence deli nquency. As mentioned above, a step down approach is adopted and variables were centered 1 Table 4 5 shows G2 affective ties to G3 not only has a promotive effect (b= 0.10 6 p<0.01) but also a protective effect (b= 0.03 1 p <0.01 ), indicating that the slope of the regression of G3 childhood delinquency on G2 adolescence risk at levels of G2 affective ties decrease by 0.03 1 unit for every one unit increase in G2 affective ties. G2 adolescence risk (b=0.012, p<0.01) again is significant here. Additionally, G3 anger/frustration (b=0.104, p<0.01) is significant. Table 4 6 shows G2 affective ties to G3 again has both promotive (b= 0.15 3 p<0.01) and protective effects (b= 0.04 2 p<0.01) in G3 adolescence model The cross product term indicates that the slope of th e regression of G3 adolescence delinquency on G2 1 Aiken and West (2001) emphasized the importance of centering continuous variables before creating product terms.
58 adolescence risk at levels of G2 affective ties decrease by 0.04 2 unit for every one unit 0.013, p=0.04 6 ) is significant in the adolescence model as well as G3 anger/frustration ( b=0.08 9 p<0.01 ). G2 consistency of discipline Table 4 7 and 4 8 showed the results from OLS regressions that assess the promotive and/or protective effects of G2 consistency of discipline on G3 childhood and adolescence del inquency. Model 1 of Table 4 7 showed that G2 consistency of discipline (b= 0.00 6 p=0. 380 ) is not a statistically significant protective factor, though the cross product term is in the expected direction. One step back, Model 2 of Table 4 7 also showed th at G2 consistency of discipline (b= 0.02 3 p=0. 055 ) is marginally a promotive factor. Model 1 of Table 4 8 showed that G2 consistency of discipline (b= 0.0 10 p=0. 263 ) is not a protective factor for adolescence model either, though the sign is in the expec ted direction. One step back, however, model 2 of Table 4 8 showed that G2 consistency of discipline (b= 0.0 42 p=0.030) is a promotive factor in the adolescence model. Model 1 of Table 4 9 showed that G2 involvement 0.00 9 p=0.3 10 ) is not a statistically significant protective factor for G3 childhood delinquency, though it is in the expected direction. One step back, model 2 of Table 4 9 showed that G2 involvement (b= 0.0 34 p=0.0 68 ) is marg inally a promotive factor. Very interestingly, Model 1 of Table 4 10 showed that 0.04 7 p=0.0 12 ) but also protective effects (b= 0.019, p=0.0 30 ) against G2 adolescence risk on G3 adole scence delinquency, indicating that the slope of the regression of G3 adolescence
59 delinquency on G2 adolescence risk at levels of G2 involvement decrease by 0.019 unit for every one unit increase in G2 involvement. G2 supervision of G3 Model 1 of Table 4 11 showed that G2 supervision of G3 (b=0.00 41 p=0. 759 ) is not a statistically significant protective factor for G3 childhood delinquent behaviors, and model 2 of Table 4 11 showed that G2 supervision (b= 0.00 2 p=0. 941 ) is also not a promotive factor. Model 1 and 2 of Table 4 12 demonstrated that G2 supervision of G3 is neither a protective (b= 0.0 14 p=0. 415 ) nor a promotive (b= 0.0 35 p=0. 213 ) for G3 adolescence delinquency. Model 1 and 2 of Table 4 13 showed tha t parent school relationship does not have either significant protective (b=0.00 1 p=0. 864 ) or promotive effects (b= 0.004, p=0. 558 ) for G3 childhood delinquency. Model 1 and 2 of Table 4 14 again showed that parent school relationship is not a protective (b=0.00 4 p=0. 396 ) or promotive factor (b=0.01 3 p=0. 164 ) for G3 adolescence delinquency. In summary, we found that G2 affective ties to G3 seems to be the most important promotive and protective factor that would buffer the intergenerational transmission both a protective and promotive factor for G3 adolescence. The only other parenting measure that has been found as having some promotive effects is G2 consistency of discipline for both G3 childhood and adolescence models. To some extent, this is unexpected because we hypothesized that the impact of parenting measures would decrease in terms of n umber of significant promotive /protective factors.
60 Comprehensive Models Now comprehensive models are estimated by the step down approach recommended by Aiken and West (1991) for G3 childhood and adolescence respectively. Model 1 in Table 4 15 shows the fu ll model with all cross product or interaction terms and promotive factors in the model. Consistent with results from individual models, we found that G2 affective ties again is both a protective factor (b= 0.0 29 p=0.02 7 ) and a promotive factor (b= 0.0 91 p<0.01) in the comprehensive model for G3 childhood. G2 adolescent risk is still significantly and positively related to G3 childhood delinquency (b=0.01 1 p = 0.01 4 ). The only other significant variable is G3 anger and frustration with b=0.099, and p<0.01. The average adjusted R 2 for this model was about 0.34, suggesting that variables included in the model accounted for about 34% of the explained variance in G3 childhood delinquent behavior among the sample. Model 2 in the Table 4 15 shows the model with all promotive factors in the model. Through examining R 2 change in model 1 and model 2, we conducted the (Aiken & West, 1991). Results show that again only G2 affective t ies to G3 (b= 0.0 90 p<0.01) is a significant promotive factor. The average adjusted R 2 for this model was about 0.3 2 suggesting that variables included in the model accounted for about 3 2 % of the explained variance in G3 childhood delinquent behavior amo ng the sample. Although the change of R 2 from 0.3 2 to 0.34 seems to be not substantial, joint Wald test, an asymptotic equivalent of log likelihood ration test (Meng & Rubin, 1992), showed that the block of cross product terms were significant, which furth er justified the existence of G2 affective ties to G3 as a protective factor in the childhood model.
61 A similar analytic procedure applies to the adolescence model. Model 1 in Table 4 16 shows the full model with interaction terms and promotive factors in the model. We found that G2 affective ties again is both a protective factor (b= 0.03 4 p=0.0 68 ) and a promotive factor (b= 0.1 36 p<0.001) in the comprehensive model for G3 adolescence. Again parenting measures as protective or promotive factors in the model. The only other significant variable is G3 anger and frustration with b=0.08 9 and p<0.01. The average adjusted R 2 for this model was about 0.3 3 suggesting that variables included in the model accounted for about 3 3 % of the explained variance in G3 childhood delinquent behavior among the sample. Model 2 in the Table 4 16 shows the model with only all promotive factors in the model. Results again show that only G2 affective ties to G3 (b= 0.162, p<0.01) is the only significan t promotive factor. The average adjusted R2 for this model was about 0.29, suggesting that variables included in the model accounted for about 29% of the explained variance in G3 childhood delinquent behavior among the sample. The joint Wald test again con firmed that the block of cross product terms were significant and justified the existence of G2 affective ties to G3 as a protective factor in the adolescence model. In summary, comprehensive models confirmed the promotive and protective effects of G2 affe ctive ties to G3 in both childhood and adolescence model. In addition, results from both models demonstrated that when there exist close ties between parent and child, no other promotive/protective factors seem to be important. Cumulative Promotive Effects As mentioned above, one of the basic premises of the risk and protective approach is that having multiple promotive factors in different domains would greatly
62 1987). In our c ase here, it means that promotive factors may add up or accumulate to increase the probability for the occurrence of intergenerational resilience of risk or delinquency. When examining cumulative effects, prior research usually only includes factors that h ave been found significantly related to the outcome in multivariate regression models (e.g. Ribeaud & Eisner, 2010; Thornberry et al., 1997; Zufferey et al., 2007). In addition, when the number of significant risk and/or promotive factors is large, cumulat ive effects have also been assessed within each domain. As shown above, only G2 affective ties to G3 was found significant in both childhood and adolescence comprehensive models. Yet individual parental measure that is not significant by itself might p erform protection effects when combining with other parental measures. Thus, a s an exploratory study, we graphed the cumulative effects of all theoretically and empirically driven promotive factors respectively for G3 childhood and adolescence 2 Accordingl y, G2 adolescence risk and parental measures were divided at the median, which is the most widely used method of dichotomization in social sciences (e.g. Ribeaud & Eisner, 2010; Thornberry et al., 1997; Zufferey et al., 2007) 3 For any G3 subjects, the number of promotive factors ranges from 1 ( high G2 adolescence risk, no promotive factor at all ) to 5 ( low or no G2 adolescence risk, all existing 5 promotive factors ). The results for G3 childhood model are presented in Figure 4 1. These data clearly show the impact of cumulative protection. When the number of promotive factors 2 Since the graphs are descriptive in nature, we used the listwise deleti on method to handle missing data. 3 See, for example, Cohen (1983), Irwin and McClelland (2003) and Osgood (2005) for why dichotomizing continuous variables is generally not recommended. Although dichotomizing continuous variable or scale may lead to bias ed standard errors and statistical significance, here we are not interested in whether, after dichotomization, individual parental measure would become a significant promotive factor or not.
63 increases, the value for delinquency decreases significantly. Of the 49 subjects with one prom otive factor in G3 childhood figure, the average value for delinquency is 0.192, whereas of the 38 subjects with 4 promotive factors, the average value for delinquency is 0.092, a 52% decrease in the value of delinquent behavior. The results for G3 adolesc ence model are presented in Figure 4 2. Again, we see a clearly general pattern of cumulative protection. Of the 39 subjects with 1 promotive factor, the average value for delinquency is 0.279, whereas of the 41 subjects with 4 promotive factors, the avera ge value for delinquency is 0.084, a 70% decrease in delinquency value. In sum, we may conclude that there exist some cumulative protection effects from multiple promotive factors. This is a research question that needs further investigation in the study o f intergenerational resilience of delinquency.
64 Figure 4 1 Cumulative promotive effects G3 childhood model 0.444 0.278 0.192 0.202 0.149 0.092 0.097 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 _1 (n=4) 0 (n=18) 1 (n=49) 2 (n=50) 3 (n=46) 4 (n=38) 5 (n=20) Delinquency Cumulative promotive effects -G3 childhood
65 Figure 4 2. Cumulative promotive effects G3 adolescence model 0.597 0.258 0.279 0.169 0.166 0.084 0.106 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 _1 (n=4) 0 (n=16) 1 (n=39) 2 (n=62) 3 (n=37) 4 (n=41) 5 (n=22) Delinquency Cumulative promotive effects -G3 adolescence
66 Ta ble 4 1. Bivariate correlations -childhood 1 2 3 4 5 6 7 8 9 1 -----2 0.226** ----3 0.344** 0.112* ----4 0.201** 0.136* 0.408** ----5 0.156** 0.107 0.218** 0.065 ----6 0.022 0.050 0.201** 0.041 0.230** ----7 0.081 0.084 0.044 0.099 0.355** 0.041 ----8 0.147** 0.274** 0.085 0.041 0.223** 0.167** 0.151** ----9 0.027 0.025 0.007 0.028 0.026 0.049 0.084 0.020 -----10 0.018 0.030 0.037 0.066 0.072 0.031 0.180** 0.031 0.073 11 0.024 0.030 0.028 0.009 0.124* 0.007 0.170** 0.061 0.057 12 0.057 0.050 0.128* 0.056 0.116* 0.105 0.248** 0.033 0.037 13 0.097 0.121** 0.050 0.015 0.061 0.041 0.025 0.138* 0.019 14 0.072 0.177** 0.082 0.124* 0.164** 0.023 0.366** 0.224** 0.059 15 0.020 0.005 0.054 0.030 0.042 0.008 0.035 0.013 0.097 16 0.003 0.110* 0.027 0.034 0.148* 0.008 0.115 0.007 0.115* 17 0.452** 0.118* 0.204** 0.120 0.100 0.026 0.171** 0.102 0.001 18 0.153** 0.045 0.047 0.077 0.091 0.007 0.010 0.011 0.026
67 Table 4 1. Continued 10 11 12 13 14 15 16 17 18 1 2 3 4 5 6 7 8 9 10 ----11 0.608** ----12 0.126* 0.025 ----13 0.013 0.059 0.073 ----14 0.032 0.067 0.047 0.070 ----15 0.179** 0.200** 0.067 0.089 0.017 ----16 0.029 0.048 0.086 0.026 0.068 0.431** -----17 0.070 0.086 0.170** 0.014 0.012 0.142* 0.017 ----18 0.091 0.589 0.053 0.088 0.062 0.096 0.023 0.096 ----*p<0.05, **p<0.01 List of v ariables for Table 4 1 1. G3 childhood delinquency 2. G2 adolescent delinquency 3. G2 affective ties to G3 4. G2 consistency of discipline 5. 6. G2 supervision of G3 7. 8. G2 concurrent delinquency 9. G3 cognitive competence 10. G2 as Black 11. G2 as Hispanic 12. 13. G3 gender 14. G2 as supervisory parent or not 15. G2 HS diploma 16. G2 beyond HS 17. G3 anger/frustration 18. G3 impulsivity
68 Table 4 2 Bivariate c orrelations adolescence 1 2 3 4 5 6 7 8 9 1 -----2 0.188** ----3 0.434** 0.089 ----4 0.209** 0.086 0.473** ----5 0.203** 0.122 0.367** 0.120 ----6 0.131* 0.004 0.254** 0.171** 0.287** ----7 0.038 0.084 0.130* 0.177** 0.305** 0.063 ----8 0.142* 0.276** 0.109 0.199** 0.147* 0.113 0.146* ----9 0.047 0.103 0.132* 0.070 0.086 0.080 0.078 0.011 ----10 0.014 0.059 0.006 0.115 0.076 0.006 0.082 0.067 0.009 11 0.034 0.038 0.050 0.117 0.103 0.018 0.069 0.033 0.035 12 0.159* 0.053 0.209** 0.030 0.129* 0.223** 0.131* 0.095 0.015 13 0.115 0.129* 0.166** 0.098 0.059 0.033 0.110 0.066 0.009 14 0.099 0.083 0.084 0.022 0.159* 0.024 0.375** 0.064 0.105 15 0.014 0.027 0.075 0.039 0.023 0.016 0.050 0.066 0.120* 16 0.036 0.104 0.008 0.001 0.168** 0.114 0.036 0.005 0.095 17 0.310** 0.101 0.110 0.036 0.044 0.066 0.086 0.124* 0.008 18 0.188** 0.048 0.102 0.084 0.015 0.073 0.025 0.059 0.064
69 Table 4 2. Continued 10 11 12 13 14 15 16 17 18 1 2 3 4 5 6 7 8 9 10 ----11 0.588** ----12 0.115* 0.034 ----13 0.015 0.048 0.091 ----14 0.062 0.073 0.046 0.018 ----15 0.186** 0.216** 0.038 0.076 0.026 ----16 0.025 0.027 0.082 0.036 0.012 0.449** ----17 0.052 0.061 0.174** 0.019 0.080 0.139* 0.031 ----18 0.075 0.038 0.090 0.082 0.003 0.078 0.020 0.121* ----*p<0.05, **p<0.01 List o f v ariables for Table 4 2 1. G3 childhood delinquency 2. G2 adolescent delinquency 3. G2 affective ties to G3 4. G2 consistency of discipline 5. 6. G2 supervision of G3 7. G2 8. G2 concurrent delinquency 9. G3 cognitive competence 10. G2 as Black 11. G2 as Hispanic 12. 13. G3 gender 14. G2 as supervisory parent or not 15. G2 HS diploma 16. G2 beyond HS 17. G3 anger/frustration 18. G3 impulsivity
70 Table 4 3 Baseline model childhood Variables b (s.e.) p value Intercept 0.1846(0.008) <0.001** G2 adolescence delinquency 0.0137(0.005) 0.004** G2 concurrent delinquency 0.0056(0.005) 0.286 G3 cognitive competence 0.0067(0.001) 0.374 G2 as Black 0.0136(0.039) 0.727 G2 as Hispanic 0.0007(0.044) 0.987 0.0118(0.005) 0.026* G3 gender 0.0340(0.020) 0.098 G2 as supervisory parent or not 0.0217(0.028) 0.442 G2 HS diploma 0.0374(0.023) 0.103 G2 beyond HS 0.0304(0.032) 0.345 G3 anger/frustration 0.1215(0.013) <0.001** G3 impulsivity 0.0315(0.030) 0.294 Adjust R 2 =0.267 *p<0.05, **p<0.01
71 Table 4 4 Baseline model adolescence Variables b (s.e .) p value Intercept 0.1901(0.012) <0.001** G2 adolescence delinquency 0.0126(0.006) 0.053* G2 concurrent delinquency 0.0143(0.008) 0.075 G3 cognitive competence 0.0005(0.001) 0.580 G2 as Black 0.0365(0.058) 0.531 G2 as Hispanic 0.0268(0.067) 0.692 0.0223(0.006) <0.001** G3 gender 0.0408(0.025) 0.102 G2 as supervisory parent or not 0.0576(0.035) 0.106 G2 HS diploma 0.0233(0.033) 0.485 G2 beyond HS 0.0064(0.045) 0.888 G3 anger/frustration 0.1025(0.019) <0.001** G3 impulsivity 0.0732(0.037) 0.053* Adjust R 2 =0.181 *p<0.05, **p<0.01
72 Table 4 5 Individual m odel (G2 affective ties to G3 -c hildhood) Model 1 Model 2 Variables b (s.e.) p value b (s.e.) p value Intercept 0.1844(0.009) <0.001** 0.1814(0.008) <0.001** G2 adolescent delinquency 0.0125(0.005) 0.013* 0.0122(0.005) 0.008** G2 affective ties to G3 0.1012(0.023) <0.001** 0.1061(0.023) <0.001** Risk a ffective ties 0.0310(0.010) 0.003** G2 concurrent delinquency 0.0037(0.005) 0.448 0.0049(0.005) 0.315 G3 cognitive competence 0.0007(0.001) 0.354 0.0003(0.001) 0.689 G2 as Black 0.0063(0.040) 0.875 0.0081(0.037) 0.826 G2 as Hispanic 0.0065(0.043) 0.879 0.0020(0.042) 0.961 0.0089(0.005) 0.077 0.0072(0.005) 0.132 G3 gender 0.0306(0.019) 0.117 0.0310(0.019) 0.116 G2 as supervisory parent or not 0.0356(0.026) 0.172 0.0364(0.026) 0.166 G2 HS diploma 0.0380(0.022) 0.088 0.0328(0.022) 0.136 G2 beyond HS 0.0324(0.032) 0.308 0.0266(0.031) 0.388 G3 anger/frustration 0.1087(0.013) <0.001** 0.1041(0.013) <0.001** G3 impulsivity 0.0346(0.029) 0.245 0.0316(0.028) 0.262 *p<0.05, **p<0.01
73 Table 4 6 Individual m odel (G2 affective ties to G3 adolescence ) Model 1 Model 2 Variables b (s.e.) p value b (s.e.) p value Intercept 0.1903(0.012) <0.001** 0.1880(0.011) <0.001** G2 adolescent delinquency 0.0116(0.006) 0.044* 0.0127(0.005) 0.002** G2 affective ties to G3 0.1680(0.029) <0.001** 0.1526(0.029) <0.001** Risk affective ties 0.0418(0.012) <0.001** G2 concurrent delinquency 0.0080(0.008) 0.298 0.0078(0.007) 0.289 G3 cognitive competence 0.0003(0.001) 0.734 0.0001(0.001) 0.985 G2 as Black 0.0340(0.052) 0.516 0.0337(0.051) 0.509 G2 as Hispanic 0.0217(0.062) 0.728 0.0200(0.061) 0.743 0.0133(0.007) 0.053* 0.0125(0.006) 0.046* G3 gender 0.0190(0.024) 0.420 0.0175(0.023) 0.444 G2 as supervisory parent or not 0.0731(0.037) 0.054* 0.0851(0.037) 0.029* G2 HS diploma 0.0369(0.032) 0.257 0.0212(0.034) 0.533 G2 beyond HS 0.0025(0.043) 0.953 0.0195(0.044) 0.660 G3 anger/frustration 0.0908(0.018) <0.001** 0.0888(0.018) <0.001** G3 impulsivity 0.0653(0.040) 0.111 0.0692(0.038) 0.081 *p<0.05, **p<0.01
74 Table 4 7 Individual m odel (G2 consistency of discipline childhood ) Model 1 Model 2 Variables b (s.e.) p value b (s.e.) p value Intercept 0.1852(0.009) <0.001** 0.1837(0.009) <0.001** G2 adolescent delinquency 0.0126(0.005) 0.012* 0.0122(0.005) 0.016* G2 consistency of discipline 0.0229(0.012) 0.055* 0.0219(0.012) 0.060* Risk consistency of discipline 0.0061(0.007) 0.380 G2 concurrent delinquency 0.0050(0.005) 0.313 0.0055(0.005) 0.269 G3 cognitive competence 0.0007(0.001) 0.393 0.0006(0.001) 0.394 G2 as Black 0.0093(0.042) 0.824 0.0055(0.040) 0.890 G2 as Hispanic 0.0029(0.044) 0.948 0.0046(0.046) 0.919 0.0116(0.005) 0.028* 0.0119(0.005) 0.024* G3 gender 0.0333(0.020) 0.099 0.0331(0.020) 0.104 G2 as supervisory parent or not 0.0321(0.029) 0.273 0.0324(0.030) 0.281 G2 HS diploma 0.0370(0.023) 0.109 0.0356(0.023) 0.117 G2 beyond HS 0.0306(0.033) 0.352 0.0274(0.033) 0.401 G3 anger/frustration 0.1191(0.014) <0.001** 0.1181(0.014) <0.001** G3 impulsivity 0.0315(0.030) 0.306 0.0341(0.029) 0.247 *p<0.05, **p<0.01
75 Table 4 8 Individual m odel (G2 consistency of discipline adolescence ) Model 1 Model 2 Variables b (s.e.) p value b (s.e.) p value Intercept 0.1928 (0.012) <0.001** 0.1921(0.012) <0.001** G2 adolescent delinquency 0.0126 (0.007) 0.061 0.0128(0.006) 0.052* G2 consistency of discipline 0.0421(0.019) 0.030* 0.0412(0.019) 0.033* Risk consistency of discipline 0.0099(0.009) 0.263 G2 concurrent delinquency 0.0099(0.008) 0.228 0.0096(0.008) 0.241 G3 cognitive competence 0.0004(0.001) 0.688 0.0004(0.001) 0.704 G2 as Black 0.0384(0.056) 0.494 0.0382(0.056) 0.498 G2 as Hispanic 0.0201(0.066) 0.762 0.0154(0.066) 0.816 0.0220(0.006) 0.001** 0.0220(0.007) 0.001** G3 gender 0.0332(0.026) 0.194 0.0324(0.025) 0.199 G2 as supervisory parent or not 0.0735(0.036) 0.042* 0.0776(0.036) 0.035* G2 HS diploma 0.0267(0.034) 0.436 0.0240(0.034) 0.484 G2 beyond HS 0.0014(0.046) 0.976 0.0045(0.046) 0.922 G3 anger/frustration 0.1051(0.018) <0.001** 0.1051(0.018) <0.001** G3 impulsivity 0.0671(0.036) 0.065 0.0646(0.036) 0.071 *p<0.05, **p<0.01
76 Table 4 9 Individual m childhood ) Model 1 Model 2 Variables b (s.e.) p value b (s.e.) p value Intercept 0.1826(0.009) <0.001** 0.1805(0.009) <0.001** G2 adolescent delinquency 0.0138(0.005) 0.004** 0.0138(0.005) 0.005** G2 involvement 0.0341(0.018) 0.068 0.038(0.018) 0.046* Risk involvement 0.0097(0.009) 0.310 G2 concurrent delinquency 0.0024(0.005) 0.641 0.0011(0.005) 0.826 G3 cognitive competence 0.0006(0.001) 0.423 0.0006(0.001) 0.429 G2 as Black 0.0167(0.042) 0.689 0.0192(0.040) 0.630 G2 as Hispanic 0.0008(0.044) 0.985 0.0011(0.044) 0.980 0.0110(0.005) 0.038* 0.0108(0.005) 0.040* G3 gender 0.0317(0.020) 0.114 0.0318(0.020) 0.113 G2 as supervisory parent or not 0.0120(0.030) 0.691 0.0081(0.032) 0.801 G2 HS diploma 0.0383(0.023) 0.096 0.0360(0.022) 0.108 G2 beyond HS 0.0399(0.032) 0.219 0.0322(0.032) 0.321 G3 anger/frustration 0.1183(0.014) <0.001** 0.1164(0.014) <0.001** G3 impulsivity 0.0361(0.029) 0.230 0.0405(0.028) 0.148 *p<0.05, **p<0.01
77 Table 4 10 Individual m odel (G2 adolescence ) Model 1 Model 2 Variables b (s.e.) p value b (s.e.) p value Intercept 0.1857(0.012) <0.001** 0.1823(0.012) <0.001** G2 adolescent delinquency 0.0121(0.006) 0.065 0.0126(0.006) 0.045* G2 involvement 0.0431(0.019) 0.027* 0.0471(0.019) 0.012* Risk involvement 0.0194(0.009) 0.030* G2 concurrent delinquency 0.0115(0.008) 0.153 0.0092(0.008) 0.267 G3 cognitive competence 0.0003(0.001) 0.772 0.0005(0.001) 0.658 G2 as Black 0.0355(0.058) 0.542 0.0314(0.056) 0.576 G2 as Hispanic 0.0211(0.067) 0.755 0.0152(0.066) 0.818 0.0206(0.006) 0.002** 0.0212(0.006) 0.001** G3 gender 0.0432(0.025) 0.084 0.0428(0.024) 0.078 G2 as supervisory parent or not 0.0256(0.037) 0.485 0.0291(0.038) 0.440 G2 HS diploma 0.0320(0.034) 0.347 0.0280(0.034) 0.410 G2 beyond HS 0.0123(0.048) 0.767 0.0019(0.049) 0.968 G3 anger/frustration 0.1007(0.019) ( ) <0.001** 0.1018(0.019) <0.001** G3 impulsivity 0.0749(0.037) 0.046* 0.0753(0.037) 0.047* *p<0.05, **p<0.01
78 Table 4 11 Individual m odel (G2 supervision of G3 childhood ) Model 1 Model 2 Variables b (s.e.) p value b (s.e.) p value Intercept 0.1845(0.009) <0.001** 0.1846(0.009) <0.001** G2 adolescent delinquency 0.0138(0.005) 0.006** 0.0138(0.005) 0.007** G2 supervision of G3 0.0021(0.029) 0.941 0.0024(0.029) 0.935 Risk supervision 0.0043(0.014) 0.759 G2 concurrent delinquency 0.0050(0.005) 0.319 0.0054(0.005) 0.299 G3 cognitive competence 0.0006(0.001) 0.411 0.0007(0.001) 0.408 G2 as Black 0.0167(0.043) 0.700 0.0157(0.043) 0.719 G2 as Hispanic 0.0026(0.045) 0.953 0.0029(0.045) 0.949 0.0119(0.005) 0.031* 0.0119(0.005) 0.031* G3 gender 0.0343(0.021) 0.104 0.0347(0.021) 0.096 G2 as supervisory parent or not 0.0228(0.028) 0.422 0.0219(0.028) 0.437 G2 HS diploma 0.0313(0.023) 0.123 0.0375(0.023) 0.309 G2 beyond HS 0.0299(0.033) 0.316 0.0318(0.033) 0.338 G3 anger/frustration 0.1217(0.014) <0.001** 0.1216(0.013) <0.001** G3 impulsivity 0.0314(0.031) 0.311 0.0316(0.031) 0.308 *p<0.05, **p<0.01
79 Table 4 12 Individual m odel (G2 supervision of G3 adolescence ) Model 1 Model 2 Variables b (s.e.) p value b (s.e.) p value Intercept 0.1911(0.012) <0.001** 0.1906(0.013) <0.001** G2 adolescent delinquency 0.0127(0.005) 0.046* 0.0134(0.006) 0.038* G2 supervision of G3 0.0345(0.028) 0.213 0.0355(0.028) 0.205 Risk supervision 0.0135(0.016) 0.415 G2 concurrent delinquency 0.0127(0.008) 0.111 0.0121(0.008) 0.129 G3 cognitive competence 0.0005(0.001) 0.599 0.0006(0.001) 0.552 G2 as Black 0.0414(0.060) 0.491 0.0440(0.060) 0.471 G2 as Hispanic 0.0329(0.070) 0.641 0.0340(0.070) () 0.633 0.0202(0.007) 0.003** 0.0204(0.007) 0.002** G3 gender 0.0399(0.025) 0.109 0.0408(0.025) 0.099 G2 as supervisory parent or not 0.0640(0.035) 0.065 0.0608(0.037) 0.102 G2 HS diploma 0.0215(0.032) 0.508 0.0169(0.032) 0.592 G2 beyond HS 0.0148(0.045) 0.742 0.0177(0.045) 0.692 G3 anger/frustration 0.1033(0.019) <0.001** 0.1020(0.019) <0.001** G3 impulsivity 0.0718(0.038) 0.062 0.0737(0.038) 0.058 *p<0.05, **p<0.01
80 Table 4 13 Individual m odel (G2 involvement childhood ) Model 1 Model 2 Variables b (s.e.) p value b (s.e.) p value Intercept 0.1839(0.009) <0.001** 0.1839(0.009) <0.001** G2 adolescent delinquency 0.0139(0.005) 0.006** 0.0138(0.005) 0.005** G2 involvement in G3 school life 0.0043(0.007) 0.557 0.0043(0.007) 0.563 Risk G3 school life 0.0006(0.003) 0.864 G2 concurrent delinquency 0.0049(0.005) 0.329 0.0049(0.005) 0.321 G3 cognitive competence 0.0062(0.001) 0.433 0.0064(0.001) 0.427 G2 as Black 0.0172(0.042) 0.685 0.0163(0.043) 0.707 G2 as Hispanic 0.0013(0.044) 0.976 0.0003(0.045) 0.994 0.0126(0.005) 0.021* 0.0126(0.005) 0.022* G3 gender 0.0349(0.020) 0.088 0.0345(0.020) 0.093 G2 as supervisory parent or not 0.0173(0.029) 0.560 0.0175(0.029) 0.555 G2 HS diploma 0.0366(0.023) 0.118 0.0369(0.023) 0.115 G2 beyond HS 0.0329(0.033) 0.319 0.0336(0.033) 0.311 G3 anger/frustration 0.1204(0.014) <0.001** 0.1202(0.014) <0.001** G3 impulsivity 0.0321(0.030) 0.295 0.0321(0.030) 0.294 *p<0.05, **p<0.01
81 Table 4 14 Individual m adolescence ) Model 1 Model 2 Variables b (s.e.) p value b (s.e.) p value Intercept 0.1922(0.012) <0.001** 0.1930(0.012) <0.001** G2 adolescent delinquency 0.0127(0.006) 0.050* 0.0124(0.006) 0.052* G2 involvement in G3 school life 0.0133(0.010) 0.164 0.0130(0.009) 0.177 Risk G3 school life 0.0036(0.004) 0.396 G2 concurrent delinquency 0.0159(0.008) 0.052* 0.0171(0.008) 0.039* G3 cognitive competence 0.0005(0.001) 0.607 0.0005(0.001) 0.656 G2 as Black 0.0327(0.057) 0.570 0.0325(0.058) 0.574 G2 as Hispanic 0.0256(0.066) 0.702 0.0218(0.068) 0.748 0.0215(0.006) 0.001** 0.0215(0.006) 0.001** G3 gender 0.0381(0.025) 0.130 0.0369(0.025) 0.143 G2 as supervisory parent or not 0.0826(0.038) 0.033* 0.0820(0.039) 0.036* G2 HS diploma 0.0187(0.033) 0.568 0.0213(0.033) 0.515 G2 beyond HS 0.0125(0.044) 0.777 0.0086(0.044) 0.853 G3 anger/frustration 0.1029(0.018) <0.001** 0.1028(0.018) <0.001** G3 impulsivity 0.0742(0.037) 0.045* 0.0704(0.037) 0.060 *p<0.05, **p<0.01
82 Table 4 15 Effect of p arenti ng measures on CBCL d elinquency outcome childhood Model 1 Model 2 Variables b coefficient p value b coefficient p value G2 adolescence delinquency 0.012 0.0107 0.011 0.0142 Affective ties 0.090 0.0003 0.091 0.0003 Consistency of discipline 0.018 0.1805 0.019 0.1530 Involvement 0.026 0.2073 0.031 0.1269 Supervision 0.036 0.2320 0.032 0.2552 Parent school relationship 0.002 0.8094 0.003 0.7283 Risk*Affective ties 0.0289 0.0266 Risk*Discipline 0.0004 0.9943 Risk*Involvement 0.0085 0.2532 Risk*Supervision 0.0084 0.5517 Risk*Parent school relationship 0.0015 0.6795 Adjust R 2 0.326 0.342
83 Table 4 16 Effect of parenting measures on CBCL delinquency outcome adolescence Model 1 Model 2 Variables b coefficient p value b coefficient p value G2 adolescence delinquency 0.010 0.0769 0.012 0.0301 Affective ties 0.162 <0.0001 0.136 0.0009 Consistency of discipline 0.011 0.6200 0.001 0.9543 Involvement 0.017 0.4731 0.026 0.2458 Supervision 0.006 0.8347 0.013 0.6144 Parent school relationship 0.011 0.3400 0.010 0.3607 Risk*Affective ties 0.034 0.0678 Risk*Discipline 0.002 0.8229 Risk*Involvement 0.012 0.3734 Risk*Supervision 0.002 0.8998 Risk*Parent school relationship 0.001 0.7764 Adjust R 2 0.295 0.324
84 CHAPTER 5 DISCUSSION AND CON C L USION Summary of Findings Past research has generally examined how risk or delinquent behaviors could transmit across generations (e.g. Ehrensaft et al., 2003; Farrington, 1993; Farrington, 2009; Kaplan & Liu, 1999; Kim et al., 2009; Thornberry, 2005; Thornberry et al., 2003; Thorn berry et al., 2009a; 2009b). In the existing literature on intergenerational continuity of delinquency, certain parenting measures, such as lack of parental involvement and monitoring, have been considered as of great importance in the process of transmiss ion. Despite the fact that there exist a great number of bad parents with good children, however, relatively few studies have taken a discontinuity perspective. The major goals of this thesis research were to explore intergenerational resilience in antisoc ial behavior and delinquency and to identify possible promotive and protective factors that would buffer parental risk that might be transmitted to their children. Using data from the Rochester Youth Development Study (RYDS) and its follow up, the Roches ter Intergenerational Study (RIGS) allowed for the exploration of intergenerational resilience of delinquency with a sample of more than 300 G3 subjects and their parents (G2). It was hypothesized that the correlation between parental (G2) adolescent delin quent history and G3 childhood and adolescent delinquent behavior would be significant but not very large in magnitude. That is, only a portion of children with parents who had adolescent delinquent history would become delinquents themselves. It was also hypothesized that good parent child relationship would play a generations. In addition, we also hypothesized that the impact of good parent child
85 relationship may decrease in G3 adolescence versus in G3 childhood in terms of the number of significant promotive/protective factors. In general, results showed mixed support for our hypotheses. Results from bivariate correlations showed that the correlation between G2 adolescence ri sk and G3 childhood and adolescence delinquency is significant but not large in magnitude (r=0.23 in childhood model and r=0.19 in adolescence model). While r value is abstract and not straightforward enough for us to understand the extent to which interge nerational resilience exists, it is necessary to discover another way to quantify intergenerational resilience. As a matter of fact, since this thesis research is among one of the very first studies that look at the issue of intergenerational discontinuity there is no existing literature on what an appropriate method could be regarding quantifying intergenerational discontinuity or, more specifically, intergenerational resilience. Masten and colleagues (1990, 1999) have specifically identified three group s of resilient phenomena: 1) overcoming the odds individuals show better than expected outcomes despite high risk status; 2) sustained competence under stress positive adaptation is attained even facing chronic environmental and interpersonal stress; and 3 we made a simple and preliminary attempt to quantify intergenerational resilience. We dichotomized both G2 adolescent delinquency and G3 childhood and adolescence delinq uency, and assessed the percentage of G3 subjects whose parent (G2) was in the high risk group but they themselves were not. We found that in G3 childhood model, 32.08% of children born to high risk parents were not in the high risk delinquency group.
86 In G 3 adolescence model, the magic number was 29.55%. The only other study, known to the author, which has made efforts to quantify intergenerational discontinuity was done by Lovergrove (2010) using the same datasets we are currently using in this thesis rese arch. Using more advanced statistical techniques, mainly trajectories group in the externalizing trajectory with the lowest level of problems behaviors across ages 8 11 (p.114) As such, 30% (more or less) might be our preliminary estimate of the degree to which intergenerational resilience of delinquency might take place. More research attention needs to be put onto this very basic but fundamental concept within this re latively new and promising field in criminology. Results from bivariate associations also showed that all our proposed parenting measures are in the expected direction in G3 childhood model, and in G3 adolescence model with the exception of G2 involvement statistically significant. Consistent with previous research on individual temperament and delinquency (e.g. Vitulano et al., 2010), G3 anger/frustration and impulsivity were found to be significantly correlated with G3 delinquency. Not surprisingly, when not competing with other variables, G2 concurrent delinquency is also significantly related to G3 delinquency both in the childhood and adolescence model with very similar r values, 0.147 and 0.142 respectively. How ever, when we look at the multivariate baseline models for G3 childhood and adolescence including important control variables, G2 concurrent delinquency is no longer significant, whereas G2 adolescence risk is still significant in both G3 childhood and ado lescence models. This preliminary finding is interesting and thought provoking
87 because when examining how bad parents may have negative influence on their offspring, few studies have examined effects of concurrent risk and intergenerational risk simultaneo usly. However, this issue could be of great policy significance. If further research confirms that when intergenerational risk exists, concurrent risk is no longer or becoming much less important, separating kids from their bad parents, for instance sendin g kids to foster homes, seems not an effective strategy to combat the problem of juvenile delinquency. Instead, those kids should be kept with their parents and establish affective ties with their parents. Again, this is only a tentative finding drawn from an exploratory study. A great number of empirical studies will be needed before any call for concrete policy changes. In addition, in both G3 childhood and adolescence baseline ntrol variables. Results from individual models supported the first hypothesis. For G3 childhood, we identified three promotive factors G2 affective ties to G3, G2 consistency of or G2 affective ties to G3. For G3 adolescence, we also identified the same three promotive factors, but with two significant protective factors G2 affective ties to G3 and G2 tection effect between promotive and protective factors is of great practical and policy importance because now we could differentiate and specifically implement protection or prevention strategies effective for the general public such as G2 consistency of discipline and strategies that would best handle the most delinquent ones. Virtually, G2 affective ties to G3 will not only additively reduce the chance for a child with bad parents
88 to become delinquent, but this protection effect will be more obvious for children with higher level of G2 risk. However, results from individual and comprehensive models are not very consistent. In both G3 childhood and adolescence comprehensive models, only G2 affective ties to G3 is statistically significant as a promotive as well as protective factor. One common explanation for this discrepancy could be the issue of multicollinearity However, this explanation seems not legitimate here: 1) bivariate correlations between all pairs of predictors were not particularly high; 2 ) the highest Variance Inflation Factors value 10 (e.g. Marquardt, 1970; Menard, 1995; Neter et al., 1989) and also below the 07); 3) after we take out the variable with the highest VIF score, standard errors of other variables in the model did not change significantly; 4) we regressed other variables onto the variable with the highest VIF score, R2 value is not very large. Thus, the possible explanation might not be the issue of multicollinearity effect of other parenting measures in comprehensive models. In other words, for instance, parental involve ment and consistency of discipline will have impact on how close G3 feel toward G2 and it is the closeness of affective ties that really matters in whether G3 would commit a crime or not. As shown in Chapter 3, questions in the Hudson scale (Hudson, 1982) on affective ties asked about how well parent and child get along, trust and understand one another; get angry or violent toward one another; enjoy or are proud of one another; and wish that the other was more like others they
89 know. It is obviously a very his/her parent. Thus, measures of effective parenting skills such as consistency of on G3 delinquent beh aviors. More advanced statistical techniques such as structural equation modeling (SEM) technique are needed to assess more complex relationship. As Kubrin et al. (2009) suggested, a good and close relationship with parents will result psychological presence even when the child is out of their direct control (p.170) Jang & Smith (1997) found that lack of supervision would lead to reduced affective ties between parent and child, though they argued that it is through the mechanism of de linquent behavior itself Moreover results from both individual and comprehensive models could have stimulate d original social bonding theory. Hirschi (1969) maintained that it is the fact of attachment to others not th e character of the people to whom one is attached that really matters. However, empirical evidence from research on peer and delinquency has largely on that The current research has provided some new evidence to support Hir Individual and comprehensive models provided much less support for the second hypothesis. Results from individual models showed that the number of significant comprehensive models, protective factor both in G3 childhood and adolescent years. Thus, hardly can we draw a conclusion that protection originated from good parent child relationship would weaken when the child b ecomes older. On the one hand, we may argue that G3
90 major sources of rewards and approval. Accordingly, protection from parents (G2) has weakened; on the other hand, we may also argue that G3 adolescents actually obtain more protection from parents (G2) because now parents realize that the range of risk childhood years. Further research is nee ded here. Limitations and Future Research Although the present study adds to the literature on intergenerational discontinuity of delinquency, it is not without its limitations. As we repeatedly mentioned, this thesis research is largely exploratory in na ture and gives only a preliminary glimpse at how parental (G2) promotive/protective factors may protect their offspring from delinquent behaviors against their own adolescent risk. First and foremost, G3 subjects studied in the Rochester Intergenerationa l Study were not a nationally representative sample. The findings of this thesis study were derived from a high risk, predominantly African American sample (G2) with their oldest biological child (G3). Thus, whether or not the findings could be generalize d to a larger scale is not yet clear. Replication is necessary to confirm the results and extend them beyond a single sample. Another potential limitation with data is the modest sample size and relatively high percentage of missing data in some parental m easures in G3 adolescence models. As mentioned earlier, the present study is intergenerational as well as longitudinal in nature. Cases were lost due to attrition over time. Although efforts were made to boost the sample size and not use listwise deletion method to handle missing data, the findings of multiple regression analysis might still be vulnerable to
91 measurement errors. As Kelley and Maxwell (2010) argued, this is particularly a problem when examining cross product or interaction terms. Another po ssible drawback of the present study is we did not include multiple caregivers in our analyses. For instance, we only examined how G2 affective ties to G3 may influence G3 delinquency, but not G1 affective ties to G3 or OCG affective ties to G3 within the same family. Intergenerational studies that did not incorporate data on multiple caregivers could be providing misleading estimates of the effects of the focal parents. This would be particularly true for G2 males since a high percentage of them did not ha include peer and community delinquent behaviors. Criminological theories and empirically studies have clearly demonstra behavior at least in adolescence and early adulthood (e.g. Akers, 1985; Loeber & Stouthamer Loeber, 1986; Liu, 2003; Thornberry et al., 2003). The same logic applies to neighbor hood and crime, though empirical support on this relationship is not as strong as that on delinquent peers and crime. Thus, in future studies of intergenerational resilience of delinquency, the influence of G1, OCG and/or G3 peer and community related fac tors should be incorporated. It is also necessary to point out that there exist minor differences between items in the measures of G3 delinquency and G2 adolescence risk (delinquency). Items in G3 delinquency measure are comparable to moderate and minor delinquency items in the measure of G2 adolescence risk. Very serious and violent delinquent behaviors such as attacked someone with a weapon or with the idea of seriously hurting or killing them
92 were not included in G3 delinquency measure because we creat ed a delinquency scale that is appropriate for both G3 children and adolescents. However, by using different measures of G2 adolescence risk and G3 delinquency, we are presented with the potential limitation of determining whether promotive/protective fact ors truly buffered G2 risk or it is simply a reflection of the changing measures. Last but not least, future research of intergenerational resilience of delinquency should attempt to incorporate genetic influence into analyses. Recent research on behavior al genetics highlights that children are differentially affected by their rearing experiences due to the so called differential susceptibility (Belsky et al., 2009; Belsky & individuals could also benefit the most from supportive environmental conditions such as healthy parent child relationship (Boyce & Ellis, 2005). In addition, Shanahan and Hofer (2005) and Rutter (2007) elaborated on a set of generic mechanisms by whi ch genes may interact with environmental circumstances to influence the occurrence of antisocial behavior across generations. It is possible that an important pathway of intergenerational discontinuity in delinquent behaviors might be attributable to gene environment interactions. Conclusion Does the apple fall far from the tree? Thornberry (2009) asked this question in the American Society of Criminology 2008 Sutherland Address. I believe that there is still no clear cut answer to this question. Limited r esearch has been done on intergenerational relationship in criminology and results from these studies were not consistent and varied by different outco me variables researchers used Studies like this, however, would hopefully lessen the degree to which res earchers and practitioners assume a
93 deterministic relationship in patterns of delinquent behaviors in successive generations. Indeed, the current study found that a good portion of children (G3) born to delinquent parents (G2) did not themselves exhibit hi gh levels of problem behaviors. In addition, such intergenerational transmission of delinquency was buffered by parental (G2) affective ties to children (G3). As such, it is hoped that this study could be used in support of any delinquency reduction progra ms that involve enhancing parent child relationship. If findings from current study are replicated in the future, juvenile parents and children should at least be modified.
94 APPENDIX A LIST OF SCALE ITEMS G3 General Delinquency 1. Cruel, bullying or mean to others? 2. 3. Gets into many fights 4. Hangs around with others who get in trouble 5. Lies or cheats 6. Physically attacks people 7. Runs away from home 8. Sets fires 9. Steals at home 10. Steals outside home 11. Swears or uses obscene language 12. Think about sex too much 13. Threatens people 14. Truant or skips schools 15. Uses alcohol or drugs for non medical purposes 16. Vandalizes 17. Destroys his/her own things 18. Destroys things belonging to his/her family or others G2 Adolescent General Delinquency 1. Run away from home? 2. Skipped classes without an excuse? 3. Lied about your age to get into some place or to buy something? (For example, lying about your age to get into a movie or buy alcohol) 4. Hitchhiked a ride with a stranger? 5. Carried a hidden weapon? 6. Been loud or rowdy in a public place where someone complained a nd you got in trouble? 7. Begged for money or things from strangers? 8. Been drunk in a public place? 9. Damaged, destroyed, marked up, or tagged somebody else's property on purpose? 10. Set fire on purpose or tried to set fire to a house, building, or car ? 11. Avoided paying for things, like a movie, taking bus rides, using a computer, or anything else? 12. Gone into or tried to go into a building to steal or damage something? 13. Tried to steal or actually stolen money or things worth $5 or less? 14. Tried to steal or actually stolen money or things worth $5 $50? 15. Tried to steal or actually stolen money or things worth between $50 $100? 16. Tried to steal or actually stolen money or things worth more than $100? 17. Tried to buy or sell things that w ere stolen?
95 18. Taken someone else's car or motorcycle for a ride without the owner's permission? 19. Stolen or tried to steal a car or other motor vehicle? 20. Forged a check or used fake money to pay for something? 21. Used or tried to use a credit card, bank card, or automatic teller card without permission? 22. Tried to cheat someone by selling them something that was not what you said it was or that was worthless? 23. Attacked someone with a weapon or with the idea of seriously hurting or killing them? 24. Hit someone with the idea of hurting them? 25. Been involved in gang or posse fights? 26. Thrown objects such as rocks or bottles at people? 27. Used a weapon or force to make someone give you money or things? 28. Made obscene phone calls? 29. Been pa id for having sexual relations with someone? 30. Physically hurt or threatened to hurt someone to get them to have sex with you? 31. Sold marijuana, reefer or pot? 32. Sold hard drugs such as crack, heroin, cocaine, LSD or acid? G2 Affective Ties to G3 1. Child is too demanding 2. Child interferes with your activities 3. You think child is terrific 4. You feel violent toward child 5. You feel very angry toward child 6. You feel proud of child 7. You wish child was more like other children that you k now 8. Child is well behaved 9. You get along well with child 10. You just do not understand child 1. How often do you read to your child 2. How often do you do things with child like playing, going to a movie or sporting event, or going out to eat 3. How often do you talk to child about what child did during the day 4. How often do you take child to visit friends or relatives 5. How often do you celebrate family events like birthday with child 6. How often do yo u take child to a play ground, park, or place to play with other children 7. How often are you too busy or unavailable to do things with child
96 G2 Concurrent Delinquency 1. Carried a hidden gun 2. Carried other hidden weapon 3. Been drunk in public place 4. Gone into or tried to go into a building to steal or damage something 5. Tried to steal or actually stolen money or things worth $50 or less 6. Tried to steal or actually stolen money or things worth over $50 7. Tried to b uy or sell stolen things 8. Tried to steal or actually stolen a car or other motor vehicle 9. Forged a check, used counterfeit money, or cashed or used bad checks on purpose 10. Used or tried to use a credit card, bank card, or automatic teller card wi thout permission 11. Attacked someone with a weapon or with the idea of seriously hurting or killing them 12. Hit someone with the idea of hurting them, for example, fist fighting 13. Used a weapon or force to make someone give you money or things 14. T ake part in illegal gambling, such as shooting dice, betting on cards, or playing the numbers 15. Driven while under the influence of drugs, beer, wine or liquor 16. Sold marijuana 17. Sold hard drugs such as crack, heroin, cocaine, LSD, or acid 18. Sold other drugs such as tranquilizers, speed, or downers 19. Sold drugs not pot G3 anger/frustration 1. Child gets angry when told he/she has to go to bed 2. 3. Child gets quite frustra ted when prevented from doing something he/she wants to do 4. 5. Child gets easily frustrated when tired 6. Child gets angry when called in from play before he/she is ready to quit G3 impulsivity 1. Child usually rushed into an activity without thinking about it 2. Child often rushed into new situations 3. Child takes a long time in approaching new situations 4. Child is slow and unhurried in deciding what to do next 5. Child slowly approached places where he/she might hurt himself/herself 6. Child is among the last children to try out a new activity
97 APPENDIX B DESIDERATA FOR MULTIPLE REGRESSION Desideratum 1. The goals of the research and how multiple regression (MR) can be useful are explicitly addressed 2. The inclusion of each of the independent variables should be justified on theoretical and/or practical grounds 3. Each criterion and regressor variable should be described in detail, including scales of mea surement, coding scheme, reliability etc. to convey how the MR model should be interpreted 4. Specific procedures for the computation and interpretation of effect sizes are delineated 5. Assumptions underlying the MR analyses and resulting inference are explicitly addressed 6. Variable selection techniques are justified 7. Sample sizes for all analyses are justified in terms of power, accuracy, and reproducibility of results 8. Methods of dealing with missing data are addressed 9. For models examining moderation, issues of interpretation, role of centering, and visualization are addressed 10. For models examining mediation, issues of interpretation and limitations due to cross sectional design are addressed 11. Visual examination of data is addressed i n order to assess model appropriateness and assumptions 12. Measurement error in predictor and/or outcome variable is addressed 13. Potential limitations of multiple regression in the current applied research context are explicitly stated 14. Alternativ es to the MR model are given
98 LIST OF REFERENCES Achenbach, T. M. (1991). Manual for the Child Be havior Checklist/4 18 and 1991 p rofile Burlington, VT: University of Vermont. Achenbach, T. M. (1992). Manual for the Child B ehavior Checklist/2 3 and 1992 p rofile Burlington, VT: University of Vermont. Achenbach, T. M., Howell, C. T., Quay, H. C., Conners, C. K., & Bates, J. E. (1991). National survey of problems and competencies amo ng four to sixteen year olds: P arents' reports for normative and clinical samples. Monographs of the Society for Research in Child Development, 56 (3), 1 130. Agresti, A., & Finlay, B. (2009). Statistical methods for the social s ciences (4th ed.). New Jersey: Prentice Hall. Aiken, L. S., & West, S. G. (1991). Multiple regression : Testing and interpreting i nteractions Newbury Park, Calif.: Sage. Akers, R. L. (1985). Deviant behavior: A social learning a pproach (3rd ed.). Belmont, CA: Wadsworth. Akers, R. L. (1998). Social learning and social structure: A general theory of crime a nd d eviance Boston: Northeastern University Press. Allison, P. D. (2002). Missing d ata Thousand Oaks, CA: Sage. Bailey, J. A., Hill, K. G., Oesterle, S., & Hawkins, J. D. (2006). Linking substance use and problem behavior across three generations. Journal of Abnormal Child Psychology 34 273 292. Bellair, P., & McNulty, T. (2010). Cognitive skills, adolescent violence and the moderating role of neighborhood disadvantage. Justice Quarterly 27 (4), 538 559. Belsky, J. (1984). The determinants of pare nting: A process model. Child Development 55 83 96. Belsky, J. (1993). Etiology of child maltreatment: A developmental ecological analysis. Psychological Bulletin 114 413 434. Belsky, J., Woodworth, S., & Crnic, K. (1996). Troubled family interaction during toddlerhood. Development and Psychopathology 8 477 495. Berlin, L. J., Malone, P. S., Ayoub, C., Ispa, J. M., Fine, M. A., Brooks Gunn, J., Brady Smith, C., & Bai, Y. (2009). Correlates and consequenc es of spanking and verbal punishment for low income white, African American, and Mexican American toddlers. Child Development 80 (5), 1403 1420.
99 Bijleveld, C., & Farrington, D. (2009). The importance of studies of intergenerational transmission of antisoci al behavior. Criminal Behavior and Mental Health 19 (2), 77 79. Bijleveld, C., & Wijkman, M. (2009). Intergenerational continuity in convictions: A five generation study. Criminal Behavior and Mental Health, 19 (2), 142 155. Bradley, R. H., & Corwyn, R. F. (2001). Home environment and behavioral development during adolescence: The mediating and moderating roles of self efficacy beliefs. Merrill Palmer Quarterly 47 165 187. Brooks Gunn, J., Duncan, G. J., Klebanov, P. K., & Sealand, N. (1993). Do neighborho ods influence child and adolescent development? American Journal of Sociology 99 353 395. Cairns, R. B., Cairns, B. D., Xie, H., Leung, M., & Hearne, S. (1998). Path across generations: Academic competence and aggressive behaviors in young mothers and th eir children. Developmental Psychology, 34 1162 1174. adjustment problems: I. A linear hypothesis II. Mothers at risk for transitions and unskilled parenting. Development al Psychology, 27 489 504. Capaldi, D. M., Pears, K. C., Kerr, D., & Owen, L. D. (2008). Intergenerational and Journal of Abnormal Child Psychology 36 347 358. Capaldi, D. M., Pears, K. C., Patterson, G. R., & Owen, L. D. (2003). Continuity of parenting practices across gen erations in an at risk sample: A prospective comparison of direct and mediated associations. Journal of Abnormal Child Psychology 31 (2), 127 142. Chassin, L., Presson, C. C., Todd, M., Rose, J. S., & Sherman, S. J. (1998). Maternal socialization of adolescent smoking: the intergenerational transmission of parenting and smoking. Developmental Psychology 34 1189 1201. Cicchetti, D., & Rogosch, F. A. (1996). Equifinality and multifinality in developmental psychopathology. Development and Psychopathology, 8 597 600. Cohen, J. (1968). Multiple regression as a general data analytic system. Psychological Bulletin, 70 426 443. Cohen, J. (1978). Partialed products are interactions; partialed vectors are curve components. Psychological Bulletin, 85 858 866. Cohen, J. (1983). The cost of dichotomization. Applied Psychological Measurement, 7 249 253
100 Conger, J. J. (1991). Adolescence and y outh: Psycholog ica l development in a changing w orld (4th ed.). New York: HarperCollins Publishers Conger, R. D., Ge, X., Elder, G. H., Lorenz, F. O., & Simons, R. L. (1994). Economic stress, coercive family process and developmental problems of adolescents. Child Developmen t 65 541 561. Cote, S. M., Vaillancourt, T., LeBlanc, J. C., Nagin, D. S., & Tremblay, R. E. (2006). The development of physical aggression from toddlerhood to pre adolescence: A nationwide longitudinal study of Canadian children. Journal of Abnormal Chi ld Psychology 34 (1), 71 85. Cramer, E. M., & Appelbaum, M. I. (1980). Nonorthogonal analysis of variance -once again. Psychological Bulletin, 87 51 57. C., Madden, P ., & Martin, N. G. (2007). Intergenerational transmission of childhood conduct problems Archives of General Psychiatry 53 243 249. Downey, G., & Coyne, J. C. (1993). Children of depressed parents: An integrative review. Psychological Bulletin 108 50 7 6. Ehrensaft, M. K., Wasserman, G. A., Verdelli, L., Greenwald, S., Miller, L. S., & Davies, M. (2003). Maternal antisocial behavior, parenting practices, and behavior problems in boys at risk for antisocial behavior. Jour nal of Child and Family Studies, 12 27 40. Elder, G. H. (1997). The life course and human development. In R. Lerner (Ed.), Handbook of c hild p sychology (Vol. 1). New York: Wiley. Elliot, D. S., Huizinga, D., & Menard, S. (1989). Multiple roblem youth: Delinquency substance use and m en tal health p roblems Springer. Elliott, D. S., Wilson, W. J., Huizinga, D., Sampson, R. J., Elliott, A., & Rankin, B. (1996). The effects of neighborhood disadvantage on adolescent development. Journal of Research in Crime and Delinquency 33 389 426. Farrin gton, D. (1993). Understanding and preventing bullying. In M. Tonry (Ed.), Crime & j ustice: A r eview of r esearch (Vol. 17, pp. 381 458). Chicago: University of Chicago Press. Farrington, D. (2009). The Cambridge Study in delinquent development. In C. Walst on (Ed.), Challenging c rime: A p ortrait of the Cambridge Institute of Criminology (pp. 64 71). London: Third Millennium. Farrington, D., Coid, J. W., & West, D. J. (2009). The development of offending from age 8 to age 50: Recent results from the Cambridge study in delinquent development. Journal of Criminology and Penal Reform 92 160 217.
101 Farrington, D., Gundry, G., & West, D. J. (1975). The familial transmission of criminality. Medicine, Science and the Law 15 177 218. Farrington, D., Lambert, S., & West, D. J. (1998). Criminal careers of two generations of family mem bers in the Cambridge Study in d elinquent d evelop ment. Studies on Crime and Crime Prevention 7 85 106. Farrington, D., Loeber, R., Jolliffe, D., & Pardini, D. A. (2008). Promotive and risk processes at different life stages. In R. Loeber (Ed.), Violence and s erious t heft (pp. 169 229). New York: Routle dge. Ferguson, C. J. (2009). An effect size primer: A guide for clinicians and researchers. Professional Psychology: Research and Practice 40 532 538. Finney, J. W., Mitchell, R. E., Cronkite, R. C., & Moos, R. H. (1984). Methodological issues in estimat ing main and interactive effects: examples from coping/social support and stress field. Journal of Health and Social Behavior, 25 85 98. Fraser, M. W., Kirby, L. D., & Smokowski, P. R. (2004). Risk and resilience in childhood. In M. W. Fraser (Ed.), Risk and r esilience in c hildhood: An e cological p erspective (2nd ed., pp. 13 66). Washington D. C.: NASW Press. Furstenberg, F. F., Levine, J. A., & Brooks Gunn, J. (1990). The children of teenage mothers: patterns of early childbearing in two generational. Family Planning Perspectives 22 54 61. Garmezy, N. (1995). Stress resistant children: the search for protective factors In J. E. Stevenson (Ed.), Aspects of c urrent c hild p sychiatry r esearch (pp. 213 233). Oxford, England: Pergamon Garnier, H. E., & Ste in, J. A. (2002). An 18 year model of family and peer effects on adolescent drug use and delinquency. Journal of Youth and Adolescent Psychiatry 31 (1), 45 56. Gecas, V., & Seff, M. A. (1990). Social class and self esteem: Psychological centrality, compens ation, and the relative effects of work and home. Social Psychology Quarterly 53 165 173. Gordon, R. A. (1987). SES versus IQ in the race IQ delinquency model. International Journal of Sociology 7 30 96. Gottfredson, M. R., & Hirschi, T. (1990). A gene ral theory of c rime Stanford, CA: Stanford University Press. Gross, D. F., L., Young, M., Ridge, A., Cowell, J. M., Richardson, R., & Sivan, A. (2006). The equivalence of the Child Behavior Checklist/1 1/2 5 across parent race/ethnicity, income level, and language. Psychological Assessment, 18 (3), 313 323.
102 Hagan, J., & Palloni, A. (1990). The social reproduction of a criminal class in working class London, Circa 1950 1980. American Journal of Sociology 96 (2), 265 299. Harrell, F. E. (2001). Regression mod eling st rategies New York: Springer. Hawkins, J. D., Catalano, R. F., & Miller, J. Y. (1992). Risk and protective factors for alcohol and other drug problems in ado lescence, and early adulthood: I mplications for substance abuse prevention. Psychological Bulletin 112 64 105. Hirschi, T. (1969). Causes of d elinquency Berkeley: University of Carlifornia Press. Hirschi, T., & Gottfredson, M. R. (1983). Age and the explanation of crime. American Journal of Sociology, 89 552 584. Hirschi, T., & Hindelang, M J. (1977). Intelligence and delinquency: A revisionist review. American Sociological Review, 42 571 587. Hops, H., Biglan, A., herman, L., Arthur, J., Friedman, L., & Osteen, V. (1987). Home observations of family interactions of depressed women. Journa l of Consulting and Clinical Psychology 55 341 346. Hops, H., Davis, B., Leve, C., & Sheeber, L. (2003). Cross generational transmission of aggressive parent behavior: a prospective, meditational examination. Journal of Abnormal child Psychology 31 161 169. Hudson, W. H. (1982). The clinical measurement package: A field m anual Homewood, IL: Dorsey Press. Hudson, W. H. (1996). WALMYR a sses sment scales scoring m anual Tempe, AZ: WALMYR Publishing Company. Huesmann, L. R., Eron, L. D., Lefkowitz, M. M., & Walder, L. O. (1984). Stability of aggression over time and generations. Developmental Psychology 6 1120 1134. Huesmann, L. R., Moise Titus, J., Podolski, C. L., & Eron, L. D. (2003). Longitudinal their aggressive and violent behavior in young adulthood. D evelopmental Psychology 39 (2), 201 221. Irwin, J. R., & McClelland, G. H. (2003). Negative consequences of dichotomizing continuous predictor variables Journal of Marketing Research, XL (August), 366 371. Jang, S. J., & Smith, C. A. (1997). A test of reciprocal causal relationships among parental supervision, affective ties and delinquency. Journal of Research in Crime and Delinquency, 34 307 336.
103 Kaplan, D. D., Kaplan, H. B., & Liu, X. (2000). Family structure and parental involvement in the intergenerational parallelism of school adversity. Journal of Educational Research 93 235 244. Kaplan, H. B., & Liu, X. (1999). Explaining transgenerational continuity in antisocial behavi or during early adolescence. In P. Cohen, S. C. & L. N. Robins (Eds.), Historical and g eographical i nfluences on p sychopathology (pp. 163 191). Mahwah, NJ: Lawrence Erlbaum Associates. Kaufman, J., & Zigler, E. (1992). The prevention of child maltreatment: Programming, research, and policy. In D. J. Willis, E. W. Holden & M. Rosenberg (Eds.), Pre vention of child maltreatment: D evelopmental and ecological perspective New York: John Wiley & Sons. Kelley K., & Maxwell, S. E. (2010). Multiple regression. In G. R. Hancock & R. O. Mueller (Eds.), The r eviewer's g uide to q uantitative m ethods in the s ocial s ciences New York: Routledge. Kim, H. K., Capaldi, D. M., Pears, K. C., Kerr, D. C., & Owen, L. D. (200 9). Intergenerational Transmission of internalizing and externalizing behaviors across three generations: Gender specific pathways. Criminal Behavior and Mental Health 19 (2), 125 141. Kotch, J. B., Browne, D. C., Dufort, V., Winsor, J., & Catellier, D. (1 999). Predicting child maltreatment in the first 4 years of life from characteristics assessed in the neonatal period. Child Abuse and Neglect, 23 305 319. Krohn, M. D., Lizotte, A. J., & Perez, C. M. (1997). The interrelationship between substance use and precocious transitions to adult statuses. Journal of Health and Social Behavior 38 87 103. Kubrin, C., Stucky, T., & Krohn, M. D. (2009). Researching theories of crime and d eviance Ne w York: Oxford University Press. Kurlychek, M. C., Krohn, M. D., Dong, B., Hall, G. P., & Lizotte, A. J. (Forthcoming). Protection from risk: an exploration of when and how neighborhood level factors can reduce violent youth outcomes. Youth Violence and Ju venile Justice Labouvie, E., & McGee, C. (1986). Relation of personality to alcohol and drug use in adolescence. Journal of Consulting and Clinical Psychology 54 289 293. Lee, C. L., & Bates, J. E. (1985). Mother child interaction at two years and perce ived difficult temperament. Child Development, 56 1314 1325. Little, R. A., & Rubin, D. B. (2002). Statistical analysis with m issing data (2nd ed.). New York: Wiley.
104 Liu, X. (2003). The moderating effects of internal and perceived external sanction threat s on the relationship between deviant peer associations and criminal offending Western Criminology Review, 4 (3), 191 202 Loeber, R., & Farrington, D. (2000). Yo ung children who commit crime: E pidemiology, developmental origins, risk factors, early interv entions, and policy implications. Development and Psychopathology 12 737 762. Loeber, R., & Stouthamer Loeber, M. (1986). Family factors as correlates and predictors of juvenile conduct problems and delinquency. Crime and Justice 7 29 149. Lovegrove, P J. (2010). Explaining discontinuous cross generational patterns of d elinquency .Unpublished manuscript. Luthar, S. S., Cicchetti, D., & Becker, B. (2000) The construct of resilience: A critical evaluation and guidelines for future work. Child Development, 71 (3), 543 562. Lynam, D. R., Moffitt, T., & Southamer Loeber, M. (1993). Explaining the relation between IQ and delinquency: Class, race, test motivation, school failure, or self control? Journal of Abnormal Psychology 102 187 196. Mano, K. E., Davies, H., Klein Tasman, B. P., & Adesso, V. J. (2009). Measurement equivalence of the child behavior checklist among parents of African American adolescents. Journal of Child and Family Studies, 18 (5), 606 620. Marquardt, D. W. (1970). Generalized i nverses, ridge regression, biased linear estimation, and nonlinear estimation. Technometrics, 12 591 612. Marquardt, D. W. (1980). You should standardize the predictor variables in your regression models Journal of the American Statistical Association, 7 5 87 91. Masten, A. S., Best, K., & Garmezy, N. (1990). Resilience and development: Contributions from the study of children who overcome adversity. Development and Psychopathology, 2 425 444. Masten, A. S., Hubbard, J. J., Gest, S. D., Tellegen, A., Gar mezy, N., & Ramirez, M. (1999). Adaptation in the context of adversity: Pathways to resilience and mal adptation from childhood to late adolescence. Development and Psychopathology, 11 143 169. McClendon, M. J. (1994). Multiple regression and causal a naly sis Long Grove, Illinois: Waveland Press. McCord, J. (1991). The cycle of crime and socialization practices. Journal of Criminal Law and Criminology 82 211 228.
105 Meng, X., & Rubin, D. B. (1992). Performing likelihood ratio tests with multiply imputed dat a sets. Biometrika, 79 (1), 103 111. Moffitt, T. (1990). The neuropsychology of delinquency: A critical review. In M. Torny & N. Morris (Eds.), Crime and j ustice: A r eview of r esearch (pp. 99 169). Chicago: University of Chicago Press. Moffitt, T. (1993). Life course persistent and adolescence limited antisocial behavior: A developmental taxonomy. Psychological Review, 100 674 701. Moffitt, T. E. (2005). The new look of behavioral genetics in developmental psychopathology: Gene environm ent interplay in antisocial behaviors. Psychological Bulletin 131 533 554. Moffitt, T. E., Caspi, A., Rutter, M., & Silva, P. (2001). Sex d ifferences in a ntisocial b ehavior: Conduct d isorder, d elinquency, and v iolence in the Dunedin l ongitudinal s tudy Ne w York: Cambridge University Press. Neter, J., Wasserman, W., & Kuntner, M. H. (1989). Applied linear regression m odels Homewood, IL: Irwin. Nye, F. I. (1958). Family relationships and delinquent b ehavior New York: Wiley. O'Brien, R. M. (2007). A caution regarding rules of thumb for variance inflation factors. Quality & Quantity, 41 673 690. and the causes of youth crime: A deceptively useful analysis? Youth Justice 9 (2), 99 114. Osgood, D. W. (2005). Making sense of crime and the life course. Annals of the American Academy of Political and Social Science 602 196 211. Patterson, G. R. (1998). Continuities a search for causal mechanisms: Comment on the special section. Developmental Psychology 34 1263 1268. Patterson, G. R., DeBaryshe, B. D., & Ramsey, E. (1989). A developmental perspective on antisocial behavior. The American Psychologist 44 329 335. Patterson, G. R., Reid, J. B., & Dishion, T. J. (1992). Antisocial b oys Eugene, OR: Castalia. Pears, K. C., & Capaldi, D. M. (2001). Intergenerational transmission of abuse: A two generational prospective study of an at risk sample. Child Abuse & Neglect: The International Journal 25 1439 1461. Pedhazur, E. J. (1982). Multiple regression in behavioral research : Explanation and p rediction (2nd ed.). New York: Holt, Rinehart, and Winston.
106 Quintano, C., Castellano, R., & Rocca, A. (2010). Influences of outliers on some multiple imputation methods. Metodoloski Zvezki 7 (1), 1 16. Reckless, W. C. (1961). The crime p roblem (3rd ed.). New York: Appleton Century Crofts. Reiss, A. J. (1951). Delinquency as the failure of persona l and social control. American Sociological Review, 16 196 207. Reppucci, N. D., Fried, C. S., & Schmidt, M. D. (2002). Youth violence: risk and protective factors. In R. R. Corrado (Ed.), Multi p roblem v iolent y outh: A f oundation for c omparative r esearch on n eeds, i nterventions, and o utcomes (pp. 3 22). Washington, D. C.: IOS Press. Rhee, S. H., & Waldman, I. D. (2002). Genetic and environmental influences on antisocial behavior: A meta analysis of twin and adoption studies. Psychological Bulletin, 128 409 529. Ribeaud, D., & Eisner, M. (2010). Risk factors for aggression in pre adolescence: Risk domains, cumulative risk and gender differences. European Journal of Criminology, 7 (6), 460 498. Richman, J. M., & Fraser, M. W. (2 001). Resilience in childhood : T he role of risk and protection. In M. W. Fraser & J. M. Richman (Eds.), The c ontext of y outh r i olence: Resilience, r isk, and p rotection (pp. 1 12): Praeger. Rose, R. A., & Fraser, M. W. (2008). A simplified framework for using multiple imputation in soc ial work research. Social Work Research, 32 (3), 171 178. Rutter, M. (1985). Resili ence in the face of adversity: P rotective factors and resistance to psychiatric disorder. British Journal of Psychiatry 147 598 611. Rutter, M. (1987). Psychological resili ence and protective mechanisms A merican Journal of Orthopsychiatry, 57 316 331. Rutter, M., & Quinton, D. (1984). Parental psychiatric disorder: Effects on children. Psychological Medicine 14 853 880. Schafer, J. L., & Graha m, J. W. (2002). Missing data: O ur view of the state of the art. Psychological Methods, 7 (2), 147 177. Serbin, L. A., Cooperman, J. M., Peters, L. L., Lehoux, P. M., Stack, D. M., & Schwartzman, A. E. (1998). Intergenerational transfer of psychosocial risk in women with childhood histories of aggression, withdrawal, or aggression and withdrawal. Developmental Psychology 34 1246 1262.
107 Serbin, L. A., & Karp, J. (2004). The intergenerational transfer of psychosocial risk: Mediators of vulnerability and Resil ience. Annual Review of Psychology 55 333 363. Smith, C. A., & Farrington, D. P. (2004). Continuities in antisocial behavior and parenting across three generations. Journal of Child Psychology & Psychiatry & Allied Disciplines 45 230 248. Smith, C. A., Krohn, M. D., Chu, R., & Best, O. (20 05). African American fathers: M yths and realities about their involvement with their firstborn children. Journal of Family Issues, 26 975 1001. Smith, C. A., Lizotte, A. J., Thornberry, T. P., & Krohn, M. D. (1995). Resilient youth: Identifying factors that prevent high risk youth from engaging in delinquency and drug use. In J. Hagan (Ed.), Delinquency and d isrepute in the l ife c ourse (pp. 217 247). Greewich, CT: JAI Press. Smith, J. R., & Brooks Gunn, J. (1997). Correlates and consequences of harsh discipline for young children. Archives of Pediatric and Adolescent Medicine 151 777 786. Smith, K. W., & Sasaki, M. S. (1979). Decreasing multicollinearity: A method for models with multiplic ative functions. Sociological Methods and Research 8 35 56. Snyder, J. (1991). Discipline as a mediator of the impact of maternal stress and mood on child conduct problems. Development and Psychopathology 3 263 276. Sterba, S. K., Prinstein, M. J., & Cox, M. J. (2007). Trajectories of internalizing problems acros s childhood: H eterogeneity, external validity and gender differences. Development and Psychopathology, 19 345 366. Stouthamer Loeber, M., Loeber, R., Farrington, D. P., Zhang, Q., Van Kammen, W. B., & Maguin, E. (1993). The double edge of protective and risk factors for delinquency: Interrelations and developmental patterns. Development and Psychopathology, 5 683 701. Stouthamer Loeber, M., Loeber, R., Wei, E., Farrington, D. P., & Wikstrom, P H. (2002). Risk and promotive effects in the explanation of persistent serious delinquency in boys. Journal of Consulting and Clinical Psychology 70 111 124. Straus, M. A. (1994). Beating the devil out of them: Corporal punishment in American f amilies New York: Lexington Books. Tate, R. L. (1984). Limitations of centering for interactive models. Sociological Methods and Research, 13 251 271.
108 Taylor, J. Y., Washington, O. G., Artinian, N. T., & Lichtenberg, P. (2007). Parental stress among African American parents and grandparents. Issues in Mental Health Nursing, 28 (4), 373 387. Thornberry, T. P. (1987). Toward an interactional theory of delinquency. Criminology, 25 863 891. Thornberry, T. P. (2005). Explaining multiple patterns of offending across life course and across generations. Annals of the American Academy of Political and Social Science 602 156 195. Interg enerational patterns of antisocial behavior the American Society of Criminology 2008 Sutherland Address. Criminology, 47 (2), 297 325. Thornberry, T. P., Freeman Gallant, A., Lizotte, A. J., Krohn, M. D., & Smit h, C. A. (2003). Linked lives: T he intergenera tional transmission of antisocial behavior. Journal of Abnormal Child Psychology 31 171 184. Thornberry, T. P., Freeman Gallant, A., & Lovegrove, P. J. (2009a). Intergenerational linkages in antisocial behavior. Criminal Behavior and Mental Health 19 8 0 93. Thornberry, T. P., Freeman Gallant, A., & Lovegrove, P. J. (2009b). The impact of parental stressors on the intergenerational transmission of antisocial behavior. Journal of Youth and Adolescence 38 312 322. Thornberry, T. P., & Krohn, M. D. (2001) The development of delinquency: An interactional perspective. In S. O. White (Ed.), Handbook of y outh and j ustice (pp. 289 305). New York: Plenum. Thornberry, T. P., & Krohn, M. D. (2005). Applying interactional theory to the explanation of continuity an d change in antisocial behavior. In D. P. Farrington (Ed.), Integra ted d evelopmental & l ife c ourse t heories of o ffending (pp. 183 209). New Brunswick, NJ: Transaction Publishers. Thornberry, T. P., Krohn, M. D., & Freeman Gallant, A. (2006). Intergenerational roots of early onset substance use. Journal of Drug Issues, 36 1 28. Thornberry, T. P., Lizotte, A. J., Krohn, M. D., Farnworth, M., & Jang, S. J. (1991). Testing interactional theory: An examination of recip rocal causal relationships among family, school, and delinquency. Journal of Criminal Law and Criminology, 82 3 35. Thornberry, T. P., Lizotte, A. J., Krohn, M. D., Farnworth, M., & Jang, S. J. (1994). Delinquent peers, beliefs, and delinquent behavior: A longitudinal test of interactional theory. Criminology, 32 601 637.
109 Thornberry, T. P., Smith, C. A., & Howard, G. J. (1997). Risk factors for teenage fatherhood Journal of Marriage and the Family 59 505 522. Tolan, P., Gorman Smith, D., & Henry, D. (2 006). Family violence. Annual Review of Psychology 57 557 583. Velleman, R. (1992). Intergenerational effects a review of environmentally oriented studies concerning the relationship between parental alcohol problems and family disharmony in the genesis of alcohol and other problems II: The intergenerational effects of family disharmony. International Journal of the Addictions 27 (4), 367 389. Verona, E., & Sachs Ericsson, N. (2005). The intergenerational transmission of externalizing be haviors in adult p articipants: T he mediating role of childhood abuse. Journal of Consulting and Clinical Psychology 73 1135 1145. Vitulano, M. L., Fite, P., J., & Rathert, J., L. (2010). Delinquent peer influence on childhood delinquency: The moderating effect of impulsivi ty. Journal of Psychopathology and Behavioral Assessment, 32 315 322. Walters, G. D. (1992). A meta analysis of the gene crime relationship. Criminology, 30 596 613. Washington, J. A., & Craig, H. K. (1999). Performances of at risk, African American pres choolers on the Peabody Picture Vocabulary Test III. Language, Speech, and Hearing Services in Schools 30 75 82. West, D. J., & Farrington, D. P. (1973). Who becomes d elinquent? London: Heinemann. West, D. J., & Farrington, D. P. (1977). The d elinquen t way of life: Third r eport of the Cambridge Study in Delinquent Development London: Heinemann. White, H. (1980). A heteroskedasticity consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica, 48 817 838. Wolfgang, M. E ., Thornberry, T. P., & Figlio, R. M. (1987). From b oy to m an, from d elinquency to c rime Chicago: University of Chicago Press. Zufferey, A., Michaud, P., Jeannin, A., Berchtold, A., Chossis, I., Melle, G. V., & Suris, J. C. (2007). Cumulative risk factors for adolescent alcohol misuse and its perceived consequences among 16 to 20 year old adolescents in Switzerland Preventive Medicine, 45 233 239. Zuravin, S., McMillen, C., DePanfilis, D., & Risley Curtiss, C. (1996). The intergenerational cycle of child maltreatment. Journal of Interpersonal Violence 11 315 334.
110 BIOGRAPHICAL SKETCH Beidi Dong attended college at the University of Hong Kong from 2005 to 2008. He graduated in August 2008, receiving his Bachelor of Social Sciences degree with a major in sociology and a minor in politics and public administration. His undergraduate resea rch investigated the delinquency problem of left over children in rural China. Currently, he is a graduate student in the Department of Sociology and Criminology and Law in the Division of Criminology, Law and Society at the University of Florida where he plans to continue his PhD study. His research interests include life course and developmental criminology, juvenile delinquency and quantitative research methods.