Citation
Fatherhood and Crime: Examining Life-Course Transitions among Men in Harlem

Material Information

Title:
Fatherhood and Crime: Examining Life-Course Transitions among Men in Harlem
Creator:
TRIPP, BRADLEY G. ( Author, Primary )
Copyright Date:
2008

Subjects

Subjects / Keywords:
Attachment behavior ( jstor )
Criminal behavior ( jstor )
Criminal punishment ( jstor )
Criminals ( jstor )
Drug design ( jstor )
Employment ( jstor )
Fatherhood ( jstor )
Fathers ( jstor )
Mathematical dependent variables ( jstor )
Modeling ( jstor )

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright Bradley G. Tripp. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Embargo Date:
5/31/2008
Resource Identifier:
660033934 ( OCLC )

Downloads

This item is only available as the following downloads:


Full Text

PAGE 1

FATHERHOOD AND CRIME: EXAMINI NG LIFE COURSE TRANSITIONS AMONG MEN IN HARLEM By BRADLEY G. TRIPP A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2007

PAGE 2

Copyright 2007 by Bradley G. Tripp

PAGE 3

iii ACKNOWLEDGMENTS I would like to thank many people fo r the various forms of support and encouragement that they have provided through graduate school and life in general. First, I would like to thank my parents. Next, I want to th ank my fianc Kristen. I would also like to thank all of my friends who are too numerous to list, but can be found in the ranks of the 4 Horsemen, the Lexington 12, a nd the ELRW. I would also like to thank the Newbomb Turk memorial library for the vast resources they have accorded me. There are many scholars who have helped me to complete this journey. First I want to thank my friend and mentor Connie Shehan. I would like to thank the members of my committee: Jodi Lane, Lonn Lanza Kaduce, Alex Piquero, and Joe Spillane.

PAGE 4

iv TABLE OF CONTENTS page ACKNOWLEDGMENTS.................................................................................................iii LIST OF TABLES...............................................................................................................x ABSTRACT.....................................................................................................................xiii CHAPTER 1 INTRODUCTION........................................................................................................1 2 REVIEW OF LITERATURE.......................................................................................3 Social Forces, Social Bonds.........................................................................................3 Employment and Education..................................................................................4 Corrections............................................................................................................7 Family..................................................................................................................12 Race, Family, Crime, and Money...............................................................................15 The Influence of Race.........................................................................................15 Families in the Lives of Men...............................................................................16 Adolescent Fathers..............................................................................................18 Prisons, Providing, and Problems........................................................................19 3 THEORY....................................................................................................................22 Social Theories on Crime...........................................................................................22 Social Control......................................................................................................22 Social Capital.......................................................................................................24 Life Course Theories..................................................................................................25 Trajectories in the Life Course............................................................................26 Transitions...........................................................................................................27 Turning Points.....................................................................................................28 Cohorts and Period Effects..................................................................................28 Life Course Criminology............................................................................................29 Criminal Careers in the Life-Course...................................................................29 The Age Crime Curve and the Criminal Career Debate......................................31 Gottfreson and Hirschi........................................................................................32 Blumstein, Cohen, and Farrington.......................................................................33 Developmental and Life Course Criminology....................................................34

PAGE 5

v DLC Issues and Measurements...........................................................................35 Desistance.....................................................................................................35 Past and future criminal behavior.................................................................37 Different DLC Theories......................................................................................39 Cummulative Disadvantage.................................................................................43 Major DLC Findings...........................................................................................45 Family...........................................................................................................45 Social Participation......................................................................................47 4 METHODS.................................................................................................................49 Methodological Issues................................................................................................49 Longitudinal Data................................................................................................49 Self Report Data..................................................................................................51 Longitudinal and Self-Report Data: Methodological Issues......................................53 Sampling..............................................................................................................54 Sampling Attrition...............................................................................................56 Reliability and Validity.......................................................................................57 Testing, Panel, and Period Effects.......................................................................58 Construct Continuity...........................................................................................59 Methods of Research..................................................................................................59 Data Source.........................................................................................................59 Data Collection....................................................................................................60 Sample.................................................................................................................63 Research Questions: Drug Use............................................................................64 Age...............................................................................................................64 Family: Marriage.......................................................................................... 64 Family: Fatherhood......................................................................................64 Residential Status.........................................................................................65 Employment.................................................................................................65 Finances........................................................................................................66 Education......................................................................................................66 Prior Behavior..............................................................................................66 Research Questions: Arrests................................................................................67 Age...............................................................................................................67 Family: Marriage..........................................................................................67 Family: Fatherhood......................................................................................67 Residential Status.........................................................................................68 Employment.................................................................................................68 Finances........................................................................................................69 Education......................................................................................................69 Prior Behavior..............................................................................................69 Research Questions: Incarceration......................................................................70 Age...............................................................................................................70 Family: Marriage..........................................................................................70 Family: Fatherhood......................................................................................70 Residential Status.........................................................................................71

PAGE 6

vi Employment.................................................................................................71 Finances........................................................................................................71 Education......................................................................................................72 Prior Behavior..............................................................................................72 Analysis...............................................................................................................73 Cross Sectional and Lagged Models............................................................ 77 Desistance Models........................................................................................77 Model and Variable Descriptions........................................................................78 Wave 1 independent variables predic ting wave 2 dependent variables.......79 Wave 2 independent variables predic ting wave 2 dependent variables.......82 Wave 2 independent variables predic ting wave 3 dependent variables.......85 Wave 3 independent variables predic ting wave 3 dependent variables.......86 Wave 3 independent variables predic ting wave 4 dependent variables.......90 Wave 4 independent variables predic ting wave 4 dependent variables.......91 Wave 4 independent variables predic ting wave 5 dependent variables.......93 Wave 5 independent variables predic ting wave 5 dependent variables.......95 Desistance.....................................................................................................97 Hypotheses..........................................................................................................99 Hypotheses: Drug Use.........................................................................................99 Age...............................................................................................................99 Family: Marriage..........................................................................................99 Family: Fatherhood....................................................................................100 Residential status........................................................................................101 Employment...............................................................................................102 Finances......................................................................................................102 Education....................................................................................................103 Prior Behavior............................................................................................103 Hypotheses: Arrests...........................................................................................104 Age.............................................................................................................104 Family: Marriage........................................................................................104 Family: Fatherhood....................................................................................104 Residential Status.......................................................................................106 Employment...............................................................................................106 Finances......................................................................................................106 Education....................................................................................................107 Prior deviance.............................................................................................107 Hypotheses: Incarceration.................................................................................108 Age.............................................................................................................108 Family: Marriage........................................................................................108 Family: Fatherhood....................................................................................109 Residential Status.......................................................................................110 Employment...............................................................................................110 Finances......................................................................................................111 Education....................................................................................................111 Prior behavior.............................................................................................112 Attrition.............................................................................................................112

PAGE 7

vii 5 RESULTS: DRUG USE...........................................................................................118 Cross-Sectional and Lagged Models........................................................................118 Wave 1 Independent Variables Predicting Wave 2 Drug Use..........................119 Wave 2 Independent Variables Predicting Wave 2 Drug Use..........................120 Wave 2 Independent Variables Predicting Wave 3 Drug Use..........................121 Wave 3 Independent Variables Predicting Wave 3 Drug Use..........................123 Strength of Social Bonds: Fatherhood.......................................................123 Strength of Social Bonds: Marriage...........................................................126 Wave 3 Independent Variables Predicting Wave 4 Drug Use..........................126 Strength of Social Bonds: Fatherhood.......................................................129 Strength of Social Bonds: Marriage...........................................................129 Wave 4 Independent Variables Predicting Wave 4 Drug Use..........................130 Wave 4 Independent Variables Predicting Wave 5 Drug Use..........................132 Wave 5 Independent Variables Predicting Wave 5 Drug Use..........................132 Strength of Social Bonds: Marriage...........................................................134 Desistance Models....................................................................................................135 Desistance 2: Wave 2 Independent Va riables Predicting Desistance Wave 3..136 Desistance 2: Wave 3 Independent Va riables Predicting Desistance Wave 3..137 Desistance 2: Wave 2 Independent Variables Predicting Desistance Waves 3-4......................................................................................................138 Desistance 2: Wave 3 Independent Variables Predicting Desistance Waves 3-4......................................................................................................139 Desistance 2: Wave 2 Independent Variables Predicting Desistance Waves 3-5......................................................................................................140 Desistance 2: Wave 3 Independent Variables Predicting Desistance Waves 3-5......................................................................................................142 Desistance 3: Wave 3 Independent Variables Predicting Desistance Wave 4...........................................................................................................142 Desistance 3: Wave 4 Independent Variables Predicting Desistance Wave 4...........................................................................................................144 Desistance 3: Wave 3 Independent Variables Predicting Desistance Wave 4-5........................................................................................................145 Desistance 3: Wave 4 Independent Variables Predicting Desistance Wave 4-5........................................................................................................145 6 RESULTS: ARRESTS.............................................................................................148 Cross-sectional and Lagged Models.........................................................................148 Wave 1 Independent Variables Predic ting Wave 2 Dependent Variables........149 Wave 2 Independent Variables Predic ting Wave 2 Dependent Variables........150 Wave 2 Independent Variables Predic ting Wave 3 Dependent Variables........150 Wave 3 Independent Variables Predic ting Wave 3 Dependent Variables........151 Strength of Social Bonds: Fatherhood.......................................................153 Strength of Social Bonds: Marriage...........................................................154 Wave 3 Independent Variables Predic ting Wave 4 Dependent Variables........154 Wave 4 Independent Variables Predic ting Wave 4 Dependent Variables........154

PAGE 8

viii Wave 4 Independent Variables Predic ting Wave 5 Dependent Variables........154 Wave 5 Independent Variables Predic ting Wave 5 Dependent Variables........156 Strength of Social Bonds: Marriage...........................................................157 Desistance Models....................................................................................................158 Desistance 2: Wave 2 Independent Va riables Predicting Desistance Wave 3..159 Desistance 2: Wave 3 Independent Va riables Predicting Desistance Wave 3..160 Desistance 2: Wave 2 Independent Variables Predicting Desistance in Waves 3-5......................................................................................................161 Desistance 2: Wave 3 Independent Variables Predicting Desistance in Waves 3-5......................................................................................................162 Desistance 3: Wave 3 Independent Variables Predicting Desistance in Waves 4-5......................................................................................................163 Desistance 3: Wave 4 Independent Variables Predicting Desistance in Waves 4-5......................................................................................................165 7 RESULTS: INCARCERATION..............................................................................167 Cross-sectional and Lagged Models.........................................................................167 Wave 1 Independent Variables Predic ting Wave 2 Dependent Variables........168 Wave 2 Independent Variables Predic ting Wave 2 Dependent Variables........168 Wave 2 Independent Variables Predic ting Wave 3 Dependent Variables........169 Wave 3 Independent Variables Predic ting Wave 3 Dependent Variables........170 Strength of Social Bonds: Fatherhood.......................................................172 Strength of Social Bonds: Marriage...........................................................173 Wave 3 Independent Variables Predic ting Wave 4 Dependent Variables........173 Strength of Social Bonds: Fatherhood.......................................................175 Strength of Social Bonds: Marriage...........................................................176 Wave 4 Independent Variables Predic ting Wave 4 Dependent Variables........176 Wave 4 Independent Variables Predic ting Wave 5 Dependent Variables........178 Wave 5 Independent Variables Predic ting Wave 5 Dependent Variables........180 Strength of Social Bonds: Marriage...........................................................180 Desistance Models....................................................................................................180 Desistance 2: Wave 2 Independent Va riables Predicting Desistance Wave 3..181 Desistance 2: Wave 3 Independent Va riables Predicting Desistance Wave 3..182 Desistance 2: Wave 2 Independent Variables Predicting Desistance Waves 3-4......................................................................................................184 Desistance 2: Wave 3 Independent Variables Predicting Desistance Waves 3-4......................................................................................................184 Desistance 2: Wave 2 Independent Variables Predicting Desistance in Waves 3-5......................................................................................................185 Desistance 2: Wave 3 Independent Variables Predicting Desistance in Waves 3-5......................................................................................................186 Desistance 3: Wave 3 Independent Variables Predicting Desistance in Waves 4..........................................................................................................188 Desistance 3: Wave 4 Independent Variables Predicting Desistance in Waves 4..........................................................................................................189

PAGE 9

ix Desistance 3: Wave 3 Independent Va riables Predicting Desistance in Waves 4-5..................................................................................................................190 Wave 4 Independent Variables Pred icting Desistance in Waves 4-5................191 8 DISCUSSION...........................................................................................................194 Introduction...............................................................................................................194 Answering Research Questions................................................................................195 Age.............................................................................................................195 Family: Marriage........................................................................................196 Family: Fatherhood....................................................................................198 Residential Status.......................................................................................203 Employment...............................................................................................206 Finances......................................................................................................208 Education....................................................................................................213 Prior Deviance............................................................................................215 Future Research........................................................................................................226 APPENDIX......................................................................................................................230 LIST OF REFERENCES.................................................................................................234 BIOGRAPHICAL SKETCH...........................................................................................258

PAGE 10

x LIST OF TABLES Table page 4-1 Harlem Longitudinal Study of Urban Black Youth Sample....................................61 4-2 Arrest and Incarceration Years fo r Cross-sectional and Lagged Models.................75 4-3 Years for Desist 2 Models........................................................................................76 4-4 Years for Desist 3 Models........................................................................................76 5-1 Wave 1 Independent Variable s Predicting Wave 2 Drug Use...............................119 5-2 Wave 2 Independent Variable s Predicting Wave 2 Drug Use...............................121 5-3 Wave 2 Independent Variable s Predicting Wave 3 Drug Use...............................122 5-4 Wave 3 Independent Variable s Predicting Wave 3 Drug Use...............................124 5-5 Wave 3 Independent Variables Pred icting Wave 3 Drug Use among Fathers.......125 5-6 Wave 3 Independent Variables Pr edicting Wave 3 Drug Use among Married Men.........................................................................................................................127 5-7 Wave 3 Independent Variable s Predicting Wave 4 Drug Use...............................128 5-8 Wave 3 Independent Variables Pred icting Wave 4 Drug Use among Fathers.......130 5-9 Wave 4 Independent Variable s Predicting Wave 4 Drug Use...............................131 5-10 Wave 4 Independent Variable s Predicting Wave 5 Drug Use...............................133 5-11 Wave 5 Independent Variable s Predicting Wave 5 Drug Use...............................134 5-12 Wave 5 Independent Variables Pr edicting Wave 5 Drug Use among Married Men.........................................................................................................................136 5-13 Desist 2: Wave 2 Independent Va riables Predicting Desistance Wave 3..............137 5-14 Desist 2: Wave 3 Independent Va riables Predicting Desistance Wave 3..............139 5-15 Desist 2: Wave 2 Independent Vari ables Predicting Desistance Waves 3-4.........140

PAGE 11

xi 5-16 Desist 2: Wave 3 Independent Vari ables Predicting Desistance Waves 3-4.........141 5-17 Desist 2: Wave 2 Independent Vari ables Predicting Desistance Waves 3-5.........142 5-18 Desist 3: Wave 3 Independent Va riables Predicting Desistance Wave 4..............143 5-19 Desist 3: Wave 4 Independent Va riables Predicting Desistance Wave 4..............144 5-20 Desist 3: Wave 3 Independent Vari ables Predicting Desistance Waves 4-5.........146 5-21 Desist 3: Wave 4 Independent Vari ables Predicting Desistance Waves 4-5.........147 6-1 Wave 1 Independent Variable s Predicting Wave 2 Arrests...................................149 6-2 Wave 2 Independent Variable s Predicting Wave 2 Arrests...................................151 6-3 Wave 2 Independent Variable s Predicting Wave 3 Arrests...................................152 6-4 Wave 3 Independent Variable s Predicting Wave 3 Arrests...................................153 6-5 Wave 3 Independent Variables Pred icting Wave 3 Arrests among Fathers..........155 6-6 Wave 4 Independent Variable s Predicting Wave 5 Arrests...................................156 6-7 Wave 5 Independent Variable s Predicting Wave 5 Arrests...................................157 6-8 Wave 5 Independent Variables Pred icting Wave 5 Arrests among Married Men.159 6-9 Desist 2: Wave 2 Independent Va riables Predicting Desistance Wave 3..............160 6-10 Desist 2: Wave 3 Independent Va riables Predicting Desistance Wave 3..............161 6-11 Desist 2: Wave 2 Independent Vari ables Predicting Desistance Waves 3-5.........163 6-12 Desist 2: Wave 3 Independent Vari ables Predicting Desistance Waves 3-5.........164 6-13 Desist 3: Wave 3 Independent Vari ables Predicting Desistance Waves 3-5.........165 6-14 Desist 3: Wave 4 Independent Vari ables Predicting Desistance Waves 3-5.........166 7-1 Wave 1 Independent Variables Predicting Wave 2 Incarceration..........................169 7-2 Wave 2 Independent Variables Predicting Wave 2 Incarceration..........................170 7-3 Wave 2 Independent Variables Predicting Wave 3 Incarceration..........................171 7-4 Wave 3 Independent Variables Predicting Wave 3 Incarceration..........................172 7-5 Wave 3 Independent Variables Pred icting Wave 3 Incarceration among Fathers.174

PAGE 12

xii 7-6 Wave 3 Independent Variables Predicting Wave 4 Incarceration..........................175 7-7 Wave 3 Independent Variables Pred icting Wave 4 Incarceration among Fathers.177 7-8 Wave 4 Independent Variables Predicting Wave 4 Incarceration..........................178 7-9 Wave 4 Independent Variables Predicting Wave 5 Incarceration..........................179 7-10 Wave 5 Independent Variables Predicting Wave 5 Incarceration..........................181 7-11 Desist 2: Wave 2 Independent Va riables Predicting Desistance Wave 3..............182 7-12 Desist 2: Wave 3 Independent Va riables Predicting Desistance Wave 3..............183 7-13 Desist 2: Wave 2 Independent Vari ables Predicting Desistance Waves 3-4.........185 7-14 Desist 2: Wave 3 Independent Vari ables Predicting Desistance Waves 3-4.........186 7-15 Desist 2: Wave 2 Independent Vari ables Predicting Desistance Waves 3-5.........187 7-16 Desist 2: Wave 3 Independent Vari ables Predicting Desistance Waves 3-5.........188 7-17 Desist 3: Wave 3 Independent Va riables Predicting Desistance Wave 4..............189 7-18 Desist 3: Wave 4 Independent Va riables Predicting Desistance Wave 4..............190 7-19 Desist 3: Wave 3 Independent Vari ables Predicting Desistance Waves 4-5.........192 7-20 Desist 3: Wave 4 Independent Vari ables Predicting Desistance Waves 4-5.........193

PAGE 13

xiii Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy FATHERHOOD AND CRIME: EXAMINI NG LIFE COURSE TRANSITIONS AMONG MEN IN HARLEM By Bradley G. Tripp May, 2007 Chair: Constance Shehan Major Department: Sociology This research examined the relationshi p between criminal careers and family participation. Using the Brunswick’s Harlem Longitudinal Study of Urban Black Youth, this project examined delinquency and incarceration among young black men. Data on participants were collected in 5 waves that spanned 26 year s. Building on the foundation of developmental life-course criminology, this research ex amined whether changes in these men’s social bonds influenced desistance from anti-social behavior. While previous research had focused on employment and marriage, this research examined the transition to fatherhood as a potential turning point in the lif e-course. Beyond paternity, residence and participation in childrearing was examined as variables of influence. While paternity was the main variable of interest, variables that had been found significant in previous research (employmen t, marriage, etc.) were also examined.

PAGE 14

1 CHAPTER 1 INTRODUCTION This research examined the relationshi ps between social bonds and deviant behavior. Framed within a developmental a nd life-course perspective, changes in the adult lives of respondents were predicted to influence criminal behavior. This perspective is founded in social control th eory, which asserts th at individuals whose bonds to society are either missing or weak are more likely to engage in deviant and criminal behaviors (Durkheim, 1897; Hirsch i, 1969; Janowitz, 1975; Kornhauser, 1978; Reiss, 1951). Therefore, the various bonds that subjects hold at different stages of the life-course were measured in order to reveal which bonds best predict participation in and cessation from criminal acts. This research examined criminal beha vior in two manners. First, the influence of social bonds on crim inal behaviors within the same wave and subsequent waves was examined. Next, the manner in which these social bonds influenced desistance from criminal behavior s through the life-course was examined. As a result, a myriad of relationships between social bonds and criminal behaviors were revealed. Sampson and Laub’s age-graded informal social control theory influenced the research goals and strategies, as well as the main hypotheses that this research sought to address. This theory purports that the accu mulation of informal social bonds affects desistance from criminal behavi ors, and that the strength of social bonds may be more influential than the acquisition of such bonds (Laub, Nagin, and Sampson, 1998; Laub and Sampson, 1993; Sampson and Laub, 1993). Pr evious research has examined social

PAGE 15

2 bonds such as marriage, employment, and education (Coombs, 1991; Currie, 1985; Doyle et al., 1999; Farrington et al., 1986; Gove et al., 1990; Grogger, 1998; Nock, 1998; Piquero, MacDonald, and Parker, 2001; Sa mpson and Laub, 1993; Shihadeh and Ousey, 1998; Shover, 1996; Waite, 1995; Warr, 1998). Ve ry little research has focused on the influence that a father-child bond may have upon desistance (Edin et al., 2004; Farrington and West, 1995; Knight, Osborn, and West, 1977). Therefore, this research built on previous life-course research by providing an examination of the ways that fatherhood influences anti-social behavior. The data for this research were provide d by the Henry A. Murray Research Center, at Radcliffe University. Dr . Ann Brunswick and colleagues began collecting data in 1968. Her study, the Harlem Longitudinal Study of Urban Black Youth, 1968-1994 , intended to focus on health issues within the New York community of Harlem. The sample consisted of 351 African American ma les who were between the ages of 12 and 17 from 1968 to 1970. All of the participants were residents of Central Harlem, New York during the initial wave of data collection. Logistic regression was used to examine th e influence of a variety of social bonds on the anti-social behaviors of these men. This method of analysis was used as all of the dependent variables sought to measure partic ipation or non-participation in various acts that were conceptualized to represent devian t or anti-social behavior. These anti-social behaviors were measured by drug use, arrest s, and incarceration. A series of crosssectional and lagged models that measured both participation and desistance from the previously mentioned meas urements were developed.

PAGE 16

3 CHAPTER 2 REVIEW OF LITERATURE Social Forces, Social Bonds The final 30 years of the twentieth ce ntury witnessed wide spread shifts in employment patterns, family structures, and correctional populations (Edin et al., 2004). As the social structures of family, crimin al justice, and employment were central measures in this research, it was important to review the changes that occurred along with some of the reasons behind these broad social upheavals. The key dependent variables of arrests, incarceration, and drug use certainly intertwined with the imprisonment binge that coincided with the end of the twentieth century. Therefore, th e incarceration boom that occurred in the past 30 years, and its effect on the poor urban black males in New York, is briefly reviewed. Research has revealed that two of the ma jor bonds that impede criminal behavior are familial involvement (Coombs, 1991; Cu rrie, 1985; Gove et al., 1990; Nock, 1998; Piquero, MacDonald, and Parker, 2001; Waite , 1995) and steady employment (Doyle et al., 1999; Farrington et al., 1986; Grogger, 1998; Sampson and Laub, 1993; Shihadeh and Ousey, 1998; Shover, 1996). The creation and maintenance of these two social bonds have been directly related to changes in so ciety. Additionally, the years during which the data for this study were collected (Data collection began in 1968 and ended in 1994) coincided with an era of change in the fam ily and employment spheres of Americans. The review of previous research also reve aled that strong connections between these aspects of social life exist. Initially social changes in the economy, corrections, and

PAGE 17

4 family life were examined. After each issue was examined on its own, connections between family life, employment, criminal behavior, and offici al responses were addressed. Employment and Education Education and employment are two of the f actors that have been seen to influence criminal behavior (Doyle et al., 1999; Farri ngton et al., 1986; Gr ogger, 1998; Sampson and Laub, 1993; Shihadeh and Ousey, 1998; Shover, 1996). Not surprisingly, the availability of these forms of social capital have waned along with the rise of mass incarceration that ha s influenced low wage earning males (Bound and Freeman, 1992; Freeman, 1996; Lichter, 1988). Social theori es have long predicte d that a lack of economic and educational opportunities may cause the frustration that can spur criminal behavior (Merton, 1968; Clow ard and Ohlin, 1960). Not su rprisingly, the 1980’s and 1990’s saw many young black men turn to crime in response to declining job opportunities (Freeman, 1996). Frequently, crime has been found to correspond with declining economic opportuni ties, joblessness, and neighborhood poverty (Bourgois, 1995; Crutchfield and Pitchford, 1997; Messner et al., 2001; LaFree and Drass, 1996; Venkatesh and Levitt, 1998). Poor academic pe rformances and low levels of attachment to school have also been common in both young and adult offenders, revealing educational opportunities as a factor in criminal involvement (Sampson and Laub, 1993; Hagan and McCarthy, 1997; Wolfgang et al., 1972). Over the last 30 years wages for low skill ed men dropped dramatically, as have the rates of employment for these same men (E din et al., 2004). A persistent gap in unemployment among black and white men bega n in the late 1960’s (Wilson et al., 1995; Bound and Freeman, 1992). Between 1972 and 1982 black men ages 18-29 experienced

PAGE 18

5 major increases in unemployment (Lichter, 1988). During the 1970’s and 1980’s the racial inequality in employment cont inued to increase (Lichter, 1988; Bound and Freeman, 1992). Declines in employment have been se verely high among young black males with low levels of education (Moss and Tilly, 1991) . Black men who failed to complete high school saw a steady decline in unemployment in the 1980’s and 1990’s, in spite of low levels of unemployment nationwide (Western an d Petit, 2000). Some of the factors that accounted for the disparate effects felt by these men were urba n deindustrialization, residential segregation, and social inequality (Massey and Denton, 1993; Oliver and Shapiro, 1997; Wilson, 1987). Historicall y, the collapse of legitimate economic opportunities led young black men with low levels of education into criminal enterprises such as drug trade (Duster, 1996). Innovative research examined labor statisti cs in a way that further revealed the ways in which the changing American econom y, in connection with changing criminal justice policies, negatively a ffected members of lower soci oeconomic status. Standard labor force data has failed to account for th e myriad of Americans who are behind bars (Western and Petit, 2000). The estimated employment rate for African Americans, young adults, and high school drop outs is reduced by an analysis that accounts for incarcerated Americans (Western and Petit, 2000). This revealed an interesting function for the American penal system. While unemployed pe ople in European nations were accounted for by the welfare system, many of the Amer ican unemployed are displaced into the criminal justice system. In this manner, the penal system intervened in the labor market by removing many of the unemployed from coun ts of those seeking labor, thus providing

PAGE 19

6 a falsely optimistic picture of the Americ an labor market in comparison with our European counterparts (Freeman, 1996; Western and Beckett, 1999). Marxist criminology referred to this function of incarce ration as “hidden jobl essness” (Jancovic, 1977). Indeed, American prisons and jails have masked masking as much as 20% of the labor force in some sections of employment (Western and Petit, 2000). Analysis of the American and European management of the labor supply revealed two important distinctions. First, the European system is redistributive, while the American system exacerbates inequality as inca rceration is used to limit the supply of labor (Western and Beckett, 1999). Comp arisons of national budgets furthered this notion. European nations spent 1/4 of th eir GDP on welfare while the United States allocated about 1/8 of their GDP for welfar e spending (Western and Beckett, 1999). Annual spending in the 1990’s on police, cour ts, and corrections was more than double the money allocated for all unemployment benefits along with employment related services (Statistical Abstract of the U.S., 1995; Western a nd Beckett, 1999). The other major difference between these two systems of labor force management is related to long term issues. In the American system, the relationship between incarceration and low levels of unemployment suggests that cont inued low unemployment rates can only be sustained by the constant expansion of the corrections population (Western and Beckett, 1999). Another long term concern for the Amer ican system is that the growing number of “ex-cons” will continue to have a hard time finding work and that this increasing proportion of society will eventually b ecome a detriment to America’s “low” unemployment rates (Weste rn and Beckett, 1999).

PAGE 20

7 Corrections The late 20th century saw an unprecedented increase in incarceration that has been referred to as an imprisonment binge as well as mass incarceration (Austin and Irwin, 2001; Garland, 2001; Wacquant, 2000). This er a of mass incarceration coincided with most of the time during which the data for this project were collecte d. As incarceration is one of the key dependent variables, a br ief review of the unprecedented spike in incarceration rates that occurred during th e study was presented below. Additionally, during this time period, the “Rockefeller Drug Laws” were passed by the New York State legislature. These laws resulted in a majo r increase in the per centage of offenders arrested and incarcerated for non-violent drug cr imes in the state of New York (Drucker, 2002). Here we see the intersection of the three main dependent variables: drug use, arrests, and incarceration. Between 1972 and 2000 the population of prisone rs in the United States increased six-fold (Harrison and Karberg, 2003). Corre spondingly, the incarcera tion rate within the United States increased every year fr om 1975 to 2001 (Beck, Karberg, and Harrison, 2002). Incarceration rates began to rise duri ng the early 1970’s and saw their most rapid growth during the 1980’s and early 1990’s (B JS, 1997). However, crime rates have fallen since 1980 according to the National Crime Survey (BJS, 1996). When comparisons are made between Western indus trialized nations, the incarceration rate within America dwarfs the rates in other na tions (Maurer, 1994; Beck and Gilliard, 1997) despite measurements that show that Ameri can crime rates are only slightly higher (van Dijk and Mayhew, 1992; Tonry, 1995; Western a nd Beckett, 1999). These trends reveal that the patterns of incarcera tion in the United States are only loosely tied to changing levels of criminal activity (Western and Beckett, 1999).

PAGE 21

8 A myriad of factors have influenced th e unprecedented growth in penal populations in America. Changes in American governance and corrections have us hered in an era of “mass incarceration” (Garland 2001 ). Governmental issues such as the delegitimization of social welfare programs, the creation of valu e politics, a variety of legislative acts, as well as the war on drugs have shaped a soci o-political environment that fosters the incarceration binge. Additionally, shifts in th e world of corrections such as changes in the theories behind incarcerati on, the crisis of reflexivity, and the economic power of the prison-industrial complex have complemented the political changes that brought forth a surge of incarceration. Each of these influe nces on the incarceration binge were briefly reviewed below. One of the major political shifts that precipitated the tremendous changes in criminal law making was the delegitimization of social welfare pr ograms (Baldwin, 1990; Burchell et al., 1991; Caplow and Simon, 1999) . The loss of social welfare as a legitimate government project left a huge vacuum that needed to be occupied by another form of governance. The failure of mode rn government to successfully regulate the social and material needs of its citizens le d to the growth and legitimization of crime control and punishment as important, if not central, government functions (Caplow and Simon, 1999; Garland, 1996). Correspondingl y, as crime control and punishment became a central feature for the American government, the manner in which we talked about crime and criminals changed into a disc ourse of “value politics.” The end of the twentieth century saw American politics move aw ay from conflicts of race and class, and into debates over values a nd identity (Beck, 1992; Caplow and Simon, 1999; Giddens, 1991). Treating crime as an issue of valu es allowed politicians to criticize the

PAGE 22

9 “underclass,” while avoiding the discourse of racial politics (Jackman, 1994; Hochschild, 1995, Tonry, 1995). While it was po litically incorrect to sp eak disparagingly against minorities, it was acceptable to berate and despise criminals. This new era of crime control as a centra l government function also saw a series of legislative acts that influenced the growth of correctional populati ons. Falling in line with opportunities to attack th e target population of criminals, legislators were quick to pass a variety of acts that helped bolster an already growing populat ion. Legislative acts such as determinant sentencing, bail reform, th e Prison Litigation Relief Act all helped to feed the American imprisonment binge (Cap low and Simon, 1999). Attempts at relieving prison overcrowding through releasing inma tes on probation, parole, and a range of intermediate sanctions only served to put offenders in a feedback loop where minor mistakes and transgressions send one back to jail (Petersilia, 1995). This showed in the fact that probationers made up a sizabl e proportion of prison admissions, ranging between 30 and 50% of the people th ey process (Parent et al., 1994). The legislative acts that can be seen as a part of “The War on Crime” can also be viewed in conjunction with the legislation that comprises “The War on Drugs.” The effects of these laws can be seen in the ch anging prisoner populations . During the span of three decades, the proportion of non-violent prisoners jumped from 1/2 to 2/3 of the prison population, and drug offenders’ repres entation among penal populations rose from 1/10 to 1/3 (Cole, 1999). The War on Crime and Drugs can be seen as a one aspect of a larger redefinition of state responsibilities (Beckett, 1997; Gans, 1995), whereby they are a part of a turn towards more punitive policies within crimin al justice systems (Blumstein and Beck, 1999).

PAGE 23

10 The war on drugs acted as a paramount force behind the incarceration binge (Blumstien and Beck, 1999; Donziger, 1996; Zimring and Hawkins, 1991). The increasingly punitive sanctions on drug offenders have seen the number of inmates skyrocket nation wide. Indeed, without the increasing punitiveness of drug laws, and the creation of a new population of offenders, the rise in imprisonment could not have happened (Blumstien and Beck, 1999; Donzig er, 1996; Zimring and Hawkins, 1991). Arguably, it is the increased severity a nd number of sentences handed out to drug offenders that stood as the number one cont ributor to the incarceration boom (Tonry, 1995). Drug offenders made up one-third of th e federal penal populati on, and one-fifth of the penal population among the st ate institutions. (Austin et al., 2000; Nadelman, 1992). From 1980 to 1997, the number of people lock ed up for drug offenses increased eight fold, increasing from 50,000 to 400,000 (Nadel man, 1998). Additionally, these policies have disproportionately affected minorities , and ensured that race, drugs, and class became heavily intertwined by the 1990’s (Sampson and Lauristen, 1997; Wacquant, 2000; 2001). The effects of the “War on Dr ugs” upon minorities in the state of New York were especially severe (Drucker, 2002) . In 1980, one-third of new drug offenders were white; this number dropped to 6% of the new drug offender population by 2000 (Drucker, 2002). Another major shift occurred within correcti ons. Concurrent with the ideologies of the war on crime and drugs, the field of correct ions saw a sharp turn away from a focus on rehabilitation towards in capacitation and punishment (Feely and Simon, 1992; Garland, 2001). One event, which was crucia l to changes in both penological practices and the legislation that contro ls correctional funding, was the infamous Martinson report.

PAGE 24

11 This report stated that correc tional treatments were not worki ng, and ushered in an era of penology where treatment was minimized as “nothing works” (Lipton et al., 1975; Martinson, 1974). While high recidivism rates ha ve been shown to consistently rise with the loss of treatment services (Petersilia , 1995; U.S. Comptroller General, General Accounting Office, 1976), a substantive return towards rehabilitati on has yet to occur despite research that has found substantive eff ects for a variety of tr eatments (Clear and Braga, 1995; Gaes et al., 1999; Langan, 1994; Pearson and Lipton, 1999; Petersilia and Turner, 1993; Wexler et al., 1999). The problem of recidivists revealed another force behi nd America’s incarceration surges; the crisis of reflexivity. Many th eorists have asserted that the changes of modernity have brought with them a conditi on called reflexivity (Beck, 1992; Giddens, 1991; Luhman, 1985). Reflexivity refers to the condition in which individuals and institutions function within a loop wherein their existence creates a new series of problems that they must solve, thereby furtheri ng their continued existence. Prisons were one of the first institutions that began to consciously func tion around the problems that it created (Caplow and Simon, 1999; Foucault, 1977; Rothman, 1972). The main problem created by prisons was a class of people called recidivists (Cap low and Simon, 1999). This failure to reform recidivists lead to claims that “nothing works” (Martinson, 1974), and lead into the era of “new penology” or managerialism (Bottoms, 1994; Feeley and Simon, 1992; Simon and Feeley, 1995). In this new era of criminal justice, corrections focused not on the reformation of offenders, but on maintaining what would seem to be a relatively stable, but frequently grow ing, population (Caplow and Simon, 1999).

PAGE 25

12 The incarceration binge created a modern ec onomic juggernaut that is referred to as the prison-industrial complex (Chambliss, 1999; Irwin and Austi n, 1997; Schlosser, 1998). This refers to the collection of ec onomic and political issues that encourage enormous spending on incarceration, despite the level of need for such projects (Schlosser, 1998). The collection of special interests and political paydays that have rallied behind this social fo rce are so powerful that there is little hope of stopping the forward progress of the in carceration juggernaut (Chamb liss, 1999; Schlosser, 1998). Overall, the sample of this study was faced with an increasingly punitive criminal justice system, where the effects were disp roportionately felt by minorities (Drucker, 2002, Maurer, 1999; Tonry, 1995). While various so cial forces influenced this change, it would seem that the influence of the “Wa r on Drugs” was felt the strongest by this sample, as a disproportionate amount of Afri can Americans were incarcerated for drug crimes in the state of New York during the study (Drucker, 2002). Family Family life also changed greatly during th e past 30 years. While the rates of marriage and divorce varied among all Americans, minorities and as low wage earning men saw a distinct decline in marriage rates (Edin et al., 2004). One influence on this statistic is the shortage of African American males in comparison to African American females (Staples and Johnson, 1993). Incarcerat ion, early death, and drug addiction were some of the causes of the shortage of “m arriageable” black men (Staples, 1988). These social forces greatly diminished the number of black males available for marriage. One result of the drop in marriage rates was an increase in father absent homes (King, 1993). Between 1960 and 1990, the percentage of children living away from their biological father more than doubled, from 17 to 36 percent (Popenoe, 1998). These

PAGE 26

13 numbers represented a change so drastic that some scholars stated that it was unprecedented in the recorded history of fa milies (Stone, 1989). Currently, about 1/5 (21%) of all American children live in a father absent home in which the mother is not a widow (US Census Bureau, 2002). More broadly, estimates ha ve stated that at some point in their lives about 1/2 of the children in the United States will live in a single parent home (Huber and Spitze, 1988). Many social ills have been attributed to father absence. In addition to a shift in family structure, the growth of single parent homes (largely single mothers) contributed to the growth in poverty and welfare depe ndence among children in the United States (Lerman, 1996). National statistics noted that 94 percent of AFDC (now TANF) caseloads were single mother families (Garfinkle and McLanahan, 1986). One study found that even when controlling for the lower income of single parent families, negative outcomes such as poor academic performance, drug use, teen pregnancies, and criminal behavior were more prevalent among children who grow up without a father in the home (McLanahan and Sandefur, 1994). However, Judith Stacey (1998) questioned the manner in which scholars used “fatherlessness” as a blanket term. Ind eed, she claimed that “the very category ‘fatherlessness’ itself would never pass soci al science muster’ (Stacey, 1998: 64). One problem with this term was that it subsumed a plethora of different family forms under one heading (McLanahan, 1998; Stacey, 1998). Certainly there are differences between families headed by never married mothers w ho conceived through intercourse, single women who used artificial reproductive t echnologies, divorced mothers, widowed mothers, and lesbian couples or singles. Th e blanket notion of fath erlessness ignores the

PAGE 27

14 diversity that accompanies a wide variety of homes that do include a father in its residence. Despite these problems, many non-married fathers seek to maintain a strong relationship with their children (Lerman, 1993; Lerman and Sorenson, 2000; McLanahan et al., 2001; Mott, 1990). This is also true among unwed incarcerated fathers as well (Edin et al., 2004; Nurse, 2004; Tripp, 2001). In 1988, nearly 1/2 of young, unwed fathers in one research sample noted that th ey visited their youngest child at least once a week (Lerman, 1993). Research examining visi tation among single fathers revealed that their reports of frequency ar e somewhat reliable as their reports of frequency are only somewhat higher than the numbers reporte d by the children’s mothers (Mott, 1990). However, among some non-residential fathers, the relationship between father and child is mediated by the child’s mother (Marsig lio, 1995; 1998; Nurse, 2004). Women may act in the role of a gatekeeper, whereby they enga ge in a range of activ ities that shapes how men relate to their children (Marsigl io, 1995; 1998). Women may encourage or discourage men’s actions as fathers, and may foster feelings of success or inadequacy when men act in their role as father. A disturbing trend is that paternal involv ement tends to decrease with time. Nonmarried fathers’ involvement with their non-resi dential children tends to decrease as the children age (Lerman, 1993; Seltzer, 1991). Prop inquity, or resident ial proximity, also holds an inverse relation with paternal visitation (Lerman, 1993; Seltzer, 1991). Some research asserts that patern al involvement among non-reside ntial, low-income, minority fathers may be an activity that fluctuates, ra ther than one which can be measured along a linear decline that leads to a termination of paternal i nvolvement (Eggebeen, 2002; Mott,

PAGE 28

15 1990; Roy, 2003). However, no research has fo und a significant connection between the well-being of children and the contact that they hold with a non-resident father (Furstenberg et al., 1987; Jordan, 1996; King, 1993). Race, Family, Crime, and Money The Influence of Race There are a disproportionate number of minorities within the criminal justice system (Mauer, 1999). Some research has asse rted that race is not a central feature in crime, as the offending trends and particip ation rates of whites and non-whites differ (Blumstein et al., 1986; Elliott, 1994; Piquer o, MacDonald, and Parker, 2002; Visher and Roth, 1986). However, a myriad of correla tes beyond participation have been found to be significantly related to the racial disparit ies in the criminal justice system. Race has been found to influence criminal particip ation indirectly th rough the structural disadvantages suffered by minorities, especi ally amongst African Americans (Sampson and Laub, 1993). Many African American male s will be arrested pr ior to adulthood, as 15 percent will be arrested for an index cr ime, and between 25 and 45 percent will be arrested for a non-traffic offens e (Blumstein et al., 1986). African Americans were less likely to be married than whites, with approximately 80.7% of whites and 47.1% of blacks married in 2000 (Census Bureau, 2000; Piquero, MacDonald, and Parker, 2002; Wilson, 1987). One factor was the loss of “marriageable men” in the African American community (Wilson, 1987: 145). Additionally, African American families had a high rate of family disruption, which was associated with higher rates of violent crime (Sampson and Laurit sen, 1994). Connections have been made when examining the relationship between race and partnerships. Marriage was found to have a negative association with non-violen t arrests with both whites and non-whites,

PAGE 29

16 however this was not found in regards to viol ent arrests, where married nonwhites than had a higher frequency of violent arrests than unmarried whites (Piquero, MacDonald, and Parker, 2002). Among non-whites commonlaw marriages have been seen to be positively associated with violent, non-viol ent, and total arrests (Piquero, MacDonald, and Parker, 2002). Families in the Lives of Men Families have functioned as a powerful form of informal social control that deter men from criminal behavior (Currie, 1985). The “fabric of inte rdependencies” woven within families have influenced the behavior s of the men who play the roles of husband, father, and son within this social group (C urrie, 1985). Marriage has yielded positive effects in the lives of men. Not only have men received a great d eal of individual and social benefits from being married, but they also have benefited a great deal more than women (Nock, 1998). Marriage has been widely cited as a factor that decreases the likelihood of criminal behavior. Criminal activity has been less likely when men get married and maintain a stable relationshi p (Laub, Nagin, and Sampson, 1998), or have been living with their wife (Horney, Osgood, a nd Marshall, 1995). Overall, married men were more healthy (both physically and mentally ), lived longer, and we re less likely to be involved in criminal activities when compar ed to their unmarried counterparts (Coombs, 1991; Gove et al., 1990; Nock, 1998; Piquer o, MacDonald, and Parker, 2001; Waite, 1995). While marriage has been widely researche d, paternity has been grossly overlooked as a turning point within life course criminology (Edin et al., 2004). Beyond Sampson and Laub’s (1990) mention of parenthood as a “dominant institution,” only a couple of scholars have examined the influence of patern ity on the criminal careers of men (Edin et

PAGE 30

17 al., 2004; Farrington and West , 1995; Knight, Osborn, and We st, 1977). This lack of attention to a major event in the lives of many men was a major motivation for this research. While fatherhood has been examined in family research as both a turning point and a trajectory (Marsiglio, 1995; 2002), its significan ce in regards to an ti-social behavior has yet to be fully examined. The transition to parenthood is one that be gins even prior to conception in the form of a procreative consciousness (Marsigli o, 1998). Procreative consciousness refers to men’s ideas, conceptions and self awareness in regards to their role as reproductive beings (Marsiglio, 1998). Beyond concep tion, pregnancy and childbirth are often overwhelming for fathers (Cowan et al ., 1985; Klein, 1985; Went e and Crockenberg, 1976). Some models have constructed the transition to fatherhood as a sequence of developmental tasks, within which mastery of each task indicates whether the transition to fatherhood is either a crisis or a time of growth (Barnhi ll et al., 1979). A longitudinal study examining men’s identities during the transition to fatherhood found that core characteristics shifted from aggressivene ss and independence to a focus on caring, provision, and empathy (Cowan and Cowan, 199 2). Other longitudinal studies have supported this idea and noted that the transition to fatherhood is accompanied by increasing maturity (Heath, 1978; Heath and Heath, 1991). Similar to marriage, paternity has influen ced the choices that men have made in their everyday lives. When fathers were highl y involved with their children they became more attentive to finding and holding employment, while also decreasing their involvement in risky behavior (Lerma n and Sorenson, 2000). Edin, Nelson, and Paranal’s (2004) qualitative study found that fatherhood altered how men perceived the

PAGE 31

18 risks and rewards of criminal behavior. Th e bonds they held with their children became highly salient, and behaviors and consequences (such as incarcerati on) that would have negatively impacted these bonds were to be avoi ded at all costs. Some of the underlying reasons behind these assertions were a fear of the loss of their ability to provide financially, as well as their desire that th e children viewed them in a positive light. Research noted that the age of the particip ant was influential ove r the amount to which the birth of a child influenced their desist ance from crime (Edin et al., 2004). However, other research found higher levels of i nvolvement among younger rather than older fathers (Danziger and Radin, 1990). Adolescent Fathers While the life course transition to father hood can be difficult for all men (Cowan et al., 1985; Klein, 1985; Wente and Crockenbe rg, 1976), it is even more trying for adolescent males (Hagan and Wheaton, 1993; Howel and Frese, 1982; Jordan, 1996). In a society in which a linear life course is e xpected (Elder, 1982), a break in the standard order of obtaining secondary education, fo llowed by marriage, followed by paternity, places a series of stresses and probl ems upon the young father (Hagan and Wheaton, 1993; Howel and Frese, 1982; Jordan, 1996). Financial and emotional immaturity, along with a lack of resources, provides many problems for young fathers (Marsiglio, 1987; Hardy and Duggan, 1988; Lerman, 1993; Elster and Lamb, 1987; Elster et al., 1987; Stoutheimer-Loeber and Wei, 1998). While some of these men may desire to take on the role of father, scant resources make th is task difficult (Pleck et al., 1995). According to national surveys, between 27% of American adolescent males were fathers; with higher rates found among inner city and minority youth (National Center for Health, 1994; Marsiglio, 1987; Sonenstein et al., 1989; Stoutheimer-Loeber and Wei,

PAGE 32

19 1998). Interestingly, adolescent fathers were more likely to be employed and have a disposable income than their cohorts; however because of their situa tion as fathers their cohorts quickly surpassed them due to edu cational attainment and job training (Belsky and Miller, 1986). While many adolescent fathers held excitement about the birth of their child, many did not continue a high level of involvement as time passes (Danziger and Radin, 1990). There were strong correlati ons between adolescent fatherhood and a spectrum of different measurements of soci oeconomic disadvantage s (education, income, etc.) (Geronimus, 1991). Teenage fathering ha s been found as both a cause and result of social conditions, but having a child did hold deleterious eff ects on the future chances of these young men (Geronimus, 1991). Some of the negative effects of a dolescent fatherhood upon these young men were an increased likelihood of problematic beha viors such as poor academic performance, behavioral problems at school, drug and alcohol abuse, sexual promiscuity, and delinquency (Elster et al., 1987; Christm on and Luckey, 1994; Dearden et al., 1995; Resnick et al., 1993; Ketterli nus et al., 1992; Springarn and DuRant, 1996; StoutheimerLoeber and Wei, 1998). Two findings from Stoutheimer-Loeber and Wei (1998) were striking in that young fathers we re twice as likely as non-fath ers to be classified as delinquents, and that deviant be havior did not decrease in eith er the year of or the year after the child’s birth. Prisons, Providing, and Problems Both criminology and family studies have examined the influence of employment and income on many different variables. One of the ways in which these social forces intersected was that men who were poor a nd/or low wage earners became first time fathers during the same years of the life course in which they were most likely to engage

PAGE 33

20 in criminal activities (Edin et al., 2004). Among fathers the ability to provide, or to perform the instrumental family role, is primary (Bozett, 1985; Gadsen and Smith, 1994; Tripp, 2001). Research finding a strong relati onship between a lack of employment and reduced paternal involvement, was theref ore unsurprising (Achatz and MacAullum, 1994; Anderson, 1993; Danziger and Radin, 1990; Furstenberg, 1995). While some research asserted a causal relationship betw een father involvement and men’s ability to find work (Levine and Pitt, 1995), others noted that finding a causal relationship between income and paternal participation among lo w-income men was difficult as the causation runs in both directions (Lerma n and Sorenson, 2000). Marriage also held similar effects among men, in that marriage has been seen to correlate with an incr ease in salary above that of similar, yet unmarried, workers (Korenman and Neumark, 1991). Additionally, greater involvement with one’s children has le ad men to have a better level of marital adjustment than uninvolved fathers (Pleck, 1985). Another relationship that ha s been established by research is that of crime and employment. Criminal involvement has tende d to increase when wages were low (Doyle et al., 1999; Grogger, 1998), job stability decreased (Sampson and Laub, 1993), entry level jobs were unavailable (Shihadeh a nd Ousey, 1998), and unemployment was high (Farrington et al., 1986). Offe nders tended to desist when they found jobs that were stable (Shover, 1996). The relationship betw een incarceration and employment has been seen to be causal in both di rections as both criminal convi ctions and serving time have been found to negatively affect employment as well as earnings (Freeman, 1991; Lott, 1990; Naging and Waldfogel, 1993; Samps on and Laub, 1993; Waldfogel, 1994; Western and Beckett, 1999; Witte and Reid, 1980). This can be seen in that former prisoners

PAGE 34

21 earned lower wages and experienced higher le vels of unemployment than similar men (Western et al., 2001). Certainly the relationships between crimin al behaviors, family participation, and economic success or failure are myriad and complex. The ways in which these three areas of social life intersect formed the core of this research, as relationships between family, employment, and criminal behavior we re fully examined. As noted earlier, the relationships between these social spheres have been examined in both family and criminological research. The following chap ter looked further into the criminological background of this research. Indeed, the th eory that guided this research holds both family participation and employment as ke y variables in the st udy of delinquent and criminal behavior.

PAGE 35

22 CHAPTER 3 THEORY The theoretical foundation for this resear ch was developmental and life-course criminology. The developmental and life-cour se criminology (DLC) perspective focuses on the development of and changes in anti-s ocial behavior throughout the life-course. DLC looks at the influences of various risk f actors and life-course events as people age. This theory differs from othe r paradigms in that examining within-individual changes is a core directive. This theory grew out of the criminal career paradigm (Blumstein et al., 1986; Farrington, 2003). It is also rooted in the theoretical so il of social control theories and the life course perspective. Each of thes e theories, along with some of their essential underlying principles, are discus sed within this chapter. Developmental life-course criminology is then reviewed in greater detail. The history of this theory along with some of its core concepts and issu es are also presented. While the myriad of distinct models that can be categorized as developmental and life-course criminology theories are review ed, the specific theory under which this research has been developed is examined at further length. Sampson and Laub’s agegraded informal social control theory (1993) directed the constructi on of this research. Finally, some of the main findings from this line of research are discussed. Social Theories on Crime Social Control As noted above, social control theory is th e organizing principle within life course criminology (Laub and Sampson, 1993; Samps on and Laub, 1993). A variety of social

PAGE 36

23 control and bonding theories were derived from the Durkheimian notion that people are more likely to conform to the rules than to break them (Agnew, 1985; Cloward and Ohlin, 1960; Cohen, 1955; Nye, 1958; Reiss, 19 51; Sykes and Matza, 1964). Hirschi’s social bonding theory has remained one of the most frequently discussed modern criminology theories, and the most utilized vers ion of social control theory (Akers, 1997; Stitt and Giacopassi, 1992). The basic tenet of social control theory is that crime and deviance result when an individual holds w eak or broken bonds with society (Durkheim, 1897; Hirschi, 1969; Janowitz, 1975; Kornhauser, 1978; Reiss, 1951). A social bond is composed of four elem ents (Hirschi, 1969). The first element, attachment, describes the degree to which an individual holds close affectionate relationships with others. Commitment is the ne xt element, this is the level of investment that an individual holds in conventional social structures, such as an occupation or an education. Involvement measures how much an individual is occupi ed with conventional activities. Finally, belief examines the level to which an individual endorses the moral rules and ideas of a given society. As people age into adulthood they develop ti es that “create inte rdependent systems of obligation and restraint that impose signifi cant costs” upon negative behaviors, such as criminal activities (Laub and Sampson, 1993: 306). Bonds do not simply materialize, as they take time to develop (Laub, Nagin, a nd Sampson, 1998). Over time, the investment that individuals make into so cial bonds grow, as does the in centive for avoiding specific behaviors (Laub, Nagin, and Sampson, 1998; Laub and Sampson, 1993; Nagin and Paternoster, 1994). The emergence of social bonds should therefore be viewed as an investment process, whereby “enduring attachme nts” to a specific role, or other people,

PAGE 37

24 influences the manner in which one act s (Laub, Nagin, and Sampson, 1998; Laub and Sampson, 1993; Nagin and Paternoster, 1994). Social bonds between individuals could be viewed as a primary relationship, wh erein interdependence and a sense of collectiveness results in a fusion of individualities (C ooley, 1909). Not surprisingly, these social bonds provide individuals with a sense of “connectedness” to the social networks within which they are embedded (Laub and Sampson, 1993). Not surprisingly, research has found the ma nner in which people are connected, or the degree to which they are socially bonded, influences delinquent and criminal behaviors. Those who hold weak connec tions with systems of interdependency are unbound to commit delinquent acts (Braithwait e, 1989). Research has found that the strength of bonds are often more influential th an the mere acquisition of bonds as in seen in the differential results of satisfying marriages and employment as opposed to the acquisition of a spouse or a job (Laub, Na gin, and Sampson, 1998; Laub and Sampson, 1993; Samspson and Laub, 1993; Sampson and Laub, 1990). Among adults, changes in criminality can be structured by life events and transitions (Rutter et al., 1990; Sampson and Laub, 1990). Additionally, the loss of empl oyment, a marital relationship, or contact with ones’ offspring has been found to be signi ficantly related to in creases in anti-social behavior (Farrington and West, 1995; Samp son and Laub, 1993; Western and Beckett, 1999). Social Capital One manner in which social bonds operate is through the development of social capital. Social capital is cr eated within the inte rpersonal relationships and connections that individuals make with institutional st ructures (Laub and Sampson, 1993). While less tangible than physical capital, social capital makes social exchanges and interactions

PAGE 38

25 possible that could not occur if it was not present (Laub and Sampson, 1993). Social investment, or capital, influe nces the strength and salien ce of bonds, thereby influencing the potency of informal social control at the micro, or individual level (Coleman, 1988; Sampson and Laub, 1990). Social capital is an aspect of social bonding that is e ssential in the creation and maintenance of ties to so cial institutions (Colema n, 1988; 1990; Laub and Sampson, 1993). Social groups that di splay interdependency and re ciprocity among members hold social ties imbued with high levels of so cial capital (Laub and Sampson, 1993). Weak social bonds are characterized by a lack of social capit al (Coleman, 1990; Laub and Sampson, 1993; Nagin and Paternoster, 1992; Sampson and Laub, 1993). As noted earlier, the strength of social bonds is a strong predictor of changes in criminal behavior (Laub, Nagin, and Sampson, 1998; Laub and Sampson, 1993; Samspson and Laub, 1993; Sampson and Laub, 1990). Social capital has be en integrated into life-course theories because criminological theories needed to address changes within important social institutions such as the family and employm ent. According to DLC theorists Laub and Sampson “social capital and turning points are important concepts in understanding processes of change in the adult life course” (1993: 302). Life Course Theories The life-course perspective has been increa singly applied to cr iminological theory (Piquero, MacDonald, and Parker, 2002; Pique ro and Mazzerole, 2001). This specific perspective has been particular ly helpful in that it addresses changes in circumstances that have been shown to influence anti-social behavior. The life-c ourse perspective has examined not only the diverse life circumstan ces within which offenders act, but it has also examined how these circumstances can va ry both in quality and strength across time

PAGE 39

26 (Horney, Osgood, and Marshall, 1995; La ub, Nagin, and Sampson, 1998; Piquero, MacDonald, and Parker, 2002; Piquero a nd Mazzerole, 2001; Sampson and Laub, 1993; Thornberry, 1997). Research prio r to the life-course perspect ive consisted of only a snap shot of social relations (Giele and Elder, 1998). Prior to its use within criminology, lifecourse theories focused on three broad issues: 1) Social meanings of age thr ough the life-course; 2) Intergenerational transmission of social patterns; 3) The e ffects of macro level events (Sampson and Laub, 1992). The following sections reviewed some of the main concepts that have been utilized in life-course resear ch. This perspective’s appl ication within criminology is discussed at length later in this chapter. There are three key concepts that are essential to a life-course study: trajectories, transitions , and turning points. All three of these concepts are discussed with a brief description of the ways they were utilized within this research. Trajectories in the Life Course A trajectory is “a pathway or line of deve lopment over the life span such as work life, marriage, parenthood, se lf-esteem, and criminal behavior” (Sampson and Laub, 1992: 65). Trajectories seek to map out the long-term patterns and directions in which people’s lives flow. The author examined a variety of trajector ies within this research. The two main trajectories under consideration were the crim inal and family trajectories. Family trajectories such as marriage and parenthood were measured in or der to discern how family involvement affected th e trajectory of criminal activit y. In addition to family and criminal trajectories, other trajectories su ch as employment and educational attainment

PAGE 40

27 were included as they have been proven influential on criminality (Farrington et al., 1986; Rutter et al., 1990). Thornberry (1997) had noted that trajectorie s hold three dimensions. First, the entrance dimension measures whethe r or not an individual partic ipates in the trajectory of interest. For example, not al l of the subjects will be a hus band or a father. The next dimension of a trajectory is timing, or mo re specifically, social timing (Elder, 1994). Social timing examines what happens, when it happened, and for how long. The social element of timing looks at age-graded expect ations that accompany certain roles and life events. The experience of life-course events , such as fatherhood, va ries in accordance with its social timing (Nock, 1998). When examining the connection between fatherhood and criminality, previous research has showed that having a child during the expected age range can result in decreases in criminal pr opensities, while fathering a child at a young age can lead to further cr iminal involvement (Farringt on and West, 1997; Nock, 1998; Thornberry, 1997). The final dimension of trajectories is that of success. Simple measures of success would be maintaining a marriage, or a j ob through out the du ration of the study. However, for this research, success was ex amined through analyzing the strengths of social bonds held. Success was measured in this research thro ugh the inclusion of variables such as marital satisfaction, j ob satisfaction, father success (two different measures) and residential fatherhood (father involvement as success). Transitions The second key concept in the life-course perspective is a transition. These are changes, or life events within a trajectory (Elder, 1985). These events tend to occur relatively suddenly, and hold a great deal of influence over the individual’s life

PAGE 41

28 trajectories (Elder, 1985; Rutter and Rutter, 1993). This research examined how a variety of transitions influence the participants’ crim inal participation. Ev ents such as being arrested, fathering a child, ga ining employment, and completion of a high school degree were measured and analyzed. Turning Points Life course perspective has also been us ed to examine turning points. Many times turning points are drastic adaptations take n in response to a major transition along a particular trajectory (Elder, 1985). Turni ng points cause individuals to change, or redirect, the path of their current traj ectory (Elder, 1986, 1994; Rutter et al., 1990; Sampson and Laub, 1990). However, given the in terpretive nature of what a major event is, and whether or not it has ch anged one’s life-course, it is difficult to determine if an event really is a turning point . Given the complex and subjec tive nature of defining what is or is not a turning point, especially with a quantitative source of data, this research focused on transitions rather than on turning points. Cohorts and Period Effects Certainly one of the key elements of the life -course perspective is that it is sensitive to the historical context within which hu mans develop and behavior unfolds. This perspective differentiates be tween the influence of cohor t and period effects (Elder, 1994). The cohort effect acknowledges that persons from different generations may experience the same events in different ways. However, historical events can also yield relatively similar change across birth cohorts as well; this is the period effect (Elder, 2000). Modern historical changes which have influenced American children are the growth of the mass media, the growth of the public school system, and fluctuations within economic cycles, especially the Great Depression (Eld er et al., 1993).

PAGE 42

29 Life Course Criminology Life-course criminology emerged from a milieu of various issues and theories. Some of the concepts acquired from life-course theory have been discussed. The notions of trajectories and transitions will be di scussed at further length as criminology’s adaptation of life-course theories is examine d. However, the theore tical journey of lifecourse theory into criminology has been characterized by a coupl e of important and influential stops. The concept of the criminal career could be envisi oned as the beginning of this sojourn. The battle over this concep t, along with the accompanying clashes over the age-crime curve and longitudinal data, have shaped the development of various theories and have held widespread influen ce on research. The concept of a criminal career is reviewed along with a brief recount ing of the scholastic skirmish about the concept and some related theoretical, conceptual, and methodological issues. Following this review, developmental and lif e-course theories are discussed. The main tenets of this branch of criminologica l theory are examined, as are some of the concepts created and the issues investigated by these theories. The various paradigms that have been considered to be DLC will then be redressed. Next, the DLC theory that shaped this research project will be discusse d. Finally, this chapter closes with a review of some of the important findings of DLC res earch, especially those which are relevant to the current research. Criminal Careers in the Life-Course The criminal career trajectory looks at how individuals have patterned their criminal behaviors over a period of time (Blu mstein et al., 1986). As noted above, the three dimensions of a trajectory are of gr eat importance when measuring the criminal trajectory. Certainly kn owledge regarding the age at which one enters into the criminal

PAGE 43

30 trajectory, one’s success at crime (measured via offenses versus arrest/encounters with the police), as well as examinations of th e timing of offenses in relation to other trajectories paints a vivi d portrayal of the individual’s criminal career. Nonetheless, there are four additional di mensions integral to understanding a criminal career (Blumstein et al., 1986). The first dimension is participation. While at first glance this would seem to measure th e same notion as the entrance dimension of trajectory, there is a subtle difference. Entr ance measures whether or not an individual has ever committed a crime, whereas participat ion looks at the continuance of criminal behavior. As subjects are interviewed multiple times, this variable assessed if an individual has committed any crimes during the preceding time span. Within this research, participation was the main dependent variable. Participation was an appropriate measure of anti-social behavior within this project as the sample was a general population (Piquero, Farrington, and Bl umstein, 2003). A deviant sample would be more appropriate for the following meas ure of criminal behavior. The next dimension, which is one of th e key issues in modern criminological debates, measures the frequency of the indivi dual’s participation in criminal activities (Blumstein et al, 1986). Frequency, or lam bda, looks at how many crimes an individual has committed over a specific period of time. Once again, frequency was not be a key consideration in this research as the sample consists of a general population. Frequency measures are best utilized in a sample of offenders, or a “deviant” sample (Piquero, Farrington, and Blumstein, 2003). The third di mension examines the seriousness of the crimes committed. This measure was not be used as data was not available on the crimes for which participants were arrested or incarc erated. However, future use of this data

PAGE 44

31 may be able to use the dimensions of freque ncy and seriousness if the main consideration is drug use. The Harlem Longitudinal data set does describe both the frequency and the types of drugs that subjects have used. Nonetheless, this research focused on participation, examining if a broad categor ization of anti-social behavior (including arrests, incarceration, and use of any illicit drugs) was infl uenced by disparate social bonds. The fourth dimension is that of durati on, a measure of the length of the criminal career from beginning till desistance (Blumste in et al., 1988). This dimension was not expressly examined but was observed as the cr iminal behavior at different time periods was noted. The Age Crime Curve and the Criminal Career Debate Data collected in the United States and ot her industrialized nations have indicated that both violent and property crime rates rise rapidly as people progress through their teenage years, peaking at 16 and 18 respect ively, followed by a decline through the rest of the life-course (Hirschi and Gottfreds on, 1983; Farrington, 1986; Sampson and Laub, 1992). Indeed, this fact has been upheld in an examination of criminal behaviors at different ages (Stattin, Magnusson, and Reiche l, 1989), that has been cited as the most comprehensive analysis on the subject thus far (Piquero, Farr ington, and Blumstein, 2003). While the peaks in offending vary betw een official data (ages 14-18) and selfreports (ages 13-16) (Elliott et al., 1983), all three major forms of data used in criminal justice research (official data, self-repor ts, and victim reports) have found an overrepresentation of youth as criminal offenders (Hindelang, 1981; Rowe and Tittle, 1977; Sampson and Laub, 1992). While the prevalence of young people as offenders is not in question, it has been the manner in which peopl e desist from crimin al activities through the life-course that has become a major dispute in criminology.

PAGE 45

32 Despite agreeing that criminal activity d ecreases with age, sc holars have battled over the meaning of different variations with in groups, between years, by crime type, and in different places (Blumstein et al., 1988a, 1988b; Farrington, 1986; Gottfredson and Hirschi, 1988; Hirschi and Gottfre dson, 1983, 1986, 1987). Blumstein, Cohen, and Farrington asserted that these differences we re important issues th at would lead to a better understanding of how individuals move th rough criminal careers (Blumstein et al., 1988a, 1988b; Farrington, 1986). On the othe r hand, Hirschi and Gottfredson claimed that these distinctions were trivial (G ottfredson and Hirschi, 1988; Hirschi and Gottfredson, 1983, 1986, 1987). While the con cerns of Gottfredson and Hirschi are noted, this research was shaped by the theo retical tenets of Blumstein, Farrington, and Cohen (along with other scholar s whose work is discussed at length later in this chapter). Gottfreson and Hirschi Gottfredson and Hirschi (1988) rejected th e construction of criminal careers on a number of fronts. First, they noted that th e concept of criminal careers was lacking in theoretical support. Gottfredson and Hirschi a ssert that the concept of a criminal career was not only theoretically bereft; it was also politically driven (Go ttfredson and Hirschi, 1986, 1987, 1988, 1990; Hirschi, 1969; Hirschi and Gottfredson, 1983). They claimed that a career must show progress. They then sited numerous works in which the seriousness of offenses did not increase thr ough out the criminal ca reer (Blumstein and Cohen, 1979; Glueck and Glueck, 1940, 1968; Wolfgang et. al., 1972). Second, they asserted that frequency can be measured through a cross-sectional study (Gottfredson and Hirschi, 1986, 1988, 1990; Hirschi, 1969). Third, according to these scholars, differences within the age-crime curv e are trivial (Goring, 1913; Guttman, 1977; Hirschi and Gottfredson, 1983, 1986). Finally, cross-secti onal studies are quicker and more cost

PAGE 46

33 effective than longitudinal study, therefore such research designs are wasteful and unnecessary (Gottfredson and Hirschi, 1986, 1987, 1988, 1990). Gottfredson and Hirschi (1990) advanced a version of control theory where the central factor was that of self-control. Th is theory asserts that self-control is an individual measure that is established during childhood and remains stable through time; therefore the cause of criminal behavior, or anti-social behavior , for all individuals remains constant throughout the life-course (Gottfredson and Hirschi, 1990). The theoretical grounding for this paradigm can be found in Gottfredson and Hirschi’s evaluation of the age-crime curve wherein they state that it is invari ant; showing the same variance in prevalence and frequency by age fo r all offenders (Hirsc hi and Gottfredson, 1983). Blumstein, Cohen, and Farrington The opposite side of this discussion asse rted three main arguments. First, Gottfredson and Hirschi both misunderstood th e criminal career concept and attached political value where there was none (Blu mstein and Cohen, 1979; Blumstein et al., 1988a, 1988b; Blumstein et al., 1982). Second, changes within subjects’ participation rates and frequency of criminal activities are an important facet of crime to measure. Additionally, these changes can only be appropriately measur ed at different points in time, adding to the logical approval of long itudinal research (Blu mstein et al., 1988a, 1988b; Farrington et al., 1986). Finally, the differences among offenders on the agecrime curve are important issues that reveal substantial information about the patterns of criminal careers (Blumstein et al., 1988a , 1988b; Farrington, 1986; Farrington, et. al., 1986). For example, the decline in the aggreg ate arrest rate as i ndividuals age does not

PAGE 47

34 require that the frequency of offending follows the same pattern (Blumstein et al., 1988a, 1988b). Developmental and Life Course Criminology Developmental and life-course criminology is an extension of the criminal career paradigm created in light of claims that th e model was atheoretical (Blumstein et al., 1986; Farrington, 2003). DLC addresses three i ssues through the incorporation of three disparate paradigms (Farrington, 2003). First, this theory d eals with the risk factors for offending at different ages through the in corporation of risk factor prevention (Farrington, 2000; Hawkins and Catalano, 1992; Loeber and Farrington, 1998). Second, DLC examines the development of offending a nd anti-social behavior with the use of developmental criminology (LeBlanc and Loeb er, 1998; Loeber and LeBlanc, 1990). Finally, this theory uses life -course criminology to address th e effects of life events and life transitions on offending (Sampson and Laub, 1993). Generally, the subjects of DLC research have been lower class, urban males residing in industrialized West ern societies (Farrington, 2003) . Broadly this research examines the development of offending over time , and tries to identify the causal factors that occur prior to or alo ng with the deviant behavior (LeBlanc and Loeber, 1998; Piquero, Farrington, and Blumst ein, 2003). Broken down further, this theory looks at anti-social behavior over the life-course and the factors that influence the movement into and away from deviant actions. One aspect of this theory that se parates it from other paradigms is the manner in which causal fact ors and the timing of various life-course events are brought forth as essential elements in shap ing a deeper understanding of criminal behavior. The life-course perspe ctive influences DLC to acknowledge that

PAGE 48

35 changes in behavior are influenced by the or derly changes that occur as the participant ages (Patterson, 1993; Thornberry, 1997). DLC Issues and Measurements Life-course theories address a number of issu es such as trajectories, transitions, and turning points. The applicati on of the life-course perspectiv e to criminology has resulted in a myriad of new concepts that seek to provide a more detailed explanation of the criminal career through the life-course. The traj ectory of criminal careers is examined in greater depth through the concepts of activat ion, aggravation, and desistance (LeBlanc and Frechette, 1989; LeBlanc and Loeber, 1998). Activation addresses the manner in which criminal behavior continues once initia ted, how it changes, its frequency, and its diversity (LeBlanc and Loeber, 1998). The th ree sub-processes of acceleration (increases or decreases in frequency), stab ilization (continuity in crimin al behavior over a period of time), and diversification (the diversity in criminal actions) develop a thorough model of the activation of criminal beha vior. The concept of aggravat ion questions whether or not subjects engage in more or less serious crim inal behaviors over ti me. Desistance is a central issue in criminology as it seeks to understand how and when people refrain from criminal activity. This re search sought to discover how young men in Harlem altered their criminal behavior as a result of cha nges in the life circumstances. Therefore, desistance was a key concept in this research. Desistance LeBlanc and Loeber (1998) describe desist ance as a “deceleration” in the frequency of criminal activities, a decrease in “speci alization,” and as a “de-escalation” of the seriousness of criminal activities. These thr ee paths of desistance, while addressing the myriad of ways in which people may redu ce their deviant behaviors, complicate the

PAGE 49

36 research agenda of those who will study desi stance. Previously this topic had been neglected (Loeber and LeBlanc, 1990), due to a variety of methodolog ical and theoretical problems (Laub and Sampson, 2001). However, many social scientists have sought to develop a deeper understanding of this issue (Farrington and Hawkins, 1991; Laub, Nagin, and Sampson, 1998; Loeber et al., 1991 ; Warr, 1998). Nonetheless, many issues regarding this subject remain unresolved. The definitional issues of desistance ar e further complicated by the discussion over whether it is an event or a process (Fagan, 1989; Laub and Sampson, 2001; Maruna, 2001). Desistance is difficult to measure, es pecially quantitativel y, if desisting from crime is a process of “making good,” and soci ally reconstructing one’s life on a daily basis (Maruna, 2001: 7). Others have suppor ted the conceptualiza tion of desistance as gradual, noting that the accu mulation of social bonds take s time to occur and solidify (Horney, Osgood, and Marshall, 1995). Additionally a conceptually acceptable “cu toff” for the measurement of subjects has remained in question (Farrington, 1979). Fo r example, has an individual desisted if they do not commit a crime during the study peri od? Can this person be labeled as one who has desisted? Or is an i ndividual only truly a de sitant if they are measured till death and have refrained from crime? While most research has labeled a participant as having desisted if they do not comm it a crime during the study period , this factor remains at large to question the validity of the measurement. Therefore, what may actually be under examination is not desistance, but periods of non-offending (Blumstein et al., 1986; Barnett et al., 1987, 1989; Bushway et al., 2001).

PAGE 50

37 One manner in which desistance has been ev aluated is to examine whether or not the behaviors of the individual show a m eaningful change (Laub and Sampson, 1993). This change may come in the form of a m odification, a drop in fre quency of offending, or as an exchange, moving from heroin use to alcohol use only (Laub and Sampson, 1993). These changes can be viewed as “deep” or “meaningful” when accompanied by the development of roles that lead to increasing investment into the so cial structures and thereby the accumulation of social cap ital (Coleman, 1988, 1990; Laub and Sampson, 1993; Nagin and Paternoster, 1992). Given the broad discussion of this topic, it is important to note the manner in which desistance will be conceptualized in this research. Overall, desistance will conceptualized as some form of non-particip ation. As a result of disparate measures throughout the various waves, non-participation will be conceptualized for each wave in accordance with the measures available. All of these measures will be described in chapter 4. Past and future criminal behavior Deviant behavior among children strongly predicts adult criminal behavior (Sampson and Laub, 1990). Research on crim inal careers has found a consistent relationship between past and future crim inality (Piquero, Farrington, and Blumstein, 2003). There have been two main explanat ions for this relationship (Nagin and Paternoster, 1991, 2000). State dependence asse rts that engaging in criminal activity alters ones life circumstances in ways that wi ll increases the probability of future criminal behavior (Nagin and Paternoster, 1991). Fo r example, deviant behavior may weaken the strength of ones conventional bonds to fa mily and friends, or incarceration may negatively impact ones’ relationship with th eir employer thereby negatively impacting

PAGE 51

38 the individual’s ties to empl oyment social structures. The other explanation for the relationship between past and future offending, persistent heterogeneity, states that individuals that hold the highest propensity to commit crime will continue to do so throughout the life course (Gottfredson and Hirschi, 1990). However, some scholars believe that both persistent heterogeniety and state depe ndence influence the positive relationship between past and future devi ant behavior (Nagin and Paternoster, 1991, 2000; Sampson and Laub, 1993; 1997). Theories that address both of these hypot heses are needed as the relationship between offending in youth and adult years has not proven to be completely stable. For example, while anti-social behavior among child ren is one of the best predictors of antisocial behavior in a dults, most delinquent children do not offend as adults (Gove, 1985). Additionally, a majority of adult offenders ha d no history with de linquency as juveniles (McCord, 1979). Research has found that there are a variety of offende r trajectories that diverge from the standard expected ag e-crime curve (Cline, 1980; Laub, Nagin, and Sampson, 1998; Sampson and Laub, 1990). One research problem has been that a large number of false negative and positive predic tions have occurred when examining the relationship between childhood and adulthood criminality (Loeber and StouthamerLoeber, 1987; Farrington and Tarling, 1985) . False positives occur when scales substantially over-predict future criminal beha vior, while false positives fail to accurately predict who will engage in future criminal actions (Loeber and Stouthamer-Loeber, 1987; Farrington and Tarling, 1985; Sampson and Laub, 1992). Nonetheless, research has continued to examine this relationship, and a re view of this research has concluded that

PAGE 52

39 both persistent heterogeniety and state dependence are importa nt factors to examine and address (Nagin and Paternoster, 2000). Different DLC Theories Many perspectives have emerged under the umbrella of developmental life-course criminology. While all of these theories provi de a distinct vision of criminal behavior, each of them contains distinct notions of the mechanisms which influence anti-social behaviors. These theories can be regarded in two distinct categories (Paternoster et al., 1997; Piquero, Farrington, and Bl umstein, 2003). First, these theories may be regarded as either static or dynamic. The distincti on between a static and dynamic theory can be summed up in a list of the questions that dynami c theories address and that static theories fail to consider (Thornberry, 1997). Dynamic theories examine the different dimensions of criminal careers, identify different types of offenders, note the risk factors and results of criminal behaviors, and address developm ental issues outside of criminal behaviors (Piquero, Farrington, and Blumst ein, 2003; Thornberry, 1997). Second, they may be either general or developmental (Pater noster et al., 1997; Piquero, Farrington, and Blumst ein, 2003). This addresses whether or not criminal behavior has a “general” cause among all subjec ts. General theories assume there is a common cause for anti-social behaviors, while developmental theories assert that there are distinct causal pathways among different groups (Paternoster et al., 1997; Piquero, Farrington, and Blumstein, 2003). Nonetheless, th e focus of this section will be to offer a brief description of the various DLC theories, followed by a more detailed discussion of Sampson and Laub’s Age-Graded Informal So cial Control Theory which guides this research.

PAGE 53

40 Moffitt’s Developmental Typology divides offenders into two distinct classes: adolescent limited and life-course persistent (Moffitt, 1993). In this theory neuropsychological dysfunctions or cognitive deficits disrupt normal development (Farrington, 2003; Moffit, 1993). Adolescent li mited delinquency ac counts for a large percentage of youthful anti-soci al behavior. This form of delinquency occurs as a result of their lack of maturity and the encouragement of peers. Primarily this behavior is a form of rebellion in which the adolescent asserts their own independence through the commission of acts normally t hought to be the province of adults (drinking, smoking, sexual behavior). The second category of offe nder, life-course pers istent, engage in a variety of anti-social behavior that includes violence early in life as a result of cognitive deficiencies that make them highly vulnerable to crimonogenic effects. While adolescent limited offenders move through the maturity ga p and adapt new perspectives on the risks and rewards of anti-social beha viors, life-course persistent delinquents fail to evolve into conventional adult roles. Catalano and Hawkins (1996) Social Developmental Model focuses on the balance between anti-social and pro-social bonds that people hold. The strengths of each bond, along with the perceived costs and benefits are the causal pathways that direct people towards or away from anti-social behavior (Catalano and Hawkins, 1996; Farrington, 2003). This theory predicts both drug use a nd delinquency. It examines four distinct phases of social development as people m ove through pre-school, elementary school, middle school, and high school. LeBlanc’s (1997) Interac tive Multilayered Contro l Theory explains the development of offending, crime rates, and criminal events. Offending develops in

PAGE 54

41 relation to an individual’s social bonds, personality development, modeling, and social constraints (Farringt on, 2003; LeBlanc, 1997). The Interactional Theory states that o ffending depends on the development of other life-course trajectories (T hornberry, 1987; Thornberry and Krohn, 2001). This theory also focuses on “reciprocal causation,” noti ng that social bonding influences offending, and that offending also affects the developm ent of social bonds (Thornberry and Krohn, 2001). According to interacti onal theory, desistance is ca used by protective factors (success in employment or education), soci al influences, and intervention programs. Recently, Farrington (2003) proposed an amalgamated theory that takes into account the various issues and pr oblems that have been revealed by previous works. The key construct, anti-social potential, addresse s how likely an individual is to commit an anti-social act. The Integrated Cognitive An ti-social Potential Theory (ICAP) looks at how anti-social potential becomes an anti-soc ial act as a result of the cognitive processes that analyze both opportunities and potential victims. Stra in, DLC, labeling, learning, rational choice, and social control theories infl uenced the development of this paradigm. Sampson and Laub’s Age Graded Informal Social Control Theory examines the various roles and interpersonal relationshi ps that people devel op throughout the lifecourse and the ways in which these connecti ons influence the choi ces that individuals make, especially in regards to criminal be haviors (Laub and Sampson, 1993). This theory is characterized by three main tene ts (Piquero, Farrington, and Blumstein, 2003). First, delinquency during adolescence and childhood can be underst ood in a structural context that is influenced by informal forms of social control. Second, there is a high level of continuity when examining devian ce from childhood to a dulthood. Third, the

PAGE 55

42 quality of social bonds that people accumulate as they en ter adulthood influence changes in criminal behavior. Overall, this th eory acknowledges earl y childhood behaviors and differences, while also advancing the notion that social bonds in adulthood also hold relevance in examining criminal behavior (Sampson and Laub, 1993). Social bonds exist in both formal and informal manners (Kornhauser, 1978; Sampson and Laub, 1993). However, this theory focuses on the forms of informal social control that influence the lives of indi viduals (Laub and Sampson, 1993; Sampson and Laub, 1993). As people form relationships they develop role recipr ocities that shape informal social controls (Kornhauser, 1978; Laub and Sampson, 1993). Sampson and Laub (1993: 18) assert that they are utilizing, “a more general conceptualization of social control as the capacity of a social group to regulate itself according to desired principles and values, and hence to make norms and ru les effective.” Here norms and rules are created and understood by members of various social groups. DLC research projects (including this project), have sought to examine the manner in which membership into such groups as a family, a workplace, or an in stitution of learning re sult in the creation of norms and rules that would lead an individua l away from anti-social behaviors. These notions of norms and rules can be furthe r understood in correla tion with the earlier discussion of social bonds and social capital. Additionally, as noted above, this theory asserts that the strength of social bonds are an influential factor in the cessation of deviant behaviors (Laub, Nagin, and Sampson, 1998; Laub and Sampson, 1993; Sampson and Laub, 1993). This is a logical derivation of the larger theory in that individuals are more likely to commit to the rules and norms of a relationship, st atus, or bond to which they feel a higher level of connection.

PAGE 56

43 The etiology of thei r perspective can be found in the investigation of a “forgotten” longitudinal data source, data collected by th e Gluecks. The Gluecks sought to examine the development of offenders. Their data are unique in that they collected a matched sample consisting of 500 delinquents and 500 non-delinquents. Participants were matched by age, race, intelligence, and residence (Laub and Sampson, 1988). This data was ideal for testing Sampson and Laub’s inform al social control theory as it examined informal social bonds, family and school espe cially, and followed the subjects over time so that the influences of changi ng social bonds could be examined. Cummulative Disadvantage Sampson and Laub’s age-graded informal social control theory extended their developmental conceptualizati on to include an element of labeling theory with their discussion of cumulative disadvantage (L aub and Sampson, 1993; Sampson and Laub, 1997). Cumulative disadvantage refers to the manner in which anti-social behavior has a “systematic attenuating effect on the social and institutional bonds linking adults to society.” (Sampson and Laub, 1997: 144). The e ffects of labeling create a cycle wherein the individual’s connections with family, p eers, school, and the criminal justice system are increasingly negatively influenced (Sampson and Laub, 1997). In this downward spiral, criminal acts directly and indirectly modify the li kelihood of future criminal behavior (Nagin and Paternoster, 1991). Directly, official reactions (arrests, incarceration) to primary deviance ca n create problems (unemployment, underemployment) that increase the likeli hood of secondary deviance (Becker, 1963; Laub and Sampson, 1993; Lemert, 1951; Tittle, 1988). Two of the underlying elements of cumula tive disadvantage are institutionalization and stigmatization (Caspi et al., 1987). In stitutionalization has b een found to weaken

PAGE 57

44 social bonds with friends and family (Wheeler, 1961; King, 1993). During institutionalization the pro cess of “prisonization” may occur wherein inmates are inculcated into the subculture of pris on that is both hypermasculine and violent (Carrabine and Longhurst, 1998; Clemmer, 1940; Messerschmidt, 1993; Newton, 1994; Sabo et al., 2001). Prisonization is often tie d to a weakening of family ties and an increase in isolationist behavi or with regards to relationshi ps outside of the prison walls (Clemmer, 1940; King, 1993). This can be s een in that fathers frequently report deterioration in closeness with their ch ildren while incarcerat ed (Lanier, 1993). Research has also found confirmation of the stigmatization mechanism of cumulative disadvantage. Those who serve l onger sentences have a more difficult time obtaining and maintaining em ployment (Sampson and Laub, 1997). Similarly, future opportunities are curt ailed for males with serious record s of youthful deviance (Caspi and Moffitt, 1993; Moffitt, 1993). The effects of labeling are harsher for lower class boys than among middle class boys (Hagan, 1991). Certainly, the cycl ical nature of cumulative disadvantage can result in a dow nward spiral for young men who have been officially sanctioned for anti-social behavior . As the negative eff ects of such labels continue to shape their socialization pa tterns and opportunities, the likelihood of accumulating informal social bonds that will steer these young men away from anti-social acts continually diminishes. This aspect of Sampson and Laub’s age-graded informal social control theory will not be incorpor ated into the models here. However, the influence of labeling (via arre sts and incarceration) on the youth in this sample is a major future interest that will be pursued.

PAGE 58

45 Major DLC Findings DLC research has examined the manner in which informal social bond have influenced deviant behavior. While some of these relationships were discussed at length in Chapter 1, a brief recounting of so me of this information is relevant. Family Many scholars have examined the role that family plays in explaining criminal behavior (Farrington, 1987; Hirschi, 1969, 1983; Loeber a nd Stouthamer-Loeber, 1986; Loury, 1987; Wilson, 1983). Specifically, the role that marriage plays in desistance has been paramount (Farrington and West, 1995; Gibbens, 1984; Knight, Osborn, and West, 1977; Laub, Nagin, and Sampson, 1998; Osbor n and West, 1979; Piquero, MacDonald, and Parker, 2002; Sampson and Laub, 1990; Warr, 1998; West, 1982). Marriage has been found to lead to an increase in “socia l stability” among males (Gibbens, 1984: 64). While some research has found a decrease in criminal offenses after marriage (Osborn and West, 1979; West, 1982), othe rs have only found a decrease in anti-social behavior (Knight, Osborn, and West, 1977). One st udy examining the relationship between marriage and criminal activities found gett ing married to be non-significant, while staying married was the most significant fact or predicting convicti ons (Farrington and West, 1995). This study also found that se paration held a strong correlation with convictions when using self-report da ta (Farrington and West, 1995). Further examination has revealed that the stability, the level of attachment, and the satisfaction level of marital relationships st rongly influences desistance ra ther than simply the marital status itself (Laub, Nagin, and Samp son, 1998; Sampson and Laub, 1990). One of the pathways through which marriag e influences criminal behavior is through the management of the husband’s soci al interaction with others. Following

PAGE 59

46 marriage the amount of time available to sp end with friends is substantially reduced (Warr, 1998). Marriage also alters the type s of people that one interacts with, seeing decreasing interactions with deviant peers (W arr, 1998). While all persons decreasingly associate with delinquent peers through the life -course, there is a more significant decline for married people (Warr, 1998). Farrington and West’s (1995) re trospective interviews with older men revealed that their desistance from criminal behavior was shaped by the desire to settle down and the ability to obtain and maintain an intimate re lationship with women. However, this goal may be difficult to achieve for older offende rs as arrests and incarceration reduce an individual’s marriagability (C ohen, 1999). This could be viewed as another example of the cumulative negative effects that devian t behavior may hold for lower class men (see cumulative disadvantage as discussed above). N onetheless, it would appear that in some cases even the desire for normal social bonds may lead individuals down a path to desistance that occurs prior to the accumulation of such bonds. While marriage has been seen to influence desistance, some scholars hold that the presence of children in the home holds great er weight (Popenoe, 1996). Some research has upheld this notion finding that having a ch ild has had a greater effect on reducing the social habits associated with offending (d rinking, using drugs) than marriage does (Knight, Osborn, and West, 1977). In accordance with the life-course concept of social timing, men who father children at least 9 months into a marriage, rath er than earlier, are more likely to be non-offenders than men who have fathered a child either outside of marriage or before 9 months have passed w ithin a marriage (Farrington and West, 1995). Additionally, fathering a child outside of marriage is significantly related to a higher

PAGE 60

47 likelihood of being an offender (Farrington a nd West, 1995). Similar to marriage, the strength of the relationship that a father holds with their child may also be influential, as separation from one’s child is significantly related to offending as well (Farrington and West, 1995). Earlier in the life-course, the parenting th at these same subjects receive as children influences the likelihood of criminal behavior (Laub and Sampson, 1988). Socioeconomic disadvantage has an adverse eff ect on parenting (Rutter and Giller, 1983). Lack of supervision, parent-child involvem ent, along with parental rejection are among the most powerful predictors of juvenile delinquency (Loeber a nd Southamer –Loeber, 1986). Laub and Sampson’s (1988) review of th e Gluecks’ data revealed that family process variables are directly and significantly related to serious and persistent delinquency. Residential mobility during child hood was also found to have a significant positive relationship with delinquency (Laub and Sampson, 1988). Social Participation As was noted earlier, participation in and tie s with social institutions predict lower levels of anti-social behavior. Research not es that “subjects with high aspirations and efforts to advance educationally and occupati onally were much less likely to engage in deviant behavior, use alcohol excessivel y, or be arrested” (Sampson and Laub, 1990: 618). Outside of the family trajectories, w hose influence has been discussed, one sees that educational and occupationa l trajectories can also modify the criminal trajectories in the lives of offenders. The effects of em ployment have been found to be strong in disparate directions, as both unemploymen t and job stability have been found to significantly influence anti-social behavior (all be it in differe nt directions) (Farrington et al., 1986; Sampson and Laub, 1990; Sampson and Laub, 1993).

PAGE 61

48 The previous research and the theo ries discussed above shaped the conceptualization and applicati on of variables. Overall, th is research advanced prior knowledge in two key areas. First, this proj ect provids a deeper level of knowledge about the ways in which fatherhood influences partic ipation in deviant behavior. While there are a few examples of studies that exam ine the influence of fatherhood on deviant behavior through the life-course (Edin et al., 2004; Farri ngton and West, 1995; Knight, Osborn, and West, 1977), the impact of fath erhood upon these behaviors seems to have been severely overlooked (Edin et al., 2004). Therefore, this proj ect adds to prior literature via an expansion of the understa nding of the ways in which fatherhood may influence deviant behavior. The second manner in which this project adds to prior research is that the sample varies distinctly from those used in prior longitudinal works within developmental lifecourse criminology. Research framed within Sampson and Laub’s Aged Graded Theory of Informal Social Control has examined samples that are composed completely, or primarily of whites (Horney et al., 1995; Nagin, Farrington, and Moffit, 1995; Nielsen, 1999; Sampson and Laub, 1990, 1993). This is key, as persons of various racial, ethnic, and social groups may experience the accumula tion and loss of social bonds in disparate ways that may hold various influences on part icipation or cessation of deviant behaviors (Laub and Sampson, 1993; Nielsen, 1999). As this research holds a sample comprised entirely of African Americans, it will provi de answers to the previously mentioned question. This research provide s an in-depth examination of the influence of social bonds on the criminal behaviors of African Amer ican men over the course of 26 years.

PAGE 62

49 CHAPTER 4 METHODS This chapter discusses the methodological issu es and directives of this research. The issues that are inherent in the use of longitudinal self-repo rt data are discussed. First, a review of longitudinal a nd self-report data is conducted separately. Second, the challenges and problems within both of these fo rms of data are examined together as the data utilized in this project are both l ongitudinal and self-report. Following this discussion of issues, the data itself is review ed. This will entail a review of the data, the manner in which they were collected, and a look at the initial sample. Next, the main research questions that this project seeks to address are reviewed. Then, the various analytical strategies that were utilized are discussed. In th is section an overview of the various models for this project are illustrate d. After the various models are discussed, an in-depth review of the myriad variables that composed these models is presented. After the variables of interest are reviewed, th e hypotheses that this project sought to investigate are noted. The end of this chap ter also addresses subj ect attrition, and how this was investigated. Methodological Issues Longitudinal Data There are two forms of longitudinal da ta: cohort and panel. A cohort study examines the changes in specific samples (i .e. a birth cohort). A panel study examines the same set of people as they are measured two or more times (Piquero, Farrington, and Blumstein, 2003). Longitudinal research designs can be eith er prospective or

PAGE 63

50 retrospective. A prospec tive longitudinal research de sign follows a cohort over an extended period of time. A retrospective desi gn asks participants to provide information about previous periods in tim e. Interestingly, the data used for this project are a prospective design that contains some re trospective data as some questions ask participants to recall events that have occurr ed more than 10 years previously. While the use of a prospective design ma y seem superior there are drawbacks to this design. Specifically, these studies are more expensive, and are more likely to be complicated by history, testing, period, and panel effects, as well as sample attriti on (Piquero, Farrington, and Blumstein, 2003). Longitudinal designs hold a variety of benefits that canno t be found in crosssectional research. First, longitudinal data allows rese archers to obtain precise time ordering of events in the part icipants’ lives (Blumstien et. al., 1988a). When researchers attempt to support hypotheses abou t causal relationships, one of th e key criteria is that the causal variable occurs before the effect that is measured. In l ongitudinal studies, each phase of data collection sets in stone a marker of when each variable of interest has occurred. This facet of longitudinal data collection is needed by criminology researchers “for estimating participation rates and the ons et, termination, and duration of careers; for distinguishing frequency from participat ion; and examining continuities (or discontinuities) in offending over tim e” (Blumstein et. al., 1988a: 30). Another core benefit of this type of research, which cannot be found in crosssectional studies, is the ability to examine changes within subjects. Analysis of the influence of age reveals different stories in cross-sectional and l ongitudinal research. Longitudinal data reveals differences within cohorts, while cross-se ctional data reveals

PAGE 64

51 differences between cohorts (Menard, 2002). Cross-sectional studies can only examine the differences between subjects. Longitudi nal studies are capable of both. The final benefit that is found solely in longitudinal data is that “there is better control of extraneous variables, because each person act s as his own control” (Blumstein et. al., 1988a: 28). Self Report Data Self-report data are one of the three major forms of crime data, and has been one of the most important innovations in research on crime in the twentieth century (Thornberry and Krohn, 2000). These reports or surveys are composed of a series of questions asked of respondents who may or may not have had pr evious contact with authorities (Jackson, 1990). Self report surveys have been used extensively within the United States and internationally (Klein, 1989; Thornberry and Krohn, 2000). Self-report data have been especially important in research on crim inal careers (Junger-Tas and Marshall, 1999; Piquero, Farrington, and Blumstein, 2003; Thornberry and Krohn, 2000). This form of data collection arose in res ponse to the alleged “dark figure” of crime (Hindelang et. al., 1981; Mosher et. al., 2002). Scholars noted that their main source of data, official records, was inept at captur ing a vast array of deviant and criminal behaviors (Merton, 1938; Sell in, 1931; Sutherland, 1939). A major problem of official data was that it only presented the activity of law enforcement personnel (Sellin, 1931). Specifically, the reliance on official data allo w for an unacceptable amount of distance between the criminal acts and their rec ognition, which provide an opportunity for a variety of biases that may di stort the validity of the data being measured (Sellin, 1931; Thornberry and Krohn, 2000). Sutherland’s (19 49) work on white-collar crime has been cited as one of the core influences on the re cognition of the need fo r alternative forms of

PAGE 65

52 crime data as it revealed that the accepted relationship between so cial class and crime was not as secure as it was once thought (Gibbens, 1979; Thor nberry and Krohn, 2000). The first published results from self-repor t data about criminal behavior appeared in the 1940’s (Porterfield, 1943, 1946). Howe ver it was the late 1950’s and early 1960’s that saw a rise in the popularity and use of self-report studies of deviance (Cernkovich, Giardano, and Pugh, 1985; Thornberry and Krohn, 2000). This occurred because scholars thought that self-repor t surveys of a general youth population would be more representative and hence more appropria te in the study of juvenile delinquency (Cernkovich, Giardano, and Pugh, 1985). Shor t and Nye’s (1957, 1958) examination of self-reports of delinquency among high school stude nts revealed that there was little to no difference in the anti-social behavior of their sample by social class. The work of Short and Nye (1957) helped not only to put se lf-reports on the me thodological map of deviance research, but also helped to reveal some of the biases that are inherent in official data. This new methodology answered the cr itiques of scholars such as Sutherland (1947) and Sellin (1931), who noted that offici al reports were unable to capture the dark figure of crime. There have been many nationwide collecti ons of self-report data, such as the National Youth Survey and Monitoring the Fu ture, among others (Elliott, Huzinga, and Ageton, 1985; Johnston, O’Malley, and Bach man, 1996; Thornberry and Krohn, 2000; William and Gold, 1972). In addition to these studies, other collections of large selfreports of deviant behavior have been conducte d in the United States. Large-scale studies such as the RAND Inmate Surveys and the Arrestee Drug Abuse Monitoring (ADAM) program have produced interesting data that helps to paint the tapestry of modern

PAGE 66

53 criminology (Chaiken and Chaiken, 1982). Th e RAND surveys have revealed that most inmates commit only a few crimes per year, a nd that there is a small population (around ten percent) that commits a multitude of cr iminal acts in a given year (Chaiken and Chaiken, 1982; Mosher et al ., 2002; Visher, 1986). The da ta from the ADAM program informs us that over 50% of the arrestees in terviewed for violent offenses had recently used a controlled substance. Additionally, the prevalence of drug use, along with the prevalence of specific drugs, vari ed from location to location. Longitudinal and Self-Report Data: Methodological Issues As noted, the data used in this research were both longitudinal and self-report. Some of the general issues that have been noted in relation with each form of data are reviewed. Next, some larger i ssues that have been noted in longitudinal self-report data are addressed. The issues of sampling, attr ition, validity, reliabil ity, testing and panel effects, as well as construct continuity are discussed at length. After these issues are illustrated, the data is reviewed. Longitudinal designs are faced with a number of methodological issues. Longitudinal studies need to address the prob lems of internal validity that occur as a result of sample attrition, mortality, and maturation (Blu mstein et. al., 1988b; Cook and Campbell, 1979). Longitudinal studies are more expensive and take longer to gather when compared to cross-sectional studie s (Gottfredson and Hi rschi, 1988; Menard, 2002). Testing and period effects, which will be discussed at length later, may be misrepresented as changes in the behavior of participants (Brame and Piquero, 2003; Jang, 1999; Lauritsen, 1998, 1999; Thornberr y, 1989; Woltman and Bushery, 1984). Specifically, repeated measures and panel c onditioning may corrupt pa rticipant responses and damage internal validity (Menard, 2002). Left-hand censo ring is another issue faced

PAGE 67

54 in longitudinal research as ch anges that have occurred prior to the first wave of data collection may hold influence that has not been noted as a factor (Menard, 2002). Over the past 30 years self-report survey s of delinquency have shown substantial methodological improvement (Thornberry, 1998). Nonetheless, this data, like all data, must confront a myriad of methodological issu es. As self-reports are a form of survey data, they contain all of the problems that are inherent in any survey research. Therefore, self-report data are in danger of suffering fr om four types of error that may occur in surveys: coverage error, sampling error, non-response error, a nd operationalization (Mosher et. al., 2002). Scholars have addresse d some of the problem s with self-report data, noting that serious offens es are either left out or underrepresented within many selfreport scales (Cernkovich, Giardano, and Pugh, 1985; Elliott and Ageton, 1980; Hindelang, Hirschi, and Weis, 1979; Thor nberry and Krohn, 2000). Overlapping items which result in inaccurate frequency counts ha ve also been a problem as events may be counted more than once (Cernkovich, Giar dano, and Pugh, 1985; Thornberry and Krohn, 2000). Finally, another concern with self-repo rt delinquency studies has been the use of categorical frequency counts that limit the am ount of interpretation av ailable, especially when dealing with chronic offenders (C ernkovich, Giardano, and Pugh, 1985; Elliott and Ageton, 1980; Thornberry and Krohn, 2000). Sampling Sampling has been both a problem and su ccess for self-report surveys. These positive and negative attributes of self-report su rveys are entrenched within the issues of sampling and selection bias. Self-report surv eys have been purported to provide a wider sample of the youthful populat ion that is not captured by o fficial statistics (Cernkovich, Giardano, and Pugh, 1985; Elliott and Ageton, 1980) . These general youth samples have

PAGE 68

55 been credited with advanci ng our knowledge about the participation and frequency of delinquency among the youth, as well as th e changing rates of such behavior (Cernkovich, Giardano, and Pugh, 1985) . In this manner, these surveys have dealt with selection bias in which the processes of the criminal justice system have created a sample of offenders who are not repr esentative of offenders at large (Blumstein et al., 1986; Piquero, Farrington, and Blumstei n, 2003). Through the proce sses of arrest, conviction, and incarceration, these offender samples dist ance themselves from the acts they attempt to measure (Piquero, Farrington, and Blumstei n, 2003; Sellin, 1931). The use of general youth samples has reduced selection bias by minimizing the distance between crime and its measurement. However, surveys using general youth samples also have been critiqued because they either under-sample, or entirely miss sampling, chronic, or more serious, offenders (Brame and Piquero, 2003; Cernkovich, Giar dano, and Pugh, 1985; Piquero, Farrington, and Blumstein, 2003). Even methodologically sophisticated genera l youth samples have failed to adequately attain sufficient numb ers of these serious juvenile offenders (Cernkovich, Giardano, and Pugh, 1985). Be yond initial sampling, these chronic youthful offenders have also been found to be difficult to retain in longitudinal studies (Brame and Piquero, 2003; Krohn and Thornberry, 1999). One methodological adjustment that has been suggested to a ddress sampling bias is to over-sample low income populations as they are more likel y to contain serious offenders (Piquero, Farrington, and Blumstein, 2003). Additionall y, these general population samples have been sited as being more appropriate for measuring participation rates rather than frequencies (Piquero, Farring ton, and Blumstein, 2003). Th e previous two statements

PAGE 69

56 have influenced the development of this research as they are fulfilled by the sample and methodology of this project. First, the enti re sample is composed of a low income neighborhood (Central Harlem, New York). A dditionally, the dependent variable in the research examines participation in antisocial behaviors rath er than frequency. Sampling Attrition Another issue with the use of longitudina l self-report data is that of subject attrition. This refers to the loss of resear ch participants as time passes (Brame and Piquero, 2003; Little, 1995; Manski, 1995). The methodological issues of longitudinal data are further complicated if the attrition of subjects occurs in a non-random fashion (Brame and Piquero, 2003, Menard, 2002; Pi quero, Farrington, and Blumstein). This form of sample attrition can result in seri ously mistaken estimations of the various dimensions of criminal careers (P iquero, Farrington, and Blumstein, 2003). Multiple studies have sought to examine the level and manner in which subject attrition may assert inappropriate influence over research findings. Cordray and Polk’s (1993) analysis of seven panels revealed th at subject attrition he ld no impact over the bivariate and multivariate relationships in question. Krohn and Thornberry (1999) also failed to find significant influence from attri tion. However, Thornberry and colleagues (1993) did find attrition that led to bias in their examination of the Rochester Youth Development Study (RYDS). There are a couple of ways to deal with missing data as due to attrition (Brame and Piquero, 2003; Menard, 2002). Fi rst, if there are missing items one may substitute a scale mean for the missing item (Brame and Pi quero, 2003; Menard, 2002). Second, the data may be analyzed without making any adjust ments (Brame and Piquero, 2003). Finally, the remaining cases may be weighted thr ough either interpolati on or extrapolation

PAGE 70

57 (Brame and Piquero, 2003; Menard, 2002). Briefly, the manner in which subjects retained varied from subjects lost was examined at the end of this chapter. Reliability and Validity A central concern in all research is the re liability and validity of the measurements utilized. Thornberry and Krohn (2000: 44) addr essed the notion of reliability stating, “No measure is absolutely, perfectly reli ablethe central question in assessing the reliability of self-reported delinquency measur es is not whether the measure is reliable but how reliable it is; reliability is always a matter of degree” (emphasis in original). The two main methods by which reliability ha s been assessed are through the test-retest method, and through an examination of the inte rnal consistency of measures (Thornberry and Krohn, 2000). Internal consistency is esta blished if multiple yet disparate measures of the same concept hold a high level of co rrelation. Analysis of self-report measures using the test-retest method have revealed a high level of reliability (Hindeland, Hirschi, and Weis, 1981; Huzinga and Elliott, 1986; Thornberry and Krohn, 2000). Validity examines the extent to which a variable measures what is under examination (Thornberry and Krohn, 2000). The validity of items ca n be corrupted in three ways. The content validity of a measurement addresses the core concern of validity in that the item measures what it is inte nded to gauge. The key concern is that the measure fully and clearly covers the domain of interest (Thorberry and Krohn, 2000). Construct validity considers the degree to whic h a variable is liked to other measures as theoretically expected (Thor nberry and Krohn, 2000). Criteri on validity looks at the degree to which a measurement reflects result s similar to the findings of a similarly measured quantity that has been previously accepted as an adequate measurement (Huzinga and Elliott, 1986; Thronberry and Krohn, 2000). Comparisons of self-reports

PAGE 71

58 of deviance to official records have found that there is a subs tantial amount of under reporting of past criminal be havior (Thornberry and Kronh, 2000). Studies questioning the criterion validity of self-report surveys have revealed that there are lower levels of validity when it co mes to reporting serious offenses (Elliott and Voss, 1974; Huzinga and Elliott, 1986; Mosher et. al., 2002). Research has also found that lower class youths (Braith waite, 1981), African American males (Hindeland, Hirschi, and Weis, 1981; Huzinga and Elliott), and some times males in general (Kim et. al., 2000) tend to underreport their involvement in delinque nt behavior, when co mpared to official records. However, other examinations of self-report data have failed to find discrepancies in the validity of self-re ported delinquency among African American males (Farrington et al., 1996). Overall, self-report surveys have been accepted as valid (Akers et. al., 1979; Hindelang et. al., 1979; Huiz inga and Elliot, 1984; Johnson, 1979; Thornberry and Krohn, 2000). Recent self-report items have b een noted as achieving an acceptable level of content validity, a high leve l of construct validity, and a moderate to strong level of criterion valid ity (Thornberry and Krohn, 2000). Testing, Panel, and Period Effects Another set of challenges that have accomp anied the use of longitudinal, self-report data are concerns that the obser ved variations in be havior are not exampl es of real change but really the results of te sting, period, and panel effect s (Brame and Piquero, 2003; Jang, 1999;Lauritsen, 1998; Thornberry, 1989; Woltma n and Bushery, 1984). Testing effects are changes in participants’ responses cau sed by prior exposure to the same item (Thornberry, 1989). In delinque ncy studies, panel effects ar e found when declines in offending are a result of the decrease in offendi ng that occurs with age (Lauritsen, 1998; Osgood et al., 1989; Piquero, Farrington, and Blumstien, 2003; Thornberry, 1989).

PAGE 72

59 Period effects occur when there is a downward trend in crime that ma y not indicate a real decline (Piquero, Farrington, and Blumstien, 2003; Thornberry, 1989). Research using this form of data must address the challenge of utilizing an outcome variable that maintains theoretical signifi cance, while at the same time avoiding the influences of testing effects and changi ng content validity as the respondent ages (Lauritsen, 1998: 149). The c ontent validity of items may va ry as respondents age, and among respondents of different ages (Lau ritsen, 1998; Piquero, Farrington, and Blumstein, 2003; Piquero, MacIntosh, and Hickman, 2002). Construct Continuity Finally, self-report surveys examining crimin al behavior over the life-course must contain items that take into account various contexts within which individuals engage in criminal behaviors (Thornberry and Krohn, 2000). The goal is to create items that hold construct continuity, or the ab ility to measure different factors that contain the same underlying feature (Thornberry and Krohn, 2000). For example, work in adulthood may measure the same social influence as school during adolescence. Th e key is to include different, yet theoretically linke d, items that address the same issue across various points in the life-course (LeBlanc, 1989). Methods of Research Data Source Data used in this study came from the Harlem Longitudinal Study of Urban Black Youth, 1968-1994 , which intended to focus on health issues within the New York community of Harlem. Dr. Ann Brunswick s ought to examine cha nges in health, drug use, and mediating factors such as family background. Research using this data set has examined changes in health and health probl ems, use patterns of and abstinence from

PAGE 73

60 illicit drugs, causal factors in cigarette sm oking, as well associations between substance use and health (Brunswick, 1976, 1980, 1984; Brunswick and Boyle, 1979; Brunswick and Messeri, 1985, 1986; Brunswick, Merzel, and Messeri, 1985). In order to delve into the lived expe riences and changing lives of adolescent African Americans in Harlem, Brunswick organized and conducted her study as a longitudinal project. The collection of data lasted for 26 years and consisted of five waves of data collection. The initial data was collected from 1968 to 1970. The subsequent four stages of data co llection occurred during 1975-1976, 1983-1984, 19891990, and 1993-1994. The Henry A. Murray Resear ch Center, at Radcliffe University, made this data available. Data Collection Data were collected in five separate wa ves that spanned 26 years. The sample consisted of youths that had been interv iewed in the Community Health Survey conducted by the Harlem Hospital and Columbia University School of Public Health. The sample consisted of an area probability sampling of the designated area. These young people represented a cross section of households in Central Harlem. During Wave 1 (1968-1970) 752 black adoles cents were interviewed. However, youths with a Latino background were dropped from the final sample, as the number of Black Hispanics was too small to be used in a separate analysis. The final sample size at the end of Wave 1 was 668 youths (See Table 4-1). In Wave 1, as well as during all subse quent waves of the study, gender-matched African American interviewers conducted data collection. Researchers worked from structured interview schedules , and interviewed participants within their own homes. Typically, these in terviews lasted between 1.5 to 2.5 hour s. Wave 1 also consisted of a

PAGE 74

61 free medical exam. The first interview cons isted of 126 close-ended questions focusing on physical health, smoking, drin king, emotional health, soci al health, aspirations and expectations about marriage and parenting, pe er and family relationships, as well as a number of demographic characteristics. Table 4-1 Harlem Longitudinal Stud y of Urban Black Youth Sample WAVE YEARS AGE SAMPLE SIZE MALE SAMPLE 1 1968-1970 12-17 668 351 2 1975-1976 18-23 536 277 3 1983-1984 26-31 426 210 4 1989-1990 31-38 364 177 5 1993-1994 35-41 347 168 Wave 2 used the same interview schedul e as Wave 1, but did not include a free medical exam. However, interviewers provide d participants with a “Request for Medical Assistance” form that could assist them in receiving needed medical care. Additionally, all participants were paid $10.00 for their inte rview. Data collection expanded in Wave 3 in order to collect a wider range of data on i ssues such as drugs us e, pregnancy history, use of obstetrics and abortion services, institutional experiences, residential mobility and composition, marital status, and sources of in come. Wave 4 utilized many of the same questions from Wave 3, but al so expanded the use of questions examining HIV/AIDS awareness and experiences. Finally, Wave 5 reproduced many Wave 4 questions and asked for summaries of each subject’s participation in the entire study. Throughout the course of the research, samp le attrition was caused mainly by an inability to relocate participants or as a result of their death. As each new wave approached, participants were located though a specific 13-step sequence. Subjects were recontacted using the following procedure: “(1) Two first-cl ass letter mailings were sent, roughly a month apart, enclosing an answer form and a stamped, addressed return

PAGE 75

62 envelope. Forwarding and address correcti on were requested from the post office. Nonreturn after two mailings was considered tentative evidence that respondent probably lived at that location. (2) An attempt at telephone contact was made when a phone number was available from the earlier inte rview. When there was no usable number, reverse telephone directoriesone current and one from th e period of the initial studywere matched to identify someone still livi ng in the respondent's old building who may know where the respondent or the family moved. (Reverse directories, sometimes called street directories, list addre sses and then identify occupants with telephones.) (3) A check was conducted with the Harlem Hospital r ecord Room and othe r usual provider of medical care, as specified in the initial interview. (4) A borou gh-by-borough post office address-correction search was undertaken. (5) A certified letter (with receipt) was sent to the new address provided by the post office s earch and also to those respondents whose two regular mailings had not been returned a nd who had not yet completed answer forms. (6) Names of all respondents who were not located through the prior five steps were forwarded for the search to the Social Serv ice Exchange, a voluntar y association of over 300 public and private social and health agen cies in New York City. In the second and larger phase of the follow-up, this step pr eceded steps 4 and 5 in order to obtain the relevant information prior to the shut down of the Social Service Exchange. (7) Marriage and birth certificates were checked at the New York City Department of Vital Statistics. (8) The death register was searched. (9) Poli ce records were searched. (10) Welfare scans and individual welfare centers searches we re conducted, especial ly for young mothers. (11) Department of Probation records were searched. (12) A special search was

PAGE 76

63 undertaken when review of th e initial interview schedule su ggested another alternative for hard-to-find case.” (Brunswick, 1984). Sample The initial sample consisted of 668 African American youths between the ages of 12 to 18 years of age. There were slightly more males (351) than females (317) in the initial sample. During the in itial wave of data collecti on these young people resided in Central Harlem, within New York City. The sample consisted of 351 males and 317 females. During Wave 1 interviews the age of participants varied as follows: 22.8% of the sample were 12 years old, 21.6% were age 13, 18.4% were age 14, 18.1% were age 15, 8.4% were age 16, 9.9% were age 17, and 0.9% were age 18. Most of the sample (79.4%) was born in New York City. Over one-third (36.8%) of the sample were a part of families that received welfare, and an additional nine percent received benef its other than welfare. However, over half of the participants (53.6%) were in families that did not receive any welfare benefits. Adolescents who completed the first interview process were more likely to have mothers who did not complete high school, as only 27.5 % of the initial sample had mothers who attained a high school degree or more. Simila r statistics were found when examining the education of participants’ fath ers as only 31% held a high sc hool degree or greater. Ninety-five percent of the sample was enrolled in school at the beginning of the surveys. Almost two-thirds of the sample was in elementary school during Wave 1. Approximately one-quarter of the sample held a “special disability” according to school records. These disabilities ranged from a reading problem to a behavioral-emotional problem. At the first interview the mean GPA of the participants was just below a 2.0, or

PAGE 77

64 “C”, average. However, the grades of female participants were higher than the marks of their male counterparts. Research Questions: Drug Use The next section of this chapter provided a list of research questions. They were divided into three sections, one for each de pendent variable. Within each of these sections conceptually linked groups of inde pendent variables were discussed. Some of the research questions below we re applied to a limited number of models. Not all of the variables were used in either lagged or desistance models for various reasons. The logic behind these decisions were discussed later wh en the variables themselves are discussed at length. Age 1. Age of Participant: a. Among men, does the age of a partic ipant influence the likelihood of using drugs in the same wave? b. Among men, does the age of a partic ipant influence the likelihood of using drugs in the subsequent wave? c. Does age influence desistance from using drugs? Family: Marriage 1. Marital Status: a. Among men, does being married in fluence the likelihood of using drugs in the same wave? b. Among men, does being married in fluence the likelihood of using drugs in the subsequent wave? c. Does being married influence desistance from using drugs? 2. Marital Satisfaction: a. Among married men, does the level of satisfaction in marital relationships influence the likelihood of using drugs in the same wave? b. Among married men, does the level of satisfaction in marital relationships influence the likelihood of using drugs in the subsequent wave? Family: Fatherhood 1. Fatherhood Status:

PAGE 78

65 a. Does fatherhood influence the likeli hood of using drugs in the same wave? b. Does fatherhood influence the lik elihood of using drugs in the subsequent wave? c. Does fatherhood influence desistance from using drugs? 2. Fatherhood Prior to the Age of 18: a. Among men, does fathering a child prio r to the age of 18 influence the likelihood of using drugs in the same wave? b. Among men, does fathering a child prio r to the age of 18 influence the likelihood of using drugs in the subsequent wave? c. Among men, does fathering a child prior to the age of 18 influence desistance from using drugs? 3. Residential Father: a. Among men, does residence with so me or all of their children influence the likelihood of usi ng drugs in the same wave? b. Among men, does residence with so me or all of their children influence the likelihood of using drugs in the subsequent wave? c. Among men, does residence with so me or all of their children influence desistance from using drugs? 4. Rate Self as Father: a. Does the way men rate themselves as fathers influence the likelihood of using drugs in the same wave? b. Does the way men rate themselves as fathers influence the likelihood of using drugs in the subsequent wave? 5. Feelings about their Relationship with their Child: a. Do the way men rate their relationship with their children influence the likelihood of using drugs in the same wave? b. Do the way men rate their relationship with their children influence the likelihood of using drugs in the subsequent wave? Residential Status 1. Residential Stability a. Among men, does the length of reside nce in the same place influence the likelihood of using drugs in the same wave? b. Among men, does the length of reside nce in the same place influence the likelihood of using drugs in the subsequent wave? c. Among men, does the length of reside nce in the same place influence desistance from using drugs? Employment 1. Employment Status a. Does employment status influence men’s likelihood of using drugs in the same wave? b. Does employment status influence men’s likelihood of using drugs in the subsequent wave?

PAGE 79

66 c. Does employment status influence desistance from using drugs? 2. Job Satisfaction a. Does the level of satisfaction in one’s job influence men’s likelihood of using drugs during the same wave? b. Does the level of satisfaction in one’s job influence men’s likelihood of using drugs during the subsequent wave? c. Does the level of satisfaction in on e’s job influence desistance from using drugs? Finances 1. Income a. Does the annual level of men’s in come influence the likelihood of using drugs within the same wave? b. Does the annual level of men’s in come influence the likelihood of using drugs within the subsequent wave? c. Does the annual level of men’s in come influence desistance from using drugs? 2. Welfare Use a. Does welfare use among men influe nce the likelihood of using drugs during the same wave? b. Does welfare use among men influe nce the likelihood of using drugs during the subsequent wave? c. Does welfare use among men influe nce desistance from using drugs? Education 1. GPA a. Does the Grade Point Average of yo ung men predict using drugs in the subsequent wave? 2. High School Diploma a. Does the attainment of a high school diploma influence the likelihood of using drugs within the same wave? b. Does the attainment of a high school diploma influence the likelihood of using drugs within the subsequent wave? c. Does the attainment of a high school diploma influence desistance from using drugs? Prior Behavior 1. Bad Behavior a. Does bad behavior in Wave 1 infl uence the likelihood of using drugs in the subsequent wave? b. Does bad behavior in Wave 1 infl uence desistance from using drugs? 2. Drug Use a. Does prior drug use influence the likelihood of using drugs in the same wave?

PAGE 80

67 b. Does prior drug use influence the likelihood of using drugs in the subsequent wave? 3. Arrests a. Do prior arrests influence the like lihood of using drugs in the same wave? b. Do prior arrests influence the li kelihood of using drugs in the subsequent wave? Research Questions: Arrests Age 1. Age of Participant: a. Among men, does the age of a partic ipant influence the likelihood of being arrested in the same wave? b. Among men, does the age of a partic ipant influence the likelihood of being arrested in the subsequent wave? c. Among men, does the age of the par ticipant influence the likelihood of desistance from being arrested? Family: Marriage 1. Marital Status: a. Among men, does being married in fluence the likelihood of being arrested in the same wave? b. Among men, does being married in fluence the likelihood of being arrested in the subsequent wave? c. Does being married influence de sistance from being arrested? 2. Marital Satisfaction: a. Among married men, does the level of satisfaction in marital relationships influence the likelihood of being arrested in the same wave? b. Among married men, does the level of satisfaction in marital relationships influence the like lihood of being arrested in the subsequent wave? Family: Fatherhood 1. Fatherhood Status: a. Does fatherhood influence the likelihood of being arrested in the same wave? b. Does fatherhood influence the likel ihood of being arrested in the subsequent wave? c. Does fatherhood influence desistance from being arrested? 2. Fatherhood Prior to the Age of 18: a. Among men, does fathering a child prio r to the age of 18 influence the likelihood of being arrest ed in the same wave?

PAGE 81

68 b. Among men, does fathering a child prio r to the age of 18 influence the likelihood of being arrested in the subsequent wave? c. Among men, does fathering a child prior to the age of 18 influence desistance from being arrested? 3. Residential Father: a. Among men, does residence with so me or all of their children influence the likelihood of being arrested in the same wave? b. Among men, does residence with so me or all of their children influence the likelihood of being a rrested in the subsequent wave? c. Among men, does residence with so me or all of their children influence desistance fr om being arrested? 4. Rate Self as Father: a. Does the way men rate themselves as fathers influence the likelihood of being arrested in the same wave? b. Does the way men rate themselves as fathers influence the likelihood of being arrested in the subsequent wave? 5. Feelings about their Relationship with their Child: a. Do the way men rate their relationship with their children influence the likelihood of being arrest ed in the same wave? b. Do the way men rate their relationship with their children influence the likelihood of being arrested in the subsequent wave? Residential Status 1. Residential Stability a. Among men, does the length of reside nce in the same place influence the likelihood of being arre sted in the same wave? b. Among men, does the length of reside nce in the same place influence the likelihood of being arrest ed in the subsequent wave? c. Among men, does the length of reside nce in the same place influence desistance from being arrested? Employment 1. Employment Status a. Does employment status influence men’s likelihood of being arrested in the same wave? b. Does employment status influence men’s likelihood of being arrested in the subsequent wave? c. Does employment status influence desistance from being arrested? 2. Job Satisfaction a. Does the level of satisfaction in one’s job influence men’s likelihood of being arrested during the same wave? b. Does the level of satisfaction in one’s job influence men’s likelihood of being arrested during the subsequent wave? c. Does the level of satisfaction in on e’s job influence desistance from being arrested?

PAGE 82

69 Finances 1. Income a. Does the annual level of men’s in come influence the likelihood of being arrested within the same wave? b. Does the annual level of men’s in come influence the likelihood of being arrested within the subsequent wave? c. Does the annual level of men’s in come influence desistance from being arrested? 2. Welfare Use a. Does welfare use among men influence the likelihood of being arrested during the same wave? b. Does welfare use among men influence the likelihood of being arrested during the subsequent wave? c. Does welfare use among men influence desistance from being arrested? Education 1. GPA a. Does the Grade Point Average of yo ung men predict being arrested in the subsequent wave? 2. High School Diploma a. Does the attainment of a high school diploma influence the likelihood of being arrested within the same wave? b. Does the attainment of a high school diploma influence the likelihood of being arrested within the subsequent wave? c. Does the attainment of a high school diploma influence desistance from being arrested? Prior Behavior 1. Bad Behavior a. Does bad behavior in Wave 1 influence the likelihood of being arrested in the subsequent wave? b. Does bad behavior in Wave 1 influence desistance from being arrested? 2. Arrests a. Do prior arrests influence the likeli hood of being arrested in the same wave? b. Do prior arrests influence the like lihood of being arrested in the subsequent wave? 3. Drug Use a. Does prior drug use influence the lik elihood of being arrested in the same wave?

PAGE 83

70 b. Does prior drug use influence the lik elihood of being arrested in the subsequent wave? Research Questions: Incarceration Age 1. Age of Participant: a. Among men, does the age of a partic ipant influence the likelihood of being incarcerated in the same wave? b. Among men, does the age of a partic ipant influence the likelihood of being incarcerated in the subsequent wave? c. Does age influence desistance from being incarcerated? Family: Marriage 1. Marital Status: a. Among men, does being married in fluence the likelihood of being incarcerated in the same wave? b. Among men, does being married in fluence the likelihood of being incarcerated in the subsequent wave? c. Does being married influence desi stance from being incarcerated? 2. Marital Satisfaction: a. Among married men, does the level of satisfaction in marital relationships influence th e likelihood of being incarcerated in the same wave? b. Among married men, does the level of satisfaction in marital relationships influence the likeli hood of being incarcerated in the subsequent wave? Family: Fatherhood 1. Fatherhood Status: a. Does fatherhood influence the likeli hood of being incarcerated in the same wave? b. Does fatherhood influence the likeli hood of being incarcerated in the subsequent wave? c. Does fatherhood influence desistan ce from being incarcerated? 2. Fatherhood Prior to the Age of 18: a. Among men, does fathering a child prio r to the age of 18 influence the likelihood of being incarcerated in the same wave? b. Among men, does fathering a child prio r to the age of 18 influence the likelihood of being incarcerated in the subsequent wave? c. Among men, does fathering a child prior to the age of 18 influence desistance from being incarcerated? 3. Residential Father:

PAGE 84

71 a. Among men, does residence with so me or all of their children influence the likelihood of being in carcerated in the same wave? b. Among men, does residence with so me or all of their children influence the likelihood of being inca rcerated in the subsequent wave? c. Among men, does residence with so me or all of their children influence desistance from being incarcerated? 4. Rate Self as Father: a. Does the way men rate themselves as fathers influence the likelihood of being incarcerated in the same wave? b. Does the way men rate themselves as fathers influence the likelihood of being incarcerated in the subsequent wave? 5. Feelings about their Relationship with their Child: a. Do the way men rate their relationship with their children influence the likelihood of being incarcerated in the same wave? b. Do the way men rate their relationship with their children influence the likelihood of being incarcerated in the subsequent wave? Residential Status 1. Residential Stability a. Among men, does the length of reside nce in the same place influence the likelihood of being incarcerated in the same wave? b. Among men, does the length of reside nce in the same place influence the likelihood of being incarcera ted in the subsequent wave? c. Among men, does the length of reside nce in the same place influence desistance from being incarcerated? Employment 1. Employment Status a. Does employment status influence men’s likelihood of being incarcerated in the same wave? b. Does employment status influence men’s likelihood of being incarcerated in the subsequent wave? c. Does employment status influence desistance from being incarcerated? 2. Job Satisfaction a. Does the level of satisfaction in one’s job influence men’s likelihood of being incarcerated during the same wave? b. Does the level of satisfaction in one’s job influence men’s likelihood of being incarcerated during the same wave? c. Does the level of satisfaction in on e’s job influence desistance from being incarcerated? Finances 1. Income

PAGE 85

72 a. Does the annual level of men’s in come influence the likelihood of being incarcerated within the same wave? b. Does the annual level of men’s in come influence the likelihood of being incarcerated within the subsequent wave? c. Does the annual level of men’s in come influence desistance from being incarcerated? 2. Welfare Use a. Does welfare use among men influence the likelihood of being incarcerated during the same wave? b. Does welfare use among men influence the likelihood of being incarcerated during th e subsequent wave? c. Does welfare use among men influence desistance from being incarcerated? Education 1. GPA a. Does the Grade Point Averag e of young men predict being incarcerated in the subsequent wave? 2. Highest Degree Earned a. Does the attainment of a high school diploma influence the likelihood of being incarcerated within the same wave? b. Does the attainment of a high school diploma influence the likelihood of being incarcerated within the subsequent wave? c. Does the attainment of a high school diploma influence desistance from being incarcerated? Prior Behavior 1. Bad Behavior a. Does bad behavior in Wave 1 influence the likelihood of being incarcerated in the subsequent wave? b. Does bad behavior in Wave 1 in fluence desistance from being incarcerated? 2. Arrests a. Do prior arrests influence the like lihood of being incarcerated in the same wave? b. Do prior arrests influence the like lihood of being incarcerated in the subsequent wave? 3. Drug Use a. Does prior drug use influence the likelihood of being incarcerated in the same wave? b. Does prior drug use influence the likelihood of being incarcerated in the subsequent wave?

PAGE 86

73 Analysis Overall, most of the research questio ns listed above are unified by one common theme: How do social bonds influence criminal behavior? The core goal of this project is to examine the ways that disparate social bonds influence criminal/a nti-social behavior throughout the life course. Or more speci fically: Do adult so cial bonds influence participation in a range of behaviors that ar e considered deviant or criminal? This was the key question that this resear ch sought to address. Partic ipation was operationalized as a dichotomous choice, wherein either the memb ers of the sample either did or did not engage in behaviors that either are criminal (drug use) or indica te criminal behavior (arrests or incarceration). Therefore, logist ic regression was the an alytical tool best suited to investigate criminal participation within this research. The methodological plan was to conduct a se ries of logistic regression analyses on a number of variables. There we re three types of models. First, a series cross-sectional and lagged models were utilized to examine if the variables of interest predicted nonparticipation in criminal activities within th e same wave and across subsequent waves. The cross-sectional models looked at criminal behavior within the same wave. As too few participants engaged in any of the conceptu alizations of criminal behavior in Wave 1, there were not any cross-sectional models predicting criminal be havior in Wave 1. Second, the lagged models displayed the manner in which the independe nt variables from the prior wave predicted criminal acts in th e following wave (See Table 4-2). Since there were no measures of arrests during Wave 4, two sets of models (Wave 3 Independent Variables predicting Wave 4 Dependent Vari ables and Wave 4 Independent Variables predicting Wave 4 Dependent Variables) on ly examined drug use, and incarceration.

PAGE 87

74 All of the dependent variables measuring arrests and incarcer ation corresponded to specific years (in the cross-sec tional and lagged models drug use is simply a measure of use within 30 day prior to the interview). As a result, a specific set of years were used for each model that predicts arrests and incarcera tion. For example, when utilizing the Wave 1 independent variables to predict Wave 2 de pendent variables the years of Wave 1 and up to the beginning of Wave 2 were used to se t the range for measures of these dependent variables. Later when Wave 2 independent variables were utilized to predict Wave 2 dependent variables, the year s within which the Wave 2 in terviews were conducted were used to set the boundaries for the years that encompassed the dependent variables. Table 4-2 displays the various years that each model’s dependent va riables represent. The use of specific years also framed the desistan ce models. These parameters are described below. The final set of models examined if the independent variables within various waves predicted desistance from criminal behavior s. There were separate models for each conceptualization of criminal behavior (drug use, arrests, and incarceration). Desistance following criminal behavior in Waves 2 and 3 was examined in two distinct sets of models. These sets of models were referred to as “Desist 2” and “Desist 3.” Desist 2 models predicted desistance from each of the three dependent variables beginning in 1983 (the first year of data collection for Wave 3). All of the part icipants included in each of the models engaged in one of the th ree criminal acts (drug use, arrests, and incarceration) prior to Wave 3 (1968-1982). There were three sets of Desist 2 models for each of the dependent variables. Desistance for different amounts of time wa s predicted. The independent variables from

PAGE 88

75 Table 4-2 Arrest and Incarceration Year s for Cross-sectional and Lagged Models MODELS YEARS DEPENDENT VARIABLES Wave 1 Independent Variables predicting Wave 2 Dependent Variables 1968-74 Arrests and Incarceration Wave 2 Independent Variables predicting Wave 2 Dependent Variables 1975-76 Arrests and Incarceration Wave 2 Independent Variables predicting Wave 3 Dependent Variables 1977-82 Arrests and Incarceration Wave 3 Independent Variables predicting Wave 3 Dependent Variables 1983-84 Arrests and Incarceration Wave 3 Independent Variables predicting Wave 4 Dependent Variables 1985-88 Incarceration Wave 4 Independent Variables predicting Wave 4 Dependent Variables 1989-90 Incarceration Wave 4 Independent Variables predicting Wave 5 Dependent Variables 1991-92 Arrests and Incarceration Wave 5 Independent Variables predicting Wave 5 Dependent Variables 1993-94 Arrests and Incarceration Waves 2 and 3 were used to predict desistan ce in all three models. First, desistance through Wave 3 was predicted (1983 through 1988). Second, desistance through Wave 4 was predicted (1983 through 1992). The final set of Desist 2 models predicted desistance through Wave 5 (1983 through 1994). These sets of years were used for both the models predicting desistance from drug use and incar ceration. Since there were no measures of arrests during Wave 4, the sec ond set of models (Desistanc e through Wave 4) were not run for arrests. Additionall y, the years available to pred ict desistance through Wave 5 were limited. Table 4-3 displays the years that are of interest in each model, as well as the sets of independent variables that were used to predict desistance.

PAGE 89

76 Table 4-3 Years for Desist 2 Models MODELS CRIME YEARS DESIST DRUG USE YEARS DESIST ARREST YEARS DESIST INCARCERATION YEARS INDEPENDENT VARIABLES Criminal Wave 2/ Desist Wave 3 1968-82 1983-88 1983-84 1983-88 WAVES 2 and 3 Criminal Wave 2/ Desist Waves 3-4 1968-82 1983-92 N/A 1983-92 WAVES 2 and 3 Criminal Wave 2/ Desist Waves 3-5 1968-82 1983-94 1983-84/ 1991-94 1983-94 WAVES 2 and 3 The final set of desistance models, Desi st 3, examined desistance beginning in Wave 4 (1989). There were two groups of De sist 3 models. The first group examined desistance through Wave 4 (1989 through 1992). Independent variables collected in Waves 3, and 4 were used to predict desistance in all Desist 3 models. However, as these was no measure of arrests at Wave 4 there we re only models predicting desistance from drug use and from incarceration. The next group predicted desi stance through Wave 5 (1989 through 1994). Table 4-4 displays the years that are of interest in each model, as well as the sets of independent variables that were used to predict desistance. Table 4-4 Years for Desist 3 Models MODELS CRIME YEARS DESIST DRUG USE YEARS DESIST ARREST YEARS DESIST INCARCERATION YEARS INDEPENDENT VARIABLES Criminal Wave 3/ Desist Wave 4 1968-88 1989-92 N/A 1989-92 Waves 3 and 4 Criminal Wave 3/ Desist Waves 4-5 1968-88 1989-94 91-94 1989-94 Waves 3 and 4

PAGE 90

77 The results are presented in three chapte rs which discuss each of the dependent variable separately. These models are presented in the following order within each chapter: Cross Sectional and Lagged Models 1. Wave 1 Independent Variables Pred icting Wave 2 Dependent Variables 2. Wave 2 Independent Variables Pred icting Wave 2 Dependent Variables 3. Wave 2 Independent Variables Pred icting Wave 3 Dependent Variables 4. Wave 3 Independent Variables Pred icting Wave 3 Dependent Variables 5. Wave 3 Independent Variables Pred icting Wave 4 Dependent Variables 6. Wave 4 Independent Variables Pred icting Wave 4 Dependent Variables 7. Wave 4 Independent Variables Pred icting Wave 5 Dependent Variables 8. Wave 5 Independent Variables Pred icting Wave 5 Dependent Variables Desistance Models 1. Wave 2 Independent Variables Pr edicting Desistance in Wave 3 2. Wave 3 Independent Variables Pr edicting Desistance in Wave 3 3. Wave 2 Independent Variables Pred icting Desistance in Waves 3-4 4. Wave 3 Independent Variables Pred icting Desistance in Waves 3-4 5. Wave 2 Independent Variables Pred icting Desistance in Waves 3-5 6. Wave 3 Independent Variables Pred icting Desistance in Waves 3-5 7. Wave 3 Independent Variables Pr edicting Desistance in Wave 4 8. Wave 4 Independent Variables Pr edicting Desistance in Wave 4 9. Wave 3 Independent Variables Pred icting Desistance in Waves 4-5 10. Wave 4 Independent Variables Pred icting Desistance in Waves 4-5

PAGE 91

78 Model and Variable Descriptions All of the types of models listed above involved a direct logistic regression, where theoretically significant variables were used to predict either criminal behavior or desistance from criminal behavior. Criminal behavior was the outco me variable within the cross sectional and lagged models. All other models s ought to predict desistance. When possible, the same predictors were used at each wave across both types of models. Multicolinnearity was examined across all of the variables utilized. Multicollinearity was assessed by examination of the Variance Inflation Factors (VIF). VIF values approaching 10 are cause for concern and require further examination (Tabachnick and Fidell, 2001). Within the va riables utilized, the highest VIF was less than 3, suggesting the absence of mulitcollinearity within the regressions included. However, mulitcollinearity did influence th e variables utilized within a number of models. A series of variables were examined for inclusion within the models in this research. Of particular inte rest were the variables noti ng prior deviance or criminal behavior. With measures of juvenile de linquency, drug use, arre sts, and incarceration available, all four served as theoretically significant predictors of future criminal behavior. When drug use, arre sts, and incarceration were in cluded together as predictors within the cross sectional a nd lagged models, prior arrests and/or prior incarceration obtained VIF scores above 10 across a number of models. Therefore, all of the cross sectional and lagged models were run including either arrest s or incarceration (drug use was included in all of these models). The Prior Arrests predictor was significant across more models, and resulted in a higher numbe r of models emerging as significant (Within the Omnibus Test of Model Coefficients). As a result, all of the cross sectional and lagged models (where measures were availabl e) included prior arrests and prior drug use

PAGE 92

79 as predictors. However, the three models where prior incarceration emerged as significant were included in Appendix A and were included in the discussion of findings within Chapter 8. Wave 1 independent variables predic ting wave 2 dependent variables The first independent variable that was included was age. This is a continuous variable that gave the age of the participant during the interview. As the sample was comprised of minors, the next independent variable, education, was highly relevant. There were two measures of education. Fi rst, whether or not each participant was currently enrolled in an educational instit ution was conceptualized as a dichotomous variable (1 = yes, 0 = no). The original va riable asked participants to describe their “current status.” The categories for this va riable were: 1 = school; 2 = work; 3 = school and work; 4 = school and looking for work; 5 = home. Those who were in categories 1, 3, and 4 were coded as ,” all others will be coded as .” Second, the quality or strength of this bond was concep tualized as the part icipant’s grade point average. This measure of achievement demonstrated the leve l of success that each participant displayed in holding the social bond of educational achievement. The next set of variables examined the ec onomic circumstances of the participants. First, the annual income of the participants ’ families was noted categorically: 1 = Less than $600; 2 = $600-999; 3 = $1,000-$1,999; 4 = $2,000-$2,999; 5 = $3,000-$3,999; 6 = $4,000-$4,999; 7 = $5,000-$5,999; 8 = $6,000-$6,999; 9 = $7,000-$7,999; 10 = $8,000$8,999; 11 = $9,000-$9,999; 12 = $10,000 or Greater. The other financial concern that was noted was whether or not the participant’ s family received some form of welfare. This variable is dichotomous, where those th at got welfare were coded as ,” and all others were coded as .” Th e original variable allowed the subjects to note that they

PAGE 93

80 either did not receive any welfare (0), that th ey received ADC (1), or that they got other forms of welfare (2). Responses of zero re mained the same, while responses of 1 and 2 were noted as .” Residential and familial circumstances were the next set of variables. First, a measure of each participant’s residential st ability was utilized. The following values were assigned to the noted amount of time each participant had spent in their residence at Wave 1: 1 = Less than 1 year; 2 = 1 to less than 2 years; 3 = 3 to le ss than 3 years; 4 = 3 to less than 5 years; 5 = more than 5 years; and 6 = all my lif e. The other variable within this group noted whether or not there was a fa ther in the home. Th e original variable offers the following choices: 0 = no older male ; 1= father-stepfather; 2= other male guardian; 3= other older male; 4= older brothe r (+15); 5 =NA. This variable was recoded into a dummy variable where participants who lived with their fathers, stepfathers, or a male guardian was coded as .” The rema ining participants were coded as .” Finally, Wave 1 measured “bad behavior.” This acted as a control for prior behavior, or low self control (Gottfredson and Hirschi, 1990). The variable bad behavior was recoded from a variable that measured ba d health. Participants were asked if they had engaged in any of the following behavior s which were damaging to one’s health: 1= don't exercise; 2= don't get enough rest; 3= don't exert myself; 4= don't dress properly; 5= eat too much candy; 6 = eat wrong food; 7 = don 't eat right; 8= eat too much; 9 = skip meals; 10 = smoke too much; 11= drink t oo much; 12= use drugs; 13 = do dangerous things; 14= do things bad for eyes; 15= no ch eck ups; 18= bad personal health habits. Participants that were recorded as 10, 11, 12, or 13, were noted as engaging in bad

PAGE 94

81 behavior. This is a dichotomous variable where bad behavior will be coded as ,” and all others will be coded as .” There were three dependent variables that measured criminal behavior in Wave 2: Drug use, arrests, and incarceration. The m easurement of drug use in Wave 2 asked participants about the number of days ago they last used a va riety of different drugs. The drugs under consideration were : Marijuana, acid, cocaine, heroin, methamphetamines, uppers, downers, glue, and “other drugs.” Th is was a continuous variable. This was recoded into a dummy variable where participants, who used any of these drugs in the last month, or 30 days, were recoded as ,” while all others were coded as .” Table 4-5 Descriptive Statis tics for Wave 1 Variables Variable N MeanStd. Dev. Age 351 13.88 1.57 In School 351 0.96 0.20 Income 351 6.79 2.29 Welfare Use 351 0.47 0.50 GPA 351 1.84 0.65 Time at Residence 351 4.15 1.72 Father in the home 351 0.44 0.49 Prior bad behavior 351 0.19 0.39 As noted above, the measures of arrests and incarceration were an amalgamation of measures of these events for a series of years. In order to maintain temporal causality, these dependent variables looked at arrest s and incarceration dur ing the years 1968 to 1974. This will predict the influence of Wave 1 variables upon these men’s lives prior to Wave 2. This variable is dichotomous, in th at any participants that were arrested or incarcerated at any time between these years was coded as ,” while the remaining participants were coded as .” While the or iginal variable asked for the significance of

PAGE 95

82 the events, and distinguished between a singul ar or a multiple event, this research condensed these categories into a dichotomous “Yes/No” response. Wave 2 independent variables predic ting wave 2 dependent variables Many of the variables from Wave 1 were measured again in Wave 2. However, some of the measurements were changed, and new variables were added. The new variables were added reflect the new social space within the life-course where these men resided. For example, Wave 2 looked at the attainment of educational degrees, employment, job satisfaction, fatherhood, a nd marriage. As these young men were now between the ages of 18 and 23, these forms of social bonds were both more influential and more likely to be attained. Age remained a continuous variable. As this is the time in the life-course where desistance by age is predicted, this variable may be more influential than in Wave 1. Normally this is also a time when yo ung men are obtaining some form of academic achievement via the attainment of a degr ee. Therefore, both enrollment and the attainment of degrees were used as measures . First, the variable “current status” was used to examine whether or not the participan t was a student during th is wave. A look at the categories within the data revealed that a few of these choices displayed that the participant is a student: 0= jail; 1= working only; 2= going to school; 3= looking for work; 4= working and going to school; 5= looking for work and going to school; 6= army; 7= housewife; 8= stay at home; 9= not hing. Categories 2 a nd 4 were recoded as ,” while the other categories were recoded as .” The ne xt variable noted the most advanced degree that each participant had ga ined. This variable was noted as: 0= no degree; 1= high school; 2= AA, AS, AAS; 3= BA , BS; 4= other. This was recoded to

PAGE 96

83 note if the participant had earned a high school degree or greater. Those who were coded as ” remained coded as ,” while all others were coded as .” The original variable measuring “current status” was used to create a measure of current employment. The choices available in the data set were: 0= jail; 1= working only; 2= going to school; 3= looking for work; 4= working and going to school; 5= looking for work and going to school; 6= army ; 7= housewife; 8= stay at home; 9= nothing. Those that were coded as 1, 4, a nd 6 were recoded as ” for the new dummy measure of employment. All others were c oded as .” Another new measure included to attend to the significance of employment duri ng this stage of the life-course was that of job satisfaction. The original four point scale was: 1= very satisfied ; 2= fairly satisfied; 3= not very satisfied; 4= not satisfied at all. This variable was reverse coded so that: 1= not satisfied at all; 2= not ve ry satisfied; 3= fairly satis fied; 4= very satisfied. Income in Wave 2 was coded as follows: 0= no income; 1= Under 2000; 2= 2000 2999; 3= 3000 4999; 4= 5000 6999; 5= 7000 9999; 6= 10000 14999; 7= 1500019999; 8= 20000 or more. Welfare use was exam ined. The data set’s measure of welfare allowed for the following responses: 1 = A ll the time; 2 = On and off; 6 = Not on Welfare; 8 = DK; 9 = NA; 666 = Welfare not a source. This variable was recoded dichotomously where categories 1 and 2 were recoded as ,” while all remaining participants were coded as .” Residential stability continued as an impor tant variable, and was measured: 1= less than 1 yr; 2= 1 to less than 2 yrs; 3= 2 to less than 3 yrs; 4= 3 to less than 4 yrs; 5= 4 to less than 5 yrs; 6= more than 5 yrs. During Wave 2 some of the measures of familial bonds appeared for the first time. One of th e main variables of interest, fatherhood, was

PAGE 97

84 examined. The measure of fatherhood was made into a dummy variable where fathers are coded as ,” and all others will be coded as .” This came from a dummy variable in the data that asked if these men had any children. Given the influence of life-course theory on this project, it wa s appropriate to examine the so cial timing of the appropriation of social bonds such as fatherhood. Therefore, the age at which these men first fathered a child was included as a variab le. The “age at first child” variable was a continuous measure, but was recoded to note fatherhood prior to the age of 18. Men who were fathers at 17 or younger were c oded as ,” while all others were coded as .” Finally, marriage was also measured as a form of so cial bonding. This variable was included in the data as a dummy variable where those that were married are coded as ,” and those that were not married are coded as .” Finally, controls for prior behavior were included. The dependent variables for Wave 1 predicting Wave 2 arrests and drug use were used as predictor for all three forms of criminal behavior. All of the remaini ng cross-sectional and la gged models used the prior measure of drug use and arrests and independent variables. These dependent variables were coded just as the dependent va riables that were used to test the relationship between Wa ve 1 independent variables and Wave 2 dependent variables. However, the year s in which arrests and incarceration were examined differed in order to satisfy the time order criteria of causality. Here incarceration and arrests duri ng the years of 1975-1976 were of interest. Drug use was coded in the same manner as before. Men w ho used drugs within the 30 days prior to the interview were coded as ” (drug users), a ll others were coded as .” Table 4-6 displayed the descriptive statis tics for Wave 2 Variables.

PAGE 98

85 Table 4-6 Descriptive Statis tics for Wave 2 Variables Variable N MeanStd. Dev. Age 27720.46 1.44 In School 2770.32 0.47 Employed 2770.45 0.49 Income 2774.26 2.06 HS Degree 2770.55 0.49 Job Satisfaction 2772.84 1.02 Welfare Use 2770.10 0.31 Time at Residence 2774.26 2.06 Fatherhood 2770.27 0.45 Father Before Age 18 2770.09 0.29 Married 2770.04 0.19 Drug Use Wave 2 2770.60 0.49 Incarcerated (2 pred. 2) 2770.03 0.17 Arrested (2 pred. 2) 2770.06 0.23 Wave 2 independent variables predic ting wave 3 dependent variables The independent variables discussed above remained unchanged, with the exception of the controls for prior behavior. For these models the dependent variables of drug use and arrests from Wave 2 predicting Wave 2 were used as controls for prior behavior. The dependent variables from Wave 3 were created from th ree distinct sets of variables. The disparate measurements of deviant behavior include d drug use, arrests, and incarceration. The following drugs were in question: Marijauna, PCP, acid, cocaine, crack, heroin, methadone, uppers, qualuudes, tran quilizers, sedatives, inhalants, nitrates, talwins, codine, and morphine. Participan ts were asked how many days have passed since they last used each of the substances listed. Conti nued usage of these substances

PAGE 99

86 was defined as consumption in the past month, or 30 days. Those who used any of these drugs in the past 30 days were coded as ,” all others were coded as .” Data about arrests and incarceration were gathered through questions concerning experiences for each year from 1960 to 1984. Participants were asked if they were arrested during each of these years, and whethe r or not these events had a great amount of influence on their lives. The same question was asked about incarcera tion. Participants were placed in one of these ten categories: 0 = No event that year ; 1 = One event, little impact; 2 = One event, some impact; 3 = One event, great impact; 4 = Two or more events, little impact; 5 = Two or more even ts, some impact; 6 = Two or more events, great impact; 8 = One event, no impact noted; 9 = Two or more events, no impact noted. Participants coded as ” remained the same, wh ile all others were coded as .” After a dichotomous variable for each year had been created, a new dummy variable was created to examine experiences between the years of 1977 to 1982. Wave 3 independent variables predicti ng wave 3 dependent variables Unfortunately the variable “age” was not coll ected in Wave 3. However, a variable noting the date of the interview was used to calculate their change in age since it was obtained in Wave 2. Attainment of educa tion, measured via the last degree earned continued to be a variable of interest. However, the operationalization changed from Wave 2 to Wave 3. The new educational atta inment variable included more categories, and offered higher levels of education as potential choices. The categories available were: 0= No diploma; 1= HS diploma; 2= HS equivalency; 3= A ssociates degree; 4= Trade/Vocational/Tech Certif icate; 5= BA/BS; 6= MA/M S; 7= Doctorate; 8= Trade/Vocational/Tech Certificate. Men who were coded as ” will remained coded as ,” all others were coded as .”

PAGE 100

87 Employment and earnings were the core concerns of the s ubsequent sets of variables. First, the employment status of the participant was noted as a dichotomous variable where employed men were coded as ,” and all others were coded as .” As in previous waves, the variable ut ilized to test employment meas ured the current “status” of the participant. The initial set of responses available were listed as: 1= Working; 2= Going to school; 3= Looking for work; 4= In armed forces; 5= In jail; 6= Housewife/mother; 7= At home/poor health ; 8= Just hanging out; 9= Other; 12= Categories 1 and 2; 23= Categories 2 and 3. Participants th at were either working (1), in the armed force (4), or working and going to school (12), were coded as ,” with the remaining participants coded as .” The other variable that examined employment sought to measure the importance of the stre ngth of the social bond via job satisfaction, assuming that satisfying employment would provide a stronger social bond. This variable was not changed from its original coding: 1= Very Dissatisfied; 2= Little Dissatisfies; 3= Fairly satisfied; 4= Very satisfied. Earnings were examined through a meas urement of the annual income of each participant, as well as a dichotomous measur e of welfare participation. Annual income was measured categorically: 0= No in come; 1= Under $5000; 2= $5000-$7499; 3= $7500-$9999; 4= $9999-14999; 5= $15000-19999; 6= $20000 or more; 7= Refused to answer. The measure of welfare use utilized for this research aske d participants about their frequency of welfare use. Participants could answer: 1 = All the time; 2 = On and off; 6 = Not on Welfare; 8 = DK; 9 = NA; 666 = Welfare not a source. This variable was recoded into a dummy variable, where those wh o either used welfare all the time, or who have used it on and off were coded as ,” while all othe rs will be coded as .”

PAGE 101

88 Residential stability was measured in the same manner as it was in Wave 2. Participants were asked how long they have lived in their current address. Responses were coded as follows: 1= less than 1 yr; 2= 1 to less than 2 yrs; 3= 2 to less than 3 yrs; 4= 3 to less than 4 yrs; 5= 4 to less than5 yr s; 6= 5 or more years. The measurement of paternal status in Wave 3 asked participants how many children they had. This variable was recoded so that all men with one or more children were coded as .” Men who had not fathered any children were coded as .” The age at first child was tested as an influence in Wave 3, and was recoded as a di chotomous variable as in Wave 2. Men who were fathers at 17 or younger were coded as ,” all others were coded as .” Residential fatherhood will be a dummy variable. Participants were asked about the people that they lived with over a series of years (1976-1983). Here the years of Wave 3 data collection (1983-1984) were used to mark co-residence with one’s child. Men were asked how many of their own children lived with them. The range varied from zero to five children in residence. Response categ ory ” meant that they lived with their children but did not provide a number. This was recoded into a dummy variable where men who lived with any of thei r children during these years we re coded as ,” all others were coded as .” A variety of measures were collected that helped to measure the strength of the fatherhood bond. The first measure regard ing the fatherhood bond asked these men if they believed that they were good parents. Responses were coded as: 0= Parent missing; 1= not so good; 2= average; 3= very good. Here “Parent Missing” refers to a lack of any relationship with one’s child. The next meas ure asked if these men were happy with their relationship with their child. They could choose from a f our point scale: 1 = very

PAGE 102

89 dissatisfied; 2= dissatisfied; 3= satisfied; 4= ve ry satisfied. These variables were used in models that only examine criminal behavior among men who are fathers. Additionally, these variables were be included in the desistance models as there are not enough participants for split desistance models. Marital status allowed the following res ponses within the data set: 0= Never married; 1= Married in past; 2= Separated; 3= Single/NA past; 4= Marr ied; 5= Live with partner; 6= Live with partner, was married. This was recoded for this research as a dummy variable where only category 4 was r ecoded as 1. All other categories were recoded as 0. The quality of this bond was examined through the ut ilization of a measure of marital satisfaction. Satis faction and dissatisfaction with in the marriage were coded as: 1= very dissatisfied; 2= dissatisfied; 3= satisfied; 4= very satisfied. This same measure was used in this research. This m easure of marital satisfaction was used only in split models that examined criminal behavi or among married men. Just as with the fatherhood quality measures, this variable was be included in the desistance models as there were not enough particip ants to run these models. Once again, the dependent variables from th e prior set of models were used as predictors of criminal behavior in the current set of models. The dependent variables of arrests and drug use from Wave 3 predicting Wave 3 were used as predictors of all three new dependent variables. Measures of drug use, arrests, and incarce ration served as the dependent variables in Wave 3. The same measurement of dr ug use from the lagged model of Wave 2 independent variables predicting Wave 3 depend ent variables was used here. Participants who used any of the controlled substances in question within the past 30 days were coded

PAGE 103

90 as ,” while all other participants were code d as .” The measurements of arrests and incarceration also maintained th e same coding. However, the years in which these events occurred differed from the prior lagged mode l. Here behaviors in the years 1983 to 1984 were measured. Table 4-7 displayed the de scriptive statistics for Wave 3 Variables. Wave 3 independent variables predic ting wave 4 dependent variables The independent variables from Wave 3 whic h were used to test Wave 3 dependent variables remained the same, with the excep tion of the controls for prior criminal behavior. The Wave 3 predicting Wave 3 de pendent variables of drug use and arrests were used as independent variables. Add itionally this section of the project also contained split models predicting criminal be havior among married men and fathers as in the previous set of models. Table 4-7 Descriptive Statis tics for Wave 3 Variables Variable N MeanStd. Dev. Age 21028.29 1.53 Employed 2100.60 0.49 Income 2104.34 1.54 HS Degree 2101.30 1.48 Job Satisfaction 2102.94 0.98 Welfare Use 2100.15 0.36 Time at Residence 2104.30 1.91 Fatherhood 2100.52 0.50 Father Before Age 18 2100.09 0.29 Residential Father 2100.27 0.44 Married 2100.22 0.42 Drug Use Wave 3 2100.54 0.50 Incarcerated (3 pred. 3) 2100.06 0.24 Arrested (3 pred. 3) 2100.08 0.27

PAGE 104

91 The Wave 4 dependent variables were not as plentiful as the other waves as there was not a measurement of arrests. Additi onally the measurements of drug use and incarceration were different from the previous measurements. The drug use variable for Wave 4 asked participants the time in months si nce they last used or injected a variety of different drugs. The drugs in question we re: Marijuana, cocaine , crack, heroin, acid, angel dust, methadone, speedballs, uppers, se datives, glue, quaaludes, talwins, codine, morphine, or “other drugs.” Participants that used any of these drugs in the past month were coded as ,” while all remaining subjec ts were coded as .” Incarceration during Wave 4 was measured by a variable that as ked participants how many months they spent in jail or prison during the years 1985 to 1989. Participants that were incarcerated for any time during these years were coded as “ 1,” all others were coded as .” Wave 4 independent variables predic ting wave 4 dependent variables The variable measuring age remained con tinuous. Educational attainment was still measured via the last degree earned by the pa rticipants. The new educational attainment variable offered the following categories: 0= No diploma; 1= HS diploma; 2= HS equivalency; 3= Associates degree; 4= Trad e/Vocational/Tech Certificate; 5= BA/BS; 6= MA/MS; 7= Doctorate; 8 = Othe r. Men who attained a high school equivalency or a high school diploma or greater were recoded as ,” all others were coded as .” Employment was measured using the “current status variable.” The choices available were: 1= Working; 2= Going to school; 3= Lo oking for work; 4= In armed forces; 5= In jail; 6= Housewife/mother; 7= At home/poor health; 8= Just hanging out; 9= Other; 12= Work and school; 13 = School and looking for work; 19 = Work and other. Categories 1, 4, 12, and 19 were recoded as ,” while the re maining categories were recoded as .” The strength of this bond continued to be m easured via job satisfaction. The scale for

PAGE 105

92 this variable was as follows: 1= Very Dissatisfied; 2= Little Dissatisfies; 3= Fairly satisfied; 4= Very satisfied. Welfare usage was examined in Wave 4 in the same manner as it was in Wave 3. Frequency of welfare us e was divided into the following categories: 1 = All the time; 2 = On and off; 6 = Not on Welfare; 8 = DK; 9 = NA; 666 = Welfare not a source. Men who were coded as ” or ” were coded as ,” and the remaining participants were coded as .” Residential stability was utilized as in previous waves. The coding remained the same as in Waves 2 and 3. Participants were as ked in this wave if they had any children. This variable remained dichotomous as in the original coding, with fathers being coded as ,” and non-fathers being coded as .” Fa therhood prior to the ag e of 18 continued to be included as a dichotomous variable. Re sidential fatherhood was noted. This question asked the men how many of their own childre n they lived with over a sequence of years (1984-1990). This variable was recoded into a dummy variable so that men that lived with any of their children dur ing the years of Wave 4 data collection (1989-1990) were coded as ,” while the other men were coded as .” Marital status was measured as a dummy variable and was used in the same manner (1= married; 0 = Not Married). Marital satisfaction, and hence a measure of the strength of this bond, was not available in Wave 4. Prior measures of arrests and drug use were used as controls for previous behavior. The Wave 3 predicting Wave 4 measure of drug use was utilized. However, as there was no Wave 4 measure of arrests, the Wave 3 pr edicting 3 measure of arrest was used as a control for prior arrests.

PAGE 106

93 The dependent variables in were Wave 4 measures drug use and incarceration. Participants that had used any of the drugs in question in the past month were coded as ,” while all others were coded as .” Incarceration during Wave 4 was measured by a variable that asked pa rticipants how many months they spent in jail or prison during the years 1985 to 1989. Participants that were in carcerated for any time during these years were coded as ,” all others were coded as .” Here incarceration from 1989-1990 was utilized as a dependent variable. Table 4-8 displayed the descriptive statistics for Wave 4 Variables: Table 4-8 Descriptive Statis tics for Wave 4 Variables Variable N MeanStd. Dev. Age 17734.33 1.56 Employed 1770.69 0.46 Income 1774.94 1.56 HS Degree 1770.79 0.41 Job Satisfaction 1773.15 0.85 Welfare Use 1770.11 0.31 Time at Residence 1774.49 1.94 Fatherhood 1770.68 0.47 Father Before Age 18 1770.10 0.30 Residential Father 1770.33 0.47 Married 1770.42 0.49 Drug Use Wave 4 1770.19 0.40 Incarcerated (4 pred. 4) 1770.11 0.31 Wave 4 independent variables predic ting wave 5 dependent variables The independent variables used in the cross sectional analysis of Wave 4 were used along with measures of prior behavior to pred ict behavior in Wave 5. Here the measure

PAGE 107

94 of arrests from Wave 3 predicting 3 was used once again. The Wave 4 predicting Wave 4 measure of drug use was used as a control for prior drug use. The dependent variables for Wave 5 consiste d of measures of drug use, arrests in the past 2 years, and incarcera tion during the past 2 years. Participants were asked about a variety of drugs. The variab le that was used here asked about the last time they had consumed a specific drug. They were aske d about marijuana, acid, coke, crack, heroin, speedball, methadone, uppers, tranquilizers, in halants, quaaludes, ta lwins, morphine, and “anything else.” All of th e questions, except for marijuana, had the same response options. Most of these drug questions offered the following choices in regards to the last time the drug was used: 1 = Today; 2 = Days ago; 3 = Weeks ago; 4 = Months ago; 5 = Years ago; 7 = Other; 8 = DK; 9 = NA. The choices available regarding the last use of marijuana were: 0 = Today; 1 = Yesterday; 2 = 2 days to one week ; 3 = 2 weeks to one month; 4 = 1 month to 3 months ; 5 = 4 months to 1 year; 6 = 1 to 2 years; 7 = 3 to 4 years; 8 = 5 to 10 years; 9 = 11 to 15 years; 10 = 16 to 20 year s; 11 = More than 20 years; 995 = Don’t smoke marijuana. For this resear ch “non-participation” was conceptualized as refraining from using any drug for over one month. Therefore, those who were coded as ,” ,” ,” and ” for any of the drugs in question were recoded as ” for recent drug users. All others were coded as .” In Wave 5, participants were asked if they had been arre sted of incarcerated in the past 2 years. Those who were not arrest ed were coded as ,” and men who were arrested in the previous 2 y ears were coded as .” The same coding was utilized when these men were questioned about being incarcer ated. This same coding was utilized in this research.

PAGE 108

95 Wave 5 independent variables predic ting wave 5 dependent variables Age was once again a continuous variable. Educational attainment was measured with the last degree the partic ipant had earned. The responses were coded as follows: 0= No diploma; 1= HS diploma; 2= HS e quivalency; 3= Associates degree; 4= Trade/Vocational/Tech Certificat e; 5= BA/BS; 6= MA/MS; 7= Doctorate; 8= Other; 9= NA. Men who attained a high school equiva lency or a high school diploma or greater were recoded as ,” all others were coded as .” The ne xt variable revealed if the participant was employed in Wave 5. When asked about their current status, the responses of participants were placed into the following categories: 1= Working; 2= Going to school; 3= Looking for work; 4= In armed forces; 5= In jail; 6= Housewife/mother; 7= At home/poor health ; 8= Just hanging out; 9 = NA. As the employment variable for this research is dichotomous, responses of ” and ” were recoded as ,” while all others were recoded as .” Job satisfaction was also measured. However, this variable was reverse coded. The initial coding was listed as: 1 = Very Satisfied; 2 = Fairly Satisfied; 3 = A Little Dissatisfied; 4 = Very Dissatisfied. This was recoded so that: 1 = Not satisfied at all; 2 = Not very satisfied; 3 = Fairly satisfied; 4 = Very Satisfied. The annual income of each participant was placed into the following categories: 0 = No income; 1 = Under $5000; 2 = $5 000-$7449; 3 = $7500-$9999; 4 = $10000-$14999; 5 = 15000-19999; 6 = $20000 or more; 7 = Refused; 8 = DK; 9 = NA. The other economic variable asked particip ants about the frequency of we lfare usage. Participants were coded as: 1 = Get it all the time; 2 = Ge t it on and off; 7 = Refused; 8 = DK; 9 = NA. This variable was recoded into a dichotom ous variable that disp layed if participants reported using welfare at all. Participants w ho were previously coded as ” or ” were

PAGE 109

96 recoded as .” All others were recoded as .” The sample was once again asked about the length of time in their current residence. Time in their current address was coded as: 1= less than 1 yr; 2= 1 to less th an 2 yrs; 3= 2 to less than 3 yrs; 4= 3 to less than 4 yrs; 5= 4 to less than5 yrs; 6= 5 or more years. This variable retained the same coding. The next set of variables displayed the various manners in which these men held family bonds. First, these men were asked if they were fathers. Here ” noted fatherhood and ” noted that they were not fathers. This variable underwent a minor amount of recoding in that res ponses of ” were changed to ” in order to create a dummy variable. Residential fatherhood wa s catalogued through a variable which asked the men how many of their children they live d with over a series of years. For the purposes of this research, resi dence with any of th eir children during th e years of Wave 5 data collection (1993-1994) was used to provi de evidence of “current” co-residence with one’s child. This was recoded from a conti nuous variable into a dummy variable where ” represented co-residence, and ” revealed that they did not live with any of their children. Marital status was represented as a dichotomous variable where married men were coded as ” and non-married men were coded as .” The measure of marital satisfaction was reverse coded from its original values of: 1 = Very Good; 2 = Pretty Good; 3 = So/So; 4 = Pretty Bad; 5 = Very Bad. The new coding was: 1 = Very Bad; 2 = Pretty Bad; 3 = So/So; 4 = Pretty Good; 5 = Very Good. This measure of marital satisfaction was once again used in separate models that pred icted drug use, arrests, and incarceration among married men. Finally, the dependent variable s used here were the same variables used in the prior lagged model. Drug use, arrests, and incar ceration remained as dichotomous variables

PAGE 110

97 where participation was coded as ,” wh ile those who did not participate in or experience any of these events was coded as .” Drug use in the past month was conceptualized as participation. Likewise, be ing arrested or incarcerated in the past 2 years was conceptualized as pa rticipation. Table 4-9 displaye d the descriptive statistics for Wave 5 Variables. Table 4-9 Descriptive Statis tics for Wave 5 Variables Variable N MeanStd. Dev. Age 16837.65 1.50 Employed 1680.63 0.49 Income 1685.00 1.85 HS Degree 1680.83 0.38 Job Satisfaction 1683.14 0.87 Welfare Use 1680.27 0.44 Time at Residence 1684.58 1.88 Fatherhood 1680.72 0.45 Residential Father 1680.32 0.47 Married 1680.41 0.49 Drug Use Wave 5 1680.29 0.46 Arrested Wave 5 1680.18 0.38 Incarcerated (5 pred. 5) 1680.17 0.38 Desistance The social bonds held by participants in each wave were utilized to predict desistance from criminal behavior. This was accomplished through the creation of a series of dummy variables. As mentioned a bove, a series of models (Desist-2 and Desist3) predicted desistance from all three criminal behaviors over a myriad of different years (See Tables 4-3 and 4-4). These models were limited to participants who engaged in criminal behavior earlier in the study. Desist2 included men who were criminal

PAGE 111

98 between 1968 and 1982. The Desist-3 models included men who were criminal between 1968 and 1988. For each of the models desistance was defined as a cessation of the criminal activity of interest through the years of interest in that model. Men who desisted were coded as ,” while men who engaged in deviant behavior were coded as .” Deviance was conceptualized as any type of participation in the aforementioned categories. Therefore, any use of drugs dur ing that time period was considered as nondesistance. This differed from the crosssectional and lagged models where recent drug use (any drug use within the past 30 days) was examined. The same held true for arrests and incarceration, where experiencing eith er of these events during the waves under consideration was conceptualized as non-desist ance. However, the models that predicted desistance via arrests had to deal with missi ng data in Wave 4. As there was no measure of arrests for Wave 4, these models could onl y examine arrests in Waves 3 and/or 5 (See Tables 4-3 and 4-4). Measures of arrests used the years in which arrests were recorded 1968-1984) as well as a measure from Wave 5 wh ich asked participants about arrests in the past 2 years. As Wave 5 interviews were conducted in 1993 and 1994, the measure of arrests spanned from 1991 though 1994. The in dependent variables used in the crosssectional and lagged models re tained the same coding. As the sample was limited to men who had engaged in prior criminal behavior, th e measures of previous drug use, arrests, and incarceration were not included. These c ontrols would have been redundant as the sample was composed entirely of offenders. Therefore, the measure of Wave 1 bad behavior was included in all models in orde r to provide a control for prior anti-social behavior. The variables used to pr edict desistance were as follows:

PAGE 112

99 Wave 2: Age, enrolled in school, highest degree earned, employed, job satisfaction, annual income, welfare use, Wave 1 bad behavior, residentia l stability, fatherhood status, age at birth of firs t child, and marital status Wave 3: Age, highest degree earned, employed, job satisfaction, annual income, welfare use, Wave 1 bad be havior, residential stabilit y, fatherhood status, age at birth of first child, residentia l fatherhood, and marital status Wave 4: Age, highest degree earned, employed, job satisfaction, annual income, welfare use, Wave 1 bad be havior, residential stabilit y, fatherhood status, age at birth of first child, residential fatherhood, , and marital status Wave 5: Age, highest degree earned, employed, job satisfaction, annual income, welfare use, Wave 1 bad behavior, re sidential stability, fatherhood status, residential fatherhood, and marital status Hypotheses The following hypotheses discuss the cross-se ctional and lagged models of deviant behaviors, as well as the m odels of desistance. These hypotheses are organized and discussed by each dependent variable. Each hyp othesis is listed along wi th literature that supports the hypothesis. The hypotheses refe rred to all of the waves in the study. Hypotheses: Drug Use Age 1. Age of Participant: a. Younger participants are expected to be more likely to use drugs in the same wave. ( Hirschi and Gottf redson, 1983; Sampson and Laub, 1992; White and Bates, 1995) b. Younger men are expected to be more likely to use drugs in the subsequent wave. ( Hirschi and Gottfredson, 1983; Sampson and Laub, 1992; White and Bates, 1995) c. Younger men are less likely to desist from using drugs. ( Hirschi and Gottfredson, 1983; Sampson and Laub, 1992; White and Bates, 1995) Family: Marriage 1. Marital Status: a. Married men are hypothesized to be less likely to use drugs in the same wave. (Farrington and We st, 1995; Rand, 1987; Sampson and Laub, 1990; 1993; Trassler, 1979; Warr, 1998) b. Married men are hypothesized to be less likely to use drugs in the subsequent wave. (Chen and Kandel, 1998; Ebbsen and Elliott, 1994; Farrington and West, 1995; La bouvie, 1996; Laub, Nagin, and

PAGE 113

100 Sampson, 1998; Osborn and West , 1979; Rand, 1987; Sampson and Laub, 1990; 1993; Trassler, 1979; Wa rr, 1998; West, 1982; White and Bates, 1995; Yamaguchi and Kandel, 1985) c. Married men are hypothesized to be mo re likely desist from drug use. (Chen and Kandel, 1998; Ebbsen and Elliott, 1994; Farrington and West, 1995; Labouvie, 1996; La ub, Nagin, and Sampson, 1998; Osborn and West, 1979; Rand, 1987; Sampson and Laub, 1990; 1993; Trassler, 1979; Warr, 1998; West, 1982; White and Bates, 1995; Yamaguchi and Kandel, 1985) 2. Marital Satisfaction: a. Among married men, those who are in marriages that are rated more favorably are hypothesized to be less likely to use drugs in the same wave. (Laub, Nagin, and Sampson, 1998; Laub and Sampson, 1993) b. Among married men, those who are in marriages that are rated more favorably are hypothesized to be le ss likely to use drugs in the subsequent wave. (Laub, Na gin, and Sampson, 1998; Laub and Sampson, 1993) Family: Fatherhood 1. Fatherhood Status: a. Fathers are predicted to be less likely to use drugs in the same wave. (Ebbsen and Elliott, 1994; Farrington and Hawkins, 1991; Farrington and West, 1995; Hughes, 1998; Kn ight, Osborn, and West, 1977; Labouvie, 1996; Sampson and Laub, 1990; Trassler, 1979; White and Bates, 1995; Yamaguchi and Kandel, 1985) b. Fathers are predicted to be less likel y to use drugs in the subsequent wave. (Ebbsen and Elliott, 1994; Farrington and Hawkins, 1991; Farrington and West, 1995; Hughes, 1998; Knight, Osborn, and West, 1977; Labouvie, 1996; Sampson a nd Laub, 1990; Trassler, 1979; White and Bates, 1995; Yamaguchi and Kandel, 1985) c. Fathers are predicted to be more likely to desist from using drugs. (Ebbsen and Elliott, 1994; Farrington and Hawkins, 1991; Farrington and West, 1995; Hughes, 1998; Kn ight, Osborn, and West, 1977; Labouvie, 1996; Sampson and Laub, 1990; Trassler, 1979; White and Bates, 1995; Yamaguchi and Kandel, 1985) 2. Fatherhood Prior to the Age of 18: a. Men that father children before the age of 18 are hypothesized to be more likely to use drugs in the same wave. (Christmon and Luckey, 1994; Dearden et al., 1995; Elster et al., 1987; Ketterlinius et al., 1992; Springarn and DuRant, 1996; Stoutheimer-Loeber and Wei, 1998) b. Men that father children before the age of 18 are hypothesized to be more likely to use drugs in the subsequent wave. (Christmon and Luckey, 1994; Dearden et al., 1995; El ster et al., 1987; Ketterlinius et al., 1992; Springarn and DuRant, 1996; Stoutheimer-Loeber and Wei, 1998)

PAGE 114

101 c. Men that father children before the age of 18 are hypothesized to be less likely to desist from usi ng drugs. (Christmon and Luckey, 1994; Dearden et al., 1995; Elst er et al., 1987; Kette rlinius et al., 1992; Springarn and DuRant, 1996; Stoutheimer-Loeber and Wei, 1998) 3. Residential Father: a. Among men, residence with some or al l of their children is expected to predict a lower likeliho od of drug use in the sa me wave. (Farrington and West, 1995) b. Among men, residence with some or al l of their children is expected to predict a lower likelihood of dr ug use in the subsequent wave. (Farrington and West, 1995) c. Among men, residence with some or al l of their children is expected to predict a higher likelihood of desi stance from drug use. (Farrington and West, 1995) 4. Rate Self as Father: a. Men who rate themselves more favorably as fathers are predicted to be less likely to use drugs in the sa me wave. (Farrington and Hawkins, 1991; Farrington and West, 1995; Maruna, 2001) b. Men who rate themselves more favorably as fathers are predicted to be less likely to use drugs in the subsequent wave. (Farrington and Hawkins, 1991; Farrington and West, 1995; Maruna, 2001) 5. Feelings about their Relationship with their Child: a. Men who rate their relationships with their children more favorably are predicted to be less likely to use dr ugs in the same wave. (Farrington and Hawkins, 1991; Farrington and West, 1995; Maruna, 2001) b. Men who rate their relationships with their children more favorably are predicted to be less likely to us e drugs in the subsequent wave. (Farrington and Hawkins, 1991; Fa rrington and West, 1995; Maruna, 2001) Residential status 1. Residential Stability a. Men who maintain the same residen ce for a longer period of time are hypothesized to be less likely to us e drugs in the same wave. (Laub and Sampson, 1988; Sampson and Laub, 1990) b. Men who maintain the same residen ce for a longer period of time are hypothesized to be less likely to us e drugs in the subsequent wave. (Laub and Sampson, 1988; Sampson and Laub, 1990) c. Men who maintain the same residen ce for a longer period of time are hypothesized to be more likely to desist from using drugs. (Laub and Sampson, 1988; Sampson and Laub, 1990)

PAGE 115

102 Employment 1. Employment Status a. Men who are employed are predicted to be less likely to use drugs in the same wave. (Farrington et al., 1986; Sampson and Laub, 1990; 1993; Waldorf, Reinarman, and Murphy, 1991) b. Men who are employed are predicted to be less likely to use drugs in the subsequent wave. (Farringt on et al., 1986; Sampson and Laub, 1990; 1993; Waldorf, Rein arman, and Murphy, 1991) c. Men who are employed are predicted to be more likely to desist from using drugs. (Farrington et al., 1986; Sampson and Laub, 1990; 1993; Waldorf, Reinarman, and Murphy, 1991) 2. Job Satisfaction a. Men with a higher level of job satis faction are predicted to be less likely to use drugs during the same wave. (Irwin, 1970; Laub and Sampson, 1993; Shover, 1985; 1996) b. Men with a higher level of job satis faction are predicted to be less likely to use drugs during the subsequent wave. (Irwin, 1970; Laub and Sampson, 1993; Shover, 1985; 1996) c. Men with a higher level of job satis faction are predicted to be more likely to desist from using dr ugs. (Irwin, 1970; Laub and Sampson, 1993; Shover, 1985; 1996) Finances 1. Income a. Men with higher incomes are hypothesi zed to be less likely to use drugs during the same wave. (Pezz in, 1995; Trassler, 1979; Waldorf, Reinarman, and Murphy, 1991) b. Men with higher incomes are hypothesi zed to be less likely to use drugs during the subsequent wave. (Pezzin, 1995; Trassler, 1979; Waldorf, Reinarman, and Murphy, 1991) c. Men with higher incomes are hypothesize d to be more likely to desist from using drugs. (Pezzin, 1995; Tr assler, 1979; Waldorf, Reinarman, and Murphy, 1991) 2. Welfare Use a. Men who use welfare are predicted to be more likely to use drugs in the same wave. (Laub and Sampson, 1988) b. Men who use welfare are predicted to be more likely to use drugs in the subsequent wave. (Laub and Sampson, 1988) c. Men who use welfare are predicted to be less likely to desist from using drugs. (Laub and Sampson, 1988)

PAGE 116

103 Education 1. GPA a. Men who held a higher Grade Point Average in Wave 1 are predicted to be less likely to use drugs in the subsequent wave. (Glueck and Glueck, 1968) 2. High School Degree a. Men with a high school degree are hypot hesized to be less likely to use drugs within the same wave. (Glueck and Glueck, 1968; Horney, Osgood, and Marshall, 1993; Rand, 1987; Sampson and Laub, 1990; 1993) b. Men with a high school degree are hypot hesized to be less likely to use drugs within the subsequent wave. (Glueck and Glueck, 1968; Horney, Osgood, and Marshall, 1993; Rand, 1987; Sampson and Laub, 1990; 1993) c. Men with a high school degree are hypot hesized to be more likely to desist from using drugs. (Glueck and Glueck, 1968; Horney, Osgood, and Marshall, 1993; Rand, 1987; Sampson and Laub, 1990; 1993) Prior Behavior: 1. Bad Behavior a. Men who engaged in bad behavior in Wave 1 are predicted to be more likely to use drugs in the subseque nt wave. (Blumstein et al., 1985; Nagin and Paternoster, 1991; 2000) b. Men who engaged in bad behavior in Wave 1 are predicted to be less likely to desist from using drugs. (Blumstein et al., 1985; Nagin and Paternoster, 1991; 2000) 2. Drug Use a. Men who previously used drugs are pr edicted to be more likely to use drugs in the same wave. (Horney et al., 1993; Nagin and Paternoster, 1991; 2000) b. Men who previously used drugs are pr edicted to be more likely to use drugs in subsequent waves. (Horney et al., 1993; Nagin and Paternoster, 1991; 2000) 3. Arrests a. Men who were previously arrested ar e predicted to be more likely to use drugs in the same waves. (B lumstein et al., 1985; Nagin and Paternoster, 1991; 2000) b. Men who were previously arrested ar e predicted to be more likely to use drugs in subsequent waves. (B lumstein et al., 1985; Nagin and Paternoster, 1991; 2000)

PAGE 117

104 Hypotheses: Arrests Age 1. Age of Participant: a. Younger participants are expected to be more likely to be arrested in the same wave. ( Hirschi and Gottfredson, 1983; Sampson and Laub, 1992) b. Younger men are expected to be more likely to be arrested in the subsequent wave. ( Hirschi and Gottfredson, 1983; Sampson and Laub, 1992) c. Younger men are predicted to be less likely to desist from being arrested. ( Hirschi and Gottfredson, 1983; Sampson and Laub, 1992) Family: Marriage 1. Marital Status: a. Married men are hypothesized to be le ss likely to be arrested in the same wave. (Farrington and We st, 1995; Rand, 1987; Sampson and Laub, 1990; 1993; Trassler, 1979; Warr, 1998) b. Married men are hypothesized to be le ss likely to be arrested in the subsequent wave. (Farrington and West, 1995; Laub, Nagin, and Sampson, 1998; Osborn and West , 1979; Rand, 1987; Sampson and Laub, 1990; 1993; Trassler, 1979; Warr, 1998; West, 1982) c. Married men are hypothesized to be more likely desist from being arrested. (Farrington and West , 1995; Laub, Nagin, and Sampson, 1998; Osborn and West, 1979; Ra nd, 1987; Sampson and Laub, 1990; 1993; Trassler, 1979; Warr, 1998; West, 1982) 2. Marital Satisfaction: a. Among married men, those who are in marriages that are rated more favorably are hypothesized to be less likely to be arrested in the same wave. (Laub, Nagin, and Sampson, 1998; Laub and Sampson, 1993) b. Among married men, those who are in marriages that are rated more favorably are hypothesized to be less likely to be arrested in the subsequent wave. (Laub, Na gin, and Sampson, 1998; Laub and Sampson, 1993) Family: Fatherhood 1. Fatherhood Status: a. Fathers are predicted to be less likely to be arrested in the same wave. (Farrington and Hawkins, 1991; Fa rrington and West, 1995; Hughes, 1998; Knight, Osborn, and West , 1977; Sampson and Laub, 1990; Trassler, 1979)

PAGE 118

105 b. Fathers are predicted to be less likely to be arrested in the subsequent wave. (Farrington and Hawkins, 1991; Farrington and West, 1995; Hughes, 1998; Knight, Osborn, a nd West, 1977; Sampson and Laub, 1990; Trassler, 1979) c. Fathers are predicted to be more lik ely to desist from being arrested. (Farrington and Hawkins, 1991; Fa rrington and West, 1995; Hughes, 1998; Knight, Osborn, and West , 1977; Sampson and Laub, 1990; Trassler, 1979) 2. Fatherhood Prior to the Age of 18: a. Men that father children before the age of 18 are hypothesized to be more likely to be arrested in th e same wave. (Christmon and Luckey, 1994; Dearden et al., 1995; Elster et al., 1987; Ketterlinius et al., 1992; Springarn and DuRant, 1996; Stoutheimer-Loeber and Wei, 1998) b. Men that father children before the age of 18 are hypothesized to be more likely to be arrested in th e subsequent wave. (Christmon and Luckey, 1994; Dearden et al., 1995; El ster et al., 1987; Ketterlinius et al., 1992; Springarn and DuRant, 1996; Stoutheimer-Loeber and Wei, 1998) c. Men that father children before the age of 18 are hypothesized to be less likely to desist from being arrested. (Christmon and Luckey, 1994; Dearden et al., 1995; Elst er et al., 1987; Kette rlinius et al., 1992; Springarn and DuRant, 1996; Stoutheimer-Loeber and Wei, 1998) 3. Residential Father: a. Among men, residence with some or al l of their children is expected to predict a lower likelihood of bei ng arrested in the same wave. (Farrington and West, 1995) b. Among men, residence with some or al l of their children is expected to predict a lower likelihood of being arrested in the subsequent wave. (Farrington and West, 1995) c. Among men, residence with some or al l of their children is expected to predict a higher likelihood of desistance from being arrested. (Farrington and West, 1995) 4. Rate Self as Father: a. Men who rate themselves more favorably as fathers are predicted to be less likely to be arrested in the same wave. (Farrington and Hawkins, 1991; Farrington and West, 1995; Maruna, 2001) b. Men who rate themselves more favorably as fathers are predicted to be less likely to use be arrested in the subsequent wave. (Farrington and Hawkins, 1991; Farrington and West, 1995; Maruna, 2001) 5. Feelings about their Relationship with their Child: a. Men who rate their relationships with their children more favorably are predicted to be less likely to be arre sted in the same wave. (Farrington and Hawkins, 1991; Farrington and West, 1995; Maruna, 2001) b. Men who rate their relationships with their children more favorably are predicted to be less likely to be arrested in the subsequent wave.

PAGE 119

106 (Farrington and Hawkins, 1991; Fa rrington and West, 1995; Maruna, 2001) Residential Status 1. Residential Stability a. Men who maintain the same residen ce for a longer period of time are hypothesized to be less likely to be arrested in the same wave. (Laub and Sampson, 1988; Sampson and Laub, 1990) b. Men who maintain the same residen ce for a longer period of time are hypothesized to be less likely to be arrested in the subsequent wave. (Laub and Sampson, 1988; Sampson and Laub, 1990) c. Men who maintain the same residen ce for a longer period of time are hypothesized to be more likely to de sist from being arrested. (Laub and Sampson, 1988; Sampson and Laub, 1990) Employment 1. Employment Status a. Men who are employed are predicted to be less likely to be arrested in the same wave. (Farrington et al., 1986; Sampson and Laub, 1990; 1993) b. Men who are employed are predicted to be less likely to be arrested in the subsequent wave. (Farringt on et al., 1986; Sampson and Laub, 1990; 1993) c. Men who are employed are predicted to be more likely to desist from being arrested. (Farrington et al., 1986; Sampson and Laub, 1990; 1993) 2. Job Satisfaction a. Men with a higher level of job satis faction are predicted to be less likely to be arrested during the same wave. (Irwin, 1970; Laub and Sampson, 1993; Shover, 1985; 1996) b. Men with a higher level of job satis faction are predicted to be less likely to be arrested during the subsequent wave. (Irwin, 1970; Laub and Sampson, 1993; Shover, 1985; 1996) c. Men with a higher level of job satis faction are predicted to be more likely to desist from being arre sted. (Irwin, 1970; Laub and Sampson, 1993; Shover, 1985; 1996) Finances 1. Income

PAGE 120

107 a. Men with higher incomes are hypothesi zed to be less likely to be arrested during the same wave. (Pezzin, 1995; Trassler, 1979; Waldorf, Reinarman, and Murphy, 1991) b. Men with higher incomes are hypothesi zed to be less likely to be arrested during the subsequent wave. (Pezzin, 1995; Trassler, 1979; Waldorf, Reinarman, and Murphy, 1991) c. Men with higher incomes are hypothesize d to be more likely to desist from being arrested. (Pezzin, 1995; Trassler, 1979; Waldorf, Reinarman, and Murphy, 1991) 2. Welfare Use a. Men who use welfare are predicted to be more likely to be arrested in the same wave. (Laub and Sampson, 1988) b. Men who use welfare are predicted to be more likely to be arrested in the subsequent wave. (Laub and Sampson, 1988) c. Men who use welfare are predicted to be less likely to desist from being arrested. (Laub and Sampson, 1988) Education 1. GPA a. Men who held a higher Grade Point Average in Wave 1 are predicted to be less likely to be arrested in the subsequent wave. (Glueck and Glueck, 1968) 2. High School Diploma a. Men with a high school diploma are hypothesized to be less likely to be arrested within the same wave . (Glueck and Glueck, 1968; Horney, Osgood, and Marshall, 1993; Rand, 1987; Sampson and Laub, 1990; 1993) b. Men with a high school diploma are hypothesized to be less likely to be arrested within the subseque nt wave. (Glueck and Glueck, 1968; Horney, Osgood, and Marshall, 1993; Rand, 1987; Sampson and Laub, 1990; 1993) c. Men with a high school diploma are hypothesized to be more likely to desist from being arrested. (G lueck and Glueck, 1968; Horney, Osgood, and Marshall, 1993; Rand, 1987; Sampson and Laub, 1990; 1993) Prior deviance 1. Bad Behavior a. Men who engaged in bad behavior in Wave 1 are predicted to be more likely to be arrested subsequent wa ves. (Blumstein et al., 1985; Nagin and Paternoster, 1991; 2000)

PAGE 121

108 b. Men who engaged in bad behavior in Wave 1 are predicted to be less likely to desist from being arreste d. (Blumstein et al., 1985; Nagin and Paternoster, 1991; 2000) 2. Drug Use a. Men were previously used drugs are predicted to be more likely to be arrested in the same wave. (Horney et al., 1993; Nagin and Paternoster, 1991; 2000) b. Men were previously used drugs are predicted to be more likely to be arrested in subsequent waves. (Horney et al., 1993; Nagin and Paternoster, 1991; 2000) 3. Arrests a. Men were previously arrested are pr edicted to be more likely to be arrested in the same wave. (Blu mstein et al., 1985; Nagin and Paternoster, 1991; 2000) b. Men were previously arrested are pr edicted to be more likely to be arrested in subsequent waves. (B lumstein et al., 1985; Nagin and Paternoster, 1991; 2000) Hypotheses: Incarceration Age 1. Age of Participant: a. Younger participants are expected to be more likely to be incarcerated in the same wave. ( Hirschi and Gottfredson, 1983; Sampson and Laub, 1992) b. Younger men are expected to be more likely to be incarcerated in the subsequent wave. ( Hirschi and Gottfredson, 1983; Sampson and Laub, 1992) c. Younger men are less likely to de sist from being incarcerated. ( Hirschi and Gottfredson, 1983; Sampson and Laub, 1992) Family: Marriage 1. Marital Status: a. Married men are hypothesized to be le ss likely to be incarcerated in the same wave. (Farrington and West, 1995; Rand, 1987; Sampson and Laub, 1990; 1993; Trassler, 1979; Warr, 1998) b. Married men are hypothesized to be le ss likely to be incarcerated in the subsequent wave. (Farringt on and West, 1995; Laub, Nagin, and Sampson, 1998; Osborn and West , 1979; Rand, 1987; Sampson and Laub, 1990; 1993; Trassler, 1979; Warr, 1998; West, 1982) c. Married men are hypothesized to be more likely desist from being incarcerated. (Farrington and We st, 1995; Laub, Nagin, and Sampson,

PAGE 122

109 1998; Osborn and West, 1979; Ra nd, 1987; Sampson and Laub, 1990; 1993; Trassler, 1979; Warr, 1998; West, 1982) 2. Marital Satisfaction: a. Among married men, those who are in marriages that are rated more favorably are hypothesized to be less likely to be incarcerated in the same wave. (Laub, Nagin, and Sampson, 1998; Laub and Sampson, 1993) b. Among married men, those who are in marriages that are rated more favorably are hypothesized to be less likely to be incarcerated in the subsequent wave. (Laub, Na gin, and Sampson, 1998; Laub and Sampson, 1993) Family: Fatherhood 1. Fatherhood Status: a. Fathers are predicted to be less likel y to be incarcerated in the same wave. (Farrington and Hawkins, 1991; Farrington and West, 1995; Hughes, 1998; Knight, Osborn, a nd West, 1977; Sampson and Laub, 1990; Trassler, 1979) b. Fathers are predicted to be less lik ely to be incarcerated in the subsequent wave. (Farrington and Hawkins, 1991; Farrington and West, 1995; Hughes, 1998; Knight, Osborn, and West, 1977; Sampson and Laub, 1990; Trassler, 1979) c. Fathers are predicted to be more likely to desist from being incarcerated. (Farrington and Hawk ins, 1991; Farrington and West, 1995; Hughes, 1998; Knight, Osborn, and West, 1977; Sampson and Laub, 1990; Trassler, 1979) 2. Fatherhood Prior to the Age of 18: a. Men that father children before the age of 18 are hypothesized to be more likely to be incarcerated in the same wave. (Christmon and Luckey, 1994; Dearden et al., 1995; El ster et al., 1987; Ketterlinius et al., 1992; Springarn and DuRant, 1996; Stoutheimer-Loeber and Wei, 1998) b. Men that father children before the age of 18 are hypothesized to be more likely to be incarcerated in the subsequent wave. (Christmon and Luckey, 1994; Dearden et al., 1995; El ster et al., 1987; Ketterlinius et al., 1992; Springarn and DuRant, 1996; Stoutheimer-Loeber and Wei, 1998) c. Men that father children before the age of 18 are hypothesized to be less likely to desist from being incarcerated. (Christmon and Luckey, 1994; Dearden et al., 1995; Elster et al., 1987; Ketterlinius et al., 1992; Springarn and DuRant, 1996; Stoutheimer-Loeber and Wei, 1998) 3. Residential Father: a. Among men, residence with some or al l of their children is expected to predict a lower likelihood of bei ng incarcerated in the same wave. (Farrington and West, 1995)

PAGE 123

110 b. Among men, residence with some or al l of their children is expected to predict a lower likelihood of bei ng incarcerated in the subsequent wave. (Farrington and West, 1995) c. Among men, residence with some or al l of their children is expected to predict a higher likelihood of de sistance from being incarcerated. (Farrington and West, 1995) 4. Rate Self as Parent: a. Men who rate themselves more favorably as fathers are predicted to be less likely to be incarcerated in the same wave. (Farrington and Hawkins, 1991; Farrington and West, 1995; Maruna, 2001) b. Men who rate themselves more favorably as fathers are predicted to be less likely to use be incarcerated in the subsequent wave. (Farrington and Hawkins, 1991; Farrington and West, 1995; Maruna, 2001) 5. Feelings about their Relationship with their Child: a. Men who rate their relationships with their children more favorably are predicted to be less likely to be incarcerated in the same wave. (Farrington and Hawkins, 1991; Fa rrington and West, 1995; Maruna, 2001) b. Men who rate their relationships with their children more favorably are predicted to be less likely to be in carcerated in the subsequent wave. (Farrington and Hawkins, 1991; Fa rrington and West, 1995; Maruna, 2001) Residential Status 1. Residential Stability a. Men who maintain the same residen ce for a longer period of time are hypothesized to be less likely to be incarcerated in the same wave. (Laub and Sampson, 1988; Sampson and Laub, 1990) b. Men who maintain the same residen ce for a longer period of time are hypothesized to be less likely to be incarcerated in the subsequent wave. (Laub and Sampson, 1988; Sampson and Laub, 1990) c. Men who maintain the same residen ce for a longer period of time are hypothesized to be more likely to desist from being incarcerated. (Laub and Sampson, 1988; Sampson and Laub, 1990) Employment 1. Employment Status a. Men who are employed are predicte d to be less likely to be incarcerated in the same wave. (F arrington et al., 1986; Sampson and Laub, 1990; 1993) b. Men who are employed are predicte d to be less likely to be incarcerated in the subsequent wa ve. (Farrington et al., 1986; Sampson and Laub, 1990; 1993)

PAGE 124

111 c. Men who are employed are predicted to be more likely to desist from being incarcerated. (Farrington et al., 1986; Sampson and Laub, 1990; 1993) 2. Job Satisfaction a. Men with a higher level of job satis faction are predicted to be less likely to be incarcerated during th e same wave. (Irwin, 1970; Laub and Sampson, 1993; Shover, 1985; 1996) b. Men with a higher level of job satis faction are predicted to be less likely to be incarcerated during th e subsequent wave. (Irwin, 1970; Laub and Sampson, 1993; Shover, 1985; 1996) c. Men with a higher level of job satis faction are predicted to be more likely to desist from being incarcerated. (Irwin, 1970; Laub and Sampson, 1993; Shover, 1985; 1996) Finances 1. Income a. Men with higher incomes are hypothesi zed to be less likely to be incarcerated during the same wa ve. (Pezzin, 1995; Trassler, 1979; Waldorf, Reinarman, and Murphy, 1991) b. Men with higher incomes are hypothesi zed to be less likely to be incarcerated during the subsequent wave. (Pezzin, 1995; Trassler, 1979; Waldorf, Reinarman, and Murphy, 1991) c. Men with higher incomes are hypothesize d to be more likely to desist from being incarcerated. (Pezzin, 1995; Trassler, 1979; Waldorf, Reinarman, and Murphy, 1991) 2. Welfare Use a. Men who use welfare are predicted to be more likely to be incarcerated in the same wave. (Laub and Sampson, 1988) b. Men who use welfare are predicted to be more likely to be incarcerated in the subsequent wave. (Laub and Sampson, 1988) c. Men who use welfare are predicted to be less likely to desist from being incarcerated. (Laub and Sampson, 1988) Education 1. GPA a. Men who held a higher Grade Point Average in Wave 1 are predicted to be less likely to be incarcerat ed in the subsequent wave. (Glueck and Glueck, 1968) 2. High School Diploma a. Men with a high school diploma are hypothesized to be less likely to be incarcerated within the same wave. (Glueck and Glueck, 1968; Horney, Osgood, and Marshall, 1993; Rand, 1987; Sampson and Laub, 1990; 1993)

PAGE 125

112 b. Men with a high school diploma are hypothesized to be less likely to be incarcerated within the subs equent wave. (Glueck and Glueck, 1968; Horney, Osgood, and Marshall, 1993; Rand, 1987; Sampson and Laub, 1990; 1993) c. Men with a high school diploma are hypothesized to be more likely to desist from being incarcerated. (Glueck and Glueck, 1968; Horney, Osgood, and Marshall, 1993; Rand, 1987; Sampson and Laub, 1990; 1993) Prior behavior 1. Bad Behavior a. Men who engaged in bad behavior in Wave 1 are predicted to be more likely to be incarcerated in the s ubsequent wave. (Blumstein et al., 1985; Nagin and Paternoster, 1991; 2000) b. Men who engaged in bad behavior in Wave 1 are predicted to be less likely to desist from being incarcerat ed. (Blumstein et al., 1985; Nagin and Paternoster, 1991; 2000) 2. Drug Use a. Men were previously used drugs are predicted to be more likely to be incarcerated in the same wave. (Horney et al., 1993; Nagin and Paternoster, 1991; 2000) b. Men were previously used drugs are predicted to be more likely to be incarcerated in subsequent waves. (Horney et al., 1993; Nagin and Paternoster, 1991; 2000) 3. Arrests a. Men were previously arrested are pr edicted to be more likely to be incarcerated in the same wave. (B lumstein et al., 1985; Nagin and Paternoster, 1991; 2000) b. Men were previously arrested are pr edicted to be more likely to be incarcerated in subsequent waves. (Blumstein et al., 1985; Nagin and Paternoster, 1991; 2000) Attrition Analyses were conducted on the indepe ndent and dependent variables (cross sectional) to assess if differe nces existed between participan ts that did and did not drop out. Depending on the type of variable unde r consideration (dichotomous, categorical, etc.) either T-tests, Chi-squared, or Mann Whitney-U tests were run. Wave 1 to Wave 2: Chi-square analyses were conduc ted to assess if differences existed on the Wave 1 variables In School, We lfare Use, Father in the Home, and Bad

PAGE 126

113 Behavior by Survey Attriti on (Continued vs. Drop-out; LOOK UP VALUES) from Wave 1 to Wave 2 (N = 351). Results indicated that there were no significant differences on In School, where one cell had an expected count of less than five, hence Yate’s Continuity Correction was used (Chi-square = .001, df = 1, p = 1.00) by Survey Attrition. Likewise there were no significant di fferences on Welfare Use (Chi-square = 1.10, df = 1, p = 0.30), Father in the House (Chi-square = .02, df = 1, p = 0.90), and Bad Behavior (Chisquare = 1.29, df = 1, p = 0.26) by Survey Attri tion. Results suggested that there were no significant differences between participants th at continued in the survey and those who dropped out from Wave 1 to Wave 2. A Mann Whitney-U test was conducted on Ti me at Current Residence by Survey Attrition (Continued vs. Drop-out) from Wave 1 to Wave 2. Results suggested that no significant differences existed, U(351) = 9699.50, p = .67, on Time at Current Address by Survey Attrition. Three t-test s were conducted to assess if mean differences existed on Age, Family Income, and GPA by Survey A ttrition (Continued vs. Drop-out) from Wave 1 to Wave 2. Results suggested that no si gnificant differences exis ted on Age, t(351) = 1.68, p = .09, Income, t(351) = -1.06, p = 0.29, and GPA, t(351) = -.034, p = .074 by Survey Attrition. Wave 2 to Wave 3: Chi-square analyses were conduc ted to assess if differences existed on Employment, In School, High Sc hool Degree, Fatherhood, Fatherhood prior to Age 18, Marital Status, Welfare Use, Wave 2 Drug Use, Wave 2 Arrests, and Wave 2 Incarceration by Survey Attriti on (Continued vs. Drop-out) from Wave 2 to Wave 3 (N = 277). Results indicated that there were no significant differences on Employment (Chisquare = 0.22, df = 1, p = 0.64), In School (Chi-square = 0.22, df = 1, p = 0.64), High

PAGE 127

114 School Degree (Chi-square = 0.00, df = 1, p = 1.00), Fatherhood (Chi-square = 2.82, df = 1, p = 0.09), Fatherhood prior to Age 18 (Chi -square = 3.23, df = 1, p = 0.07), Welfare Use (Chi-square = 0.30, df = 1, p = 0.58), and Wave 2 Drug Use (Chi-square = 1.41, df = 1, p = 0.23). Marital Status, Wave 2 Incarcera tion, and Wave 2 Arrests all had one cell that had an expected count of less than fi ve, therefore Yates Continuity Correction was used. There were no significant difference s on Marital Status (C hi-square = 1.38, df = 1, p = 0.24), Wave 2 Incarceration (Chi-square = 2.00, df = 1, p = 0.16), and Wave 2 Arrests (Chi-square = 2.00, df = 1, p = 0.18) by Survey Attrition. A Mann Whitney-U test was conducted on Ti me at Current Residence by Survey Attrition (Continued vs. Drop-out) from Wave 2 to Wave 3. Results suggested that no significant difference existed, U(277) = 7455.00, p = .56, on Time at Current Address by Survey Attrition. Three t-te sts were conducted to assess if a mean difference existed on Age, Income, and Job Satisfaction by Surv ey Attrition (Continued vs. Drop-out) from Wave 2 to Wave 3. Results suggested th at no significant difference existed on Age, t(277) = 0.14, p = 0.89, Income, t(277) = 1.30, p = 0.20, or Job Satisfaction, t(277) = 0.70, p = 0.48, by Survey Attrition. Results s uggested that there were no significant differences between participants that conti nued in the survey and those who dropped out from Wave 2 to Wave 3. Wave 3 to Wave 4: Chi-square analyses were conduc ted to assess if differences existed on Fatherhood, Marital Status, High Sc hool Degree, Wave 3 Drug Use, Wave 3 Arrests, Wave 3 Incarceration, Welfare Us e, Residential Father hood, Fatherhood prior to Age 18, and Employment by Survey Attrition (Continued vs. Drop-Out) from Wave 3 to Wave 4 (N=210). Results indicated that there were no significant differences on

PAGE 128

115 Fatherhood (Chi-square = 0.40, df = 1, p = 0.53), Marital Status (Chi-square = 2.56, df = 1, p = 0.10), High School Degree (Chi-square = 0.32, df = 1, p = 0.57), Wave 3 Drug Use (Chi-square = 0.09, df = 1, p = 0.77), Wave 3 Incarceration (Chi-square = 0.37, df = 1, p = 0.54), Welfare Use (Chi-square = 0.03, df = 1, p = 0.86), and Residential Fatherhood (Chi-square = 0.73, df = 1, p = 0.39) by Survey Attrition. Fatherhood prior to Age 18 had one cell that had an expected count of le ss than five, therefore Yates Continuity Correction was used. There were no signifi cant differences on Fatherhood prior to Age 18 (Chi-square = 0.70, df = 1, p = 0.42) by Survey Attrition. A significant difference existed on Wave 3 Arrests (Chi-square = 4.20, df = 1, p < 0.05) by Survey Attrition. As one cell had an expected count of less than five, Yate’s Continuity Correction was used. Individua ls who dropped out ha d a more frequent occurrence of Wave 3 Arrests than those w ho continued in the su rvey. A significant difference also existed on Employment (Chi -square = 11.51, df = 1, p < .01). Participants who remained in the study were more fre quently employed than those who dropped out from Wave 3 to Wave 4. A Mann Whitney-U test was conducted on Ti me at Current Residence by Survey Attrition (Continued vs. Drop-out) from Wave 3 to Wave 4. Results suggested that no significant difference existed, U(210) = 3789.50, p = .76, on Time at Current Address by Survey Attrition. Three t-te sts were conducted to assess if a mean difference existed on Age, Income, and Job Satisfaction by Surv ey Attrition (Continued vs. Drop-out) from Wave 3 to Wave 4. Results suggested th at no significant difference existed on Age, t(210) = 0.02, p = 0.99, Income, t(210) = 0.46, p = 0.65, or Job Satisfaction, t(210) = 0.65, p = 0.52, by Survey Attrition.

PAGE 129

116 Wave 4 to Wave 5: Chi-square analyses were conduc ted to assess if differences existed on Marital Status, Fatherhood, Reside ntial Father, Fatherhood prior to Age 18, High School Degree, Employment, Welfare Use, Wave 4 Drug Use, Wave 3 Arrests, and Wave 4 Incarceration by Surv ey Attrition (Continued vs. Drop-out) from Wave 4 to Wave 5. Results indicated th at there were no significant differences on Marital Status (Chi-square = 0.28, df = 1, p = 0.59), Fath erhood (Chi-square = 0.49, df = 1, p = 0.49), Residential Father (Chi-square = 0.48, df = 1, p = 0.49), Fatherhood prior to Age 18 (Chisquare = 3.27, df = 1, p = 0.70), High School Degree (Chi-square = 0.48, df = 1, p = .49), Employment (Chi-square = 3.65, df = 1, p = 0.06), Welfare Use (Chi-square = 0.05, df = 1, p = 0.83), Wave 4 Drug Use (Chi-square = 0.01, df = 1, p = 0.97), and Wave 4 Incarceration (Chi-square = 0.72, df = 1, p = 0.79) by Survey Attrition. Wave 3 Arrests (No Wave 4 measure of Arrest s was available) had one cell with an expected count less than five, hence Yate’s Continuity Corr ection was used. There was no significant difference on Wave 3 Arrests (Chi-square = 0.45, df = 1, p = 0.51) by Survey Attrition. A Mann Whitney-U test was conducted on Ti me at Current Residence by Survey Attrition (Continued vs. Drop-out) from Wave 4 to Wave 5. Results suggested that no significant difference existed, U(177) = 3413.50, p = .27, on Time at Current Address by Survey Attrition. Three t-te sts were conducted to assess if a mean difference existed on Age, Income, and Job Satisfaction by Surv ey Attrition (Continued vs. Drop-out) from Wave 4 to Wave 5. Results suggested th at no significant difference existed on Job Satisfaction, t(177) = -0.44, p = 0.66, by Survey Attrition. Results suggested that differences existed on Age, t( 177) = -2.42, p < 0.05, as partic ipants that dropped out of the study has a higher mean age (Mean = 34.61, SD = 1.64), compared to those who

PAGE 130

117 stayed in the study (Mean = 34.05, SD = 1.43). Re sults also indicated that a significant difference existed on Income, t(177) = 2.71, p < 0.01, as participants who stayed in the study had a higher mean income (Mean = 5.29, SD = 1.40) compared to those who dropped out of the study (Mean = 4.55, SD = 1.90) from Wave 4 to Wave 5. Three variables (Wave 3 Arrests, Wave 4 Age, and Wave 4 Income) suffered from non-random attrition were adjusted. Howeve r, prior research suggested that if participants that stayed in the survey and those that dr opped out of the program were “indistinguishable at th e start of the study, then overall c onclusions at the end of the study are strengthened even with lo ss of attrition (Goodwin, 2001).” In this study there were no significant differences from Wave 1 to Wave 2 between those that stayed in the study and those that dropped out. Future research on this data set will substitute a scale mean for the missing items (Brame and Piquero, 2003; Menard, 2002).

PAGE 131

118 CHAPTER 5 RESULTS: DRUG USE This chapter examines drug use and the va riables that were utilized to predict recent substance use. Each model utilized a variety of theoretically supported variables to attempt to predict drug use in each wa ve. While most of the same drugs were considered for each wave, there were minor va riations. Therefore, the brief description of each model is accompanied by a list of the drugs whose use was measured. The first section of this chapter examin es the various cross-sectional and lagged models that were run. The following section displays the models th at sought to predict desistance from drug use. Prior to each secti on an in depth discussion of what follows is provided. Cross-Sectional and Lagged Models Drug use was conceptualized as using any of the drugs in question within 30 days of the interview. A list of the drugs for each wave preceded each model. Measurements of drug use in waves 2 through 5 allowed the dependent variables to specify whether or not each participant had used a ny of the drugs in question in the last 30 days. In addition to the standard models, models were also pr ovided that examined the strength of social bonds among specific populations. Waves that contained variables which were conceptualized as measures of the strengt h of social bonds among husbands and fathers contained addition models that used these a dditional variables to predict drug use. Two measures of the strength of the fatherhood bond were available and utilized for Wave 3 independent variables predicting both Wave 3 and Wave 4 dependent variables. One

PAGE 132

119 measure of marital success was utilized in the following models: Wave 3 independent variables predicting Wave 3 dependent vari ables, and Wave 5 independent variables predicting Wave 5 dependent variables. W ithin all of these models, only the men who held these bonds (fathers or husbands) were selected as cases for analysis. Wave 1 Independent Variables Predicting Wave 2 Drug Use In this model the variables from Wave 1 (1968-1970) were used to predict drug use during Wave 2 (1975-1976). A direct logist ic regression was conducted on the outcome variable of Drugs Use in Wave 2 and the fo llowing Wave 1 predictors: Age, being in school, family income, welfare use, GPA, time at current residence, presence of a father in the home, and a measure of juvenile delinquency. The drugs under consideration were: marijuana, acid, cocaine , heroin, methadone, uppers, do wners, glue, and opiates. Participants who used any drugs within the past 30 days were considered to be users and coded as ,” all others were coded as .” Table 5-1 Wave 1 Independent Vari ables Predicting Wave 2 Drug Use VARIABLE B S.E Wald Sig. Exp(B) Age -.085 .089 .917 .338 .918 In School .082 .664 .015 .902 1.086 Income .134 .068 3.934 .047 1.144 Welfare Use .352 .289 1.490 .222 1.422 GPA .189 .190 .997 .318 1.208 Time at Res. -.081 .074 1.170 .279 .923 Father in home -.166 .294 .321 .571 .847 Bad Behav. .408 .351 1.353 .245 1.504 CONSTANT .441 1.682 .069 .793 1.554

PAGE 133

120 The test of the full model (all predictors ) against a constant only model was not significant (Chi-square = 9.66, df = 8, p > .05) indicating that th e set of Wave 1 predictors did not reliably dis tinguish between participants th at did and did not use drugs. Table 5-1 presents regression coefficients where Income re liably predicted Wave 2 Drug Use (N = 277). For every one unit increase in income, the likelihood of Wave 2 Drug Use increased 1.14 times. Wave 2 Independent Variables Predicting Wave 2 Drug Use In this model the variables from Wave 2 (1975-1976) were used to predict drug use during Wave 2 (1975-1976). A di rect logistic regression wa s conducted on the out come variable of Wave 2 Drug Use and the following Wave 2 predictors: Age, being in school, employment, attainment of a high school degr ee, job satisfaction, welfare use, income, time at current residence, fatherhood, fa therhood prior to age 18, marriage, a measurement of juvenile deli nquency, and prior arrests (N o measure of prior drug use was available). Participants who used any dr ugs within the past 30 days were considered to be users and coded as ,” all othe rs were coded as .” The drugs under consideration were: marijuan a, acid, cocaine, heroin, me thadone, uppers, downers, glue, and opiates. The test of the full model (all predictors ) against a constant only model was not significant (Chi-square = 12.82, df = 13, p > .05) indicating that th e set of Wave 2 predictors did not reliably dis tinguish between participants th at did and did not use drugs during Wave 2. Table 5-2 presents regressi on coefficients where none of the thirteen Wave 2 variables were significant pred ictors of Wave 2 Drug Use (N = 277).

PAGE 134

121 Table 5-2 Wave 2 Independent Vari ables Predicting Wave 2 Drug Use VARIABLE B S.E. Wald Sig. Exp(B) Age -.114 .099 1.315 .251 .892 In School -.192 .297 .420 .517 .825 Employed .264 .275 .925 .336 1.302 HS Degree -.147 .289 .259 .611 .863 Job Satisfaction -.084 .127 .434 .510 .920 Welfare Use .107 .423 .064 .801 1.113 Income -.048 .099 .241 .624 .953 Time at Res. .004 .065 .004 .950 1.004 Father .569 .376 2.290 .130 1.766 Fath. b/f 18 .251 .592 .179 .672 1.285 Married .156 .776 .041 .840 1.169 Bad Behav. .277 .357 .604 .437 1.319 Prior Arrests -.929 .545 2.908 .088 .395 CONSTANT 3.057 2.152 2.018 .155 21.267 Wave 2 Independent Variables Predicting Wave 3 Drug Use In this model the variables from Wave 2 (1975-1976) were used to predict drug use during Wave 3 (1983-1984). A direct logist ic regression was conducted on Wave 3 Drug Use as the outcome variable and the following Wave 2 predictors: Age, being in school, employment, attainment of a high school degr ee, job satisfaction, welfare use, income, time at current residence, fatherhood, father hood prior to age 18, marriage, prior drug use, and prior arrests. Participants who us ed any drugs within the past 30 days were considered to be users and coded as ,” al l others were coded as .” The drugs under consideration were: marijuan a, acid, pcp, cocaine, heroin, methadone, uppers, quaaludes, tranquilizers, sedatives, inhalants, nitrat es, talwins, codeine, and morphine.

PAGE 135

122 A direct logistic regre ssion was conducted on Wave 3 Drug Use as the outcome variable and the following Wave 2 predic tors: Age, being in school, employment, attainment of a high school degree, job satisfa ction, welfare use, income, time at current residence, fatherhood, fatherhood prior to age 18, marriage, prior drug use, and prior arrests. The test of the full model (all pr edictors) against a cons tant only model was significant (Chi-square = 48.59, df = 13, p < .001) indicating that the set of predictors reliably distinguished between participants that did and did not use drugs during Wave 3. Table 5-3 presents regression coefficients where Prior Drug Use reliably predicted Wave 3 Drug Use (N = 210). Participants w ho used drugs during the prior period of Table 5-3 Wave 2 Independent Vari ables Predicting Wave 3 Drug Use VARIABLE B S.E Wald Sig. Exp(B) Age .003 .128 .000 .984 1.003 In School -.176 .372 .225 .635 .838 Employed .192 .353 .296 .587 1.212 HS Degree -.483 .373 1.672 .196 .617 Job Sat. -.076 .160 .225 .635 .927 Welfare Use -.700 .529 1.753 .186 .496 Income .138 .126 1.191 .275 1.148 Time at Res. .044 .079 .318 .573 1.045 Father .925 .498 3.458 .063 2.523 Father b/f 18 -.510 .804 .402 .526 .601 Married -7.781 15.537 .251 .617 .000 Prior Drug Use 1.606 .322 24.802 .000 4.982 Prior Arrests -.743 .701 1.126 .289 .476 CONSTANT -1.165 2.877 .164 .685 .312

PAGE 136

123 measurement had odds of Wave 3 drug use th at were 4.98 greater than men who did not engage in previous drug use. Wave 3 Independent Variables Predicting Wave 3 Drug Use In this model the variables from Wave 3 (1983-1984) were used to predict drug use during Wave 3 (1983-1984). A direct logistic regression was conducted on the outcome variable of Wave 3 Drug Use and the follo wing Wave 3 predictors: Age, employment status, income, welfare use, job satisfaction, attainment of a high school degree, time at current residence, fatherhood, marital status , residential fatherhood, fatherhood prior to age 18, prior drug use, and prior arrests. Part icipants who used any drugs within the past 30 days were considered to be users and coded as ,” all others were coded as .” The drugs under consideration were: marijuana, acid, pcp, cocaine, heroin, methadone, uppers, quaaludes, tranquilizers, sedatives, inha lants, nitrates, talwins, codeine, and morphine. The test of the full model (all predicto rs) against the constant only model was significant (Chi-square = 46.45, df = 13, p < .001) indicating that the set of predictors reliably distinguished between participants th at did and did not use drugs in Wave 3. Table 5-4 presents the regression coefficients where Time at Current Residence and Drug Use Wave 2 reliably predicted Wave 3 Drug Us e (N = 210). As par ticipants tended to use drugs in Wave 2, the odds that the partic ipants will use drugs in Wave 3 increased by 4.67. For every one unit increase in Time at Residence, the odds decreased by 17.6% that the participant would use drugs in Wave 3. Strength of Social Bonds: Fatherhood The following model examined the strength of fathers’ social bonds with their children. The measurements of strength were conceptualized in two distinct manners.

PAGE 137

124 Table 5-4 Wave 3 Independent Vari ables Predicting Wave 3 Drug Use VARIABLE B S.E Wald Sig. Exp(B) Age -.030 .120 .062 .803 .970 Employed .299 .400 .560 .454 1.349 Income .054 .126 .187 .666 1.056 Welfare Use .684 .507 1.821 .177 1.981 Job Sat. -.108 .164 .435 .510 .897 HS Degree -.247 .378 .428 .513 .781 Time at Res. -.193 .087 4.980 .026 .824 Father .492 .406 1.456 .226 1.635 Married -.441 .491 .806 .369 .643 Res. Father .441 .518 .723 .395 1.554 Father b/f 18 -.198 .623 .101 .751 .821 Prior Drug Use 1.541 .324 22.565 .000 4.667 Prior Arrests .651 .387 2.831 .092 1.917 CONSTANT .341 2.496 .019 .891 1.406 First, men were asked to rate themselves as fathers. Second, men were asked how they feel about their relationship with their children. For bot h of these variables, higher numbers indicated better situations (An in-d epth description of the coding of these variables can be found in Chapter 4: Methods ). A model was run incorporating both of these variables. In this m odel the cases under considerati on were limited to fathers who participated in the wave in which the dependent variable was collected. This model used Wave 3 independent vari ables to predict drug use in Wave 3. A direct logistic regression was conducted on the outcome variable of Wave 3 Drug Use and the following Wave 3 predictors: Age, em ployment status, income, welfare use, job satisfaction, attainment of a high school diploma, time at current residence, marital status,

PAGE 138

125 residential fatherhood, fatherhood prior to age 18, prior arrests, prior drug use, a rating of one’s success as a parent, a nd a rating of one’s relations hip with their child(ren). Table 5-5 Wave 3 Independent Variables Predicting Wave 3 Drug Use among Fathers VARIABLE B S.E. Wald Sig. Exp(B) Age -.084 .184 .207 .649 .920 Employed .975 .662 2.168 .141 2.652 Income -.088 .210 .177 .674 .915 Welfare Use 1.009 .821 1.509 .219 2.742 Job Satisfaction -.371 .284 1.700 .192 .690 HS Degree -.165 .548 .091 .763 .848 Time at Res. -.393 .156 6.319 .012 .675 Married -1.127 .694 2.640 .104 .324 Res. Father .967 .682 2.012 .156 2.630 Father b/f 18 -.467 .721 .419 .517 .627 Prior Arrests .564 .572 .974 .324 1.758 Prior Drug Use 1.522 .564 7.274 .007 4.579 Rate Self as parent -.620 .427 2.110 .146 .538 Relationship w/ Child -.254 .329 .595 .440 .776 CONSTANT 6.162 4.288 2.065 .151 474.296 The test of the full model (all predictors) against a constant only model was significant (Chi-square = 30.94, df = 14, p < .05) indicating that the set of predictors reliably distinguished between fathers w ho did and did not use drugs during Wave 3. Table 5-5 presents regression coefficients where Time at Current Residence and Prior Drug Use reliably predicted Wave 3 Drug Us e among fathers (N = 106). For every one

PAGE 139

126 unit increase in Time at Residence, the odds decrease by 32.5% that the participant will use drugs in Wave 3. Fathers who used drugs in Wave 2 held odds of Wave 3 drug Use that were 4.58 higher than fathers who refrained from drug use in Wave 2. Strength of Social Bonds: Marriage The next model examined the strength of marital bonds. Men were asked how satisfied they were with their relationship w ith their spouse. As w ith the prior measures of the strength of social bonds , higher numbers were measures of higher strength or in this case satisfaction (An in-d epth description of the coding of this variable can be found in Chapter 4: Methods). A direct logistic regressi on was conducted on the outcome variable of Wave 3 Drug Use and the follo wing Wave 3 predictors: Age, employment status, income, job satisfacti on, attainment of a high school diploma, time at current residence, fatherhood, residential fatherhood, fa therhood prior to age 18, prior arrests, prior drug use, and a rating of one’s marital satisfaction. The full model could not be run as the variable noting welfar e use was problematic. When a cross tabulation was run, one of the cells contained zero part icipants (All of the particip ants who used welfare used drugs during Wave 3). The test of the full model (all predictors ) against a constant only model was not significant (Chi-square = 18.04, df = 13, p > .05) i ndicating that the set of predictors did not reliably distinguish between married men who did and did not use drugs during Wave 3. Table 5-6 presents regressi on coefficients where none of the twelve Wave 3 predictors reliably predicted Wave 3 Drug Use (N = 46). Wave 3 Independent Variables Predicting Wave 4 Drug Use In this model the variables from Wave 3 (1983-1984) were used to predict drug use during Wave 4 (1989-1990). A direct logist ic regression was conducted on the outcome

PAGE 140

127 variable of Wave 4 Drug Use and the follo wing Wave 3 predictors: Age, employment status, income, welfare use, job satisfaction, attainment of a high school degree, time at current residence, fatherhood, marital status , residential fatherhood, fatherhood prior to age 18, prior drug use, and prior arrests. Pa rticipants who used or injected any drugs within the past 30 days were considered to be users and coded as ,” all others were coded as .” The drugs under consider ation were: marijuana, acid, dust, crack, powdered cocaine, heroin, speedball, methadone, uppers, quaaludes, sedatives, tranquilizers, inhalants, glue, talwins, codein e, morphine, ecstasy, and “any other drug.” Table 5-6 Wave 3 Independent Variables Predicting Wave 3 Drug Use among Married Men VARIABLE B S.E. Wald Sig. Exp(B) Age -.021 .320 .004 .949 .980 Employed 1.428 1.691 .714 .398 4.171 Income -.761 .461 2.724 .099 .467 Job Satisfaction -.689 .600 1.319 .251 .502 HS Degree .321 1.115 .083 .773 1.379 Time at Res. -.228 .259 .770 .380 .796 Fatherhood -1.441 1.616 .795 .373 .237 Res. Father 2.050 1.526 1.806 .179 7.771 Father b/f 18 .093 1.813 .003 .959 1.097 Prior Arrests -1.155 1.166 .982 .322 .315 Prior Drug Use 1.095 .862 1.615 .204 2.991 Marital Satisfaction -1.239 .661 3.515 .061 .290 CONSTANT 9.300 6.620 1.973 .160 10932.831

PAGE 141

128 The test of the full model (all predictors) against a constant only model was significant (Chi-square = 29.68, df = 13, p < .05) indicating that the set of predictors reliably distinguished between participants that did and did not e ngage in Wave 4 drug use. Table 5-7 presents the regression coefficients where Income and Prior Drug Use reliably predicted Wave 4 Drug Use (N = 177) . For every one unit increase in Income, the odds decreased by 38% that the participan t would engage in Wave 4 drug use. As participants tended to use drugs in the prior period of meas urement, the odds of Wave 4 drug use increased by 4.58. Table 5-7 Wave 3 Independent Vari ables Predicting Wave 4 Drug Use VARIABLE B S.E. Wald Sig. Exp(B) Age -.135 .165 .672 .412 .874 Employed .030 .522 .003 .955 1.030 Income -.478 .173 7.621 .006 .620 Welfare .334 .616 .294 .588 1.397 Job Sat. -.090 .230 .155 .694 .914 HS Degree .800 .532 2.260 .133 2.225 Time at Res. .150 .124 1.484 .223 1.162 Father .271 .521 .270 .603 1.311 Married -.700 .753 .865 .352 .496 Res. Father -.420 .670 .393 .531 .657 Father b/f 18 .550 .744 .547 .460 1.733 Prior Drug Use 1.521 .490 9.611 .002 4.575 Prior Arrests .819 .867 .892 .345 2.267 CONSTANT 1.313 3.367 .152 .697 3.716

PAGE 142

129 Strength of Social Bonds: Fatherhood The following model examined the strength of fathers’ social bonds with their children. The measurements of strength were conceptualized in two distinct manners. First, men were asked to rate themselves as fathers. Second, men were asked how they felt about their relationship w ith their children. For both of these variables, higher numbers indicated better situations (An in-d epth description of the coding of these variables can be found in Chapter 4: Methods ). A model was run incorporating both of these variables. A direct logistic regressi on was conducted on the outcome variable of Wave 4 Drug Use and the following Wave 3 predictors: Age, employment status, income, welfare use, job satisfaction, attainment of a high school di ploma, time at current residence, marital status, residential fatherhood, fatherhood prior to age 18, prior arrests, prior drug use, a rating of one’s success as a parent, and a rati ng of one’s relationship with their child(ren). The test of the full model (all predictors) against a constant only model was significant (Chi-square = 26.24, df = 14, p < .05) indicating that the set of predictors reliably distinguished between fathers did a nd did not use drugs dur ing Wave 4. Table 58 presents regression coefficients where Income and High School Degrees reliably predicted Wave 4 Drug Use (N = 83). Men who held High School Diplomas by Wave 3 had 13.42 greater odds of using drugs than me n who did not graduate. Additionally, for every one unit increase in Income, the odds of drug use in Wave 4 decreased by 44.1%. Strength of Social Bonds: Marriage There were not enough participants availa ble to successfully run this model.

PAGE 143

130 Table 5-8 Wave 3 Independent Variables Predicting Wave 4 Drug Use among Fathers VARIABLE B S.E. Wald Sig. Exp(B) Age -.318 .303 1.105 .293 .727 Employed -.750 .821 .836 .361 .472 Income -.582 .268 4.724 .030 .559 Welfare Use 1.528 .984 2.411 .121 4.607 Job Satisfaction -.056 .375 .023 .880 .945 HS Degree 2.596 .995 6.807 .009 13.415 Time at Res. .174 .198 .774 .379 1.191 Married -1.179 1.030 1.309 .253 .308 Res. Father .169 .809 .044 .835 1.184 Father b/f 18 .504 .905 .311 .577 1.656 Prior Arrests .788 1.048 .566 .452 2.199 Prior Drug Use 1.516 .925 2.685 .101 4.554 Rate Self as parent -.519 .468 1.229 .268 .595 Relationship w/ Child .391 .512 .582 .445 1.478 CONSTANT 5.765 6.197 .865 .352 318.938 Wave 4 Independent Variables Predicting Wave 4 Drug Use In this model the variables from Wave 4 (1989-1990) were used to predict drug use during Wave 4 (1989-1990). A direct logist ic regression was conducted on the outcome variable of Wave 4 Drug Use and the followi ng predictor variables: Age, employment status, income, welfare use, job satisfaction, attainment of a high school degree, time at current residence, fatherhood, marital status , residential fatherhood, fatherhood prior to age 18, prior drug use, and prior arrests. Pa rticipants who used or injected any drugs

PAGE 144

131 within the past 30 days were considered to be users and coded as ,” all others were coded as .” The drugs under considerat ion were: marijuana, acid, dust, crack, powdered cocaine, heroin, speedball, methadone, uppers, quaaludes, sedatives, tranquilizers, inhalants, glue, talwins, codein e, morphine, ecstasy, and “any other drug.” Table 5-9 Wave 4 Independent Vari ables Predicting Wave 4 Drug Use VARIABLE B S.E. Wald Sig. Exp(B) Age -.213 .142 2.241 .134 .808 Employed .032 .528 .004 .952 1.032 Income .065 .163 .160 .690 1.067 Welfare 1.058 .687 2.372 .124 2.882 Job Sat. -.046 .258 .032 .858 .955 HS Degree .527 .546 .930 .335 1.693 Time at Res. -.058 .110 .280 .597 .943 Father .961 .550 3.054 .081 2.615 Married -.714 .504 2.006 .157 .490 Res. Father -1.016 .554 3.368 .066 .362 Father b/f 18 -.371 .721 .265 .607 .690 Prior Drug Use 1.445 .471 9.414 .002 4.241 Prior Arrests .423 .822 .264 .607 1.526 CONSTANT 4.368 4.888 .799 .371 78.905 The test of the full model (all predictors) against a constant only model was significant (Chi-square = 24.69, df = 13, p < .05) indicating that the set of predictors reliably distinguished between participants that did and did not use drugs in Wave 4. Table 5-9 presents the regression coefficien ts where Prior Drug Use reliably predicted Wave 4 Drug Use (N = 177). As participants tended to engage in prior drug use, the odds of Wave 4 drug use increased by 4.24.

PAGE 145

132 Wave 4 Independent Variables Predicting Wave 5 Drug Use In this model the variables from Wave 4 (1989-1990) were used to predict drug use during Wave 5 (1993-1994). A direct logist ic regression was conducted on the outcome variable of Wave 5 Drug Use and the follo wing Wave 4 predictors: Age, employment status, income, welfare use, job satisfaction, attainment of a high school degree, time at current residence, fatherhood, marital status , residential fatherhood, fatherhood prior to age 18, prior drug use, and prior arrests. Pa rticipants who used or injected any drugs within the past 30 days were considered to be users and coded as ,” all others were coded as .” The drugs under considerat ion were: marijuana, acid, dust, pcp, crack, powdered cocaine, heroin, speedball, methadone, uppers, quaaludes, sedatives, tranquilizers, inhalants, glue, aerosols, talwin s, codeine, morphine, and “any other drug.” The test of the full model (all predictors) against a constant only model was significant (Chi-square = 22.60, df = 13, p < .05) indicating that the set of predictors reliably distinguished between participants that did and did not e ngage in Wave 5 drug use. Table 5-10 presents the regression co efficients where Age and Fatherhood before Age 18 reliably predicted Wave 5 Drug Use (N = 168). For every one unit increase in age, the odds decreased by 34% that the partic ipant would engage in wave 5 drug use. As men tended to be fathers before the age of 18, the odds of Wave 5 drug use increased by 5.96 Wave 5 Independent Variables Predicting Wave 5 Drug Use In this model the variables from Wave 5 (1993-1994) were used to predict drug use during Wave 5 (1993-1994). A direct logist ic regression was conducted on the outcome variable of Wave 5 Drug Use and the follo wing Wave 5 predictors: Age, employment status, income, welfare use, job satisfaction, attainment of a high school degree, time at

PAGE 146

133 current residence, fatherhood, marital status , residential fatherhood, prior drug use, and prior arrests. Participants who used or inj ected any drugs within the past 30 days were considered to be users and coded as ,” al l others were coded as .” The drugs under consideration were: mariju ana, acid, dust, pcp, crack, powdered cocaine, heroin, speedball, methadone, uppers, quaaludes, seda tives, tranquilizers, inhalants, glue, aerosols, talwins, codeine, morphine, and “any other drug.” Table 5-10 Wave 4 Independent Vari ables Predicting Wave 5 Drug Use VARIABLE B S.E. Wald Sig. Exp(B) Age -.416 .188 4.891 .027 .660 Employed -.247 .666 .138 .710 .781 Income -.473 .256 3.412 .065 .623 Welfare -.039 1.089 .001 .971 .961 Job Sat. .178 .345 .266 .606 1.195 HS Degree .326 .546 .358 .550 1.386 Time at Res. -.162 .154 1.103 .294 .851 Father .669 .679 .970 .325 1.953 Married -.760 .749 1.029 .310 .468 Res. Father .640 .810 .624 .429 1.897 Father b/f 18 1.785 .790 5.112 .024 5.961 Prior Drug Use -.620 .716 .751 .386 .538 Prior Arrests -.825 1.191 .480 .489 .438 CONSTANT 15.451 6.684 5.343 .021 5131530.2 The test of the full model (all predictors ) against a constant only model was not significant (Chi-square = 19.25, df = 12, p >.05) indicating that the set of Wave 5 predictors did not reliably di stinguished between participants that did and did not use drugs in Wave 5. Table 5-11 presents regr ession coefficients where Job Satisfaction

PAGE 147

134 reliably predicted Wave 5 Drug Use (N = 168). For every one unit increase in job satisfaction, the odds of Wave 5 drug use decreased by 33.9%. Table 5-11 Wave 5 Independent Vari ables Predicting Wave 5 Drug Use VARIABLES B S.E Wald Sig. Exp(B) Age -.105 .125 .712 .399 .900 Employed -.546 .462 1.399 .237 .579 Income .125 .115 1.179 .278 1.133 Welfare Use .330 .445 .550 .458 1.391 HS Degree -.808 .503 2.582 .108 .446 Job Sat. -.414 .205 4.083 .043 .661 Time at Res. .042 .099 .180 .671 1.043 Father .376 .432 .756 .385 1.456 Res. Father -.637 .486 1.717 .190 .529 Married -.146 .432 .114 .735 .864 Prior Drug Use .517 .586 .778 .378 1.677 Prior Arrests -.840 1.186 .502 .478 .432 CONSTANT 4.128 4.677 .779 .377 62.077 Strength of Social Bonds: Marriage The next model examined the strength of marital bonds. Men were asked how satisfied they were with their relationship w ith their spouse. As w ith the prior measures of the strength of social bonds , higher numbers were measures of higher strength or in this case satisfaction (An in-d epth description of the coding of this variable can be found in Chapter 4: Methods). Only men who were married in Wave 5 were included in this model. A direct logistic regression was conducted on the out come variable of Wave 5 Drug Use and the following Wave 5 predictors : Age, employment status, income, job satisfaction, welfare use, attain ment of a high school diploma, time at current residence,

PAGE 148

135 residential fatherhood, prior arrests, prior drug use, and a rating of one’s marital satisfaction. The full model could not be run as the variable noting fatherhood was problematic. When a cross tabulation wa s run, one of the cells contained zero participants (All of the men who were not fathers did not use drugs during Wave 5). The test of the full model (all predictors ) against a constant only model was not significant (Chi-square = 11.95, df = 11, p > .05) i ndicating that the set of predictors did not reliably distinguish between married men who did and did not us e drugs during Wave 5. Table 5-12 presents regre ssion coefficients where attain ment of a High School Degree reliably predicted Wave 5 Drug Use (N = 69). Married men who had attained a high school degree by Wave 5 were 92.3% less likely to use drugs than married men who had not completed high school. Desistance Models This section of the chapter di splays all of the models th at examined desistance from drug use. In the following models the meas urement of drug use differed from the prior models in that any drug use during each wave was counted, as opposed to the prior crosssectional and lagged models which examined recent drug use (drug use within the past 30 days). There were two distinct measurements of desistance; Desist ance 2 (or Desist 2) and Desistance 3 (or Desist 3). These two m easurements differed with regards to the waves within which deviant behavior occurr ed and the waves within which desistance from such behaviors occurred. Desistance 2 consisted of participants who engaged in criminal behaviors in wave 2, but did not engage in a specif ic behavior in the following waves. Desistance 3 referred to men who engage d in criminal behavior in wave 3 but did not engage in a specific behavi or in the following waves. This chapter examined men who desisted from drug use. For both sets of models, desistance was examined through

PAGE 149

136 each wave up through the end of the study. De sistance 2 models pred icted desistance in Wave 3, in Waves 3 and 4, as well as in Wave s 3, 4, and 5. Desistance 3 models predict desistance in Wave 4, as well as in Waves 4 and 5. Table 5-12 Wave 5 Independent Variables Predicting Wave 5 Drug Use among Married Men VARIABLE B S.E. Wald Sig. Exp(B) Age .078 .235 .110 .740 1.081 Employed -1.089 1.357 .644 .422 .337 Income .288 .366 .390 .532 1.256 Job Satisfaction .116 .433 .072 .788 1.123 Welfare Use -3.287 1.784 3.394 .065 .037 HS Degree -2.563 1.250 4.203 .040 .077 Time at Res. .352 .259 1.848 .174 .1422 Res. Father -.567 .737 .591 .442 .567 Prior Arrests .397 1.410 .079 .779 1.487 Prior Drug Use .140 1.068 .017 .895 1.151 Marital Satisfaction .454 .342 1.765 .184 1.574 CONSTANT -4.990 9.737 .263 .608 .007 Desistance 2: Wave 2 Independent Vari ables Predicting Desistance Wave 3 As with all of the Desistance 2 models, this model used theoretically significant variables to predict the cessa tion of criminal behavior af ter 1982. This model utilized predictors from Wave 2 (1975-1976) to pred ict desistance from drug use from 1983 to 1988. A direct logistic regression was conduc ted on the outcome variable of Desistance and the following Wave 2 predictors: Age, School enrollment, Employment Status, Attainment of a High School Diploma, Job Sa tisfaction, Welfare Use, Income, Time at

PAGE 150

137 Current Residence, Fatherhood, Fatherhood prior to Age 18, Marital Status, and Juvenile delinquency. The test of the full model (all predictors ) against the constant only model was not significant (Chi-square = 10.78, df = 12, p > .05) indicating that the Wave 2 predictors did not reliably distinguish between particip ants who did and did not desist. Table 5-13 presents the regression coefficients where Ti me at current Residence reliably predicted desistance (N = 172). For every one unit increase in time at current residence, the odds of desistance from drug use decreased by 22.5%. Table 5-13 Desist 2: Wave 2 Independent Variables Predicting Desistance Wave 3 VARIABLES B S.E. Wald Sig. Exp(B) Age .053 .206 .067 .796 1.055 In School .583 .562 1.074 .300 1.791 Employed .364 .545 .445 .505 1.439 HS Degree -.612 .567 1.163 .281 .542 Job Sat. .073 .256 .081 .776 1.075 Welfare Use .738 .696 1.124 .289 2.091 Income .078 .209 .141 .708 1.081 Time at Res. -.255 .117 4.780 .029 .775 Father -.963 .857 1.262 .261 .382 Father b/f 18 .492 1.384 .127 .722 1.636 Married .130 1.277 .010 .919 1.138 Bad Behav. -.134 .745 .033 .857 .874 CONSTANT -2.682 4.552 .347 .556 .068 Desistance 2: Wave 3 Independent Vari ables Predicting Desistance Wave 3 This model also predicted desistance from drug use from 1983 to 1988. However, the predictors utilized here were collected during Wave 3 (1983-1984). A direct logistic regression was conducted on the outcome variab le of Desistance and the following Wave

PAGE 151

138 2 predictors: Age, Employment Status, A ttainment of a High School Diploma, Job Satisfaction, Welfare Use, Income, Time at Current Residence, Fatherhood, Residential Fatherhood, Marital Status, and Juvenile delinqu ency. The full model could not be run as the variable noting fatherhood prior to the age of 18 was problematic. When a cross tabulation was run, one of the cells contained zero participants (None of the participants were fathers prior to the age of 18 at wave 3 who also desisted from drug use during wave 3). The test of the full model (all predicto rs) against the constant only model was significant (Chi-square = 20.01, df = 11, p < .05) indicating that th e set of Wave 3 predictors reliably distinguish ed between men who did and did not desist. Table 5-14 presents regression coefficients where In come, Fatherhood, and attainment of a High School Degree reliably predicted desistance. For every one unit increase in income, a participant had 37.8% decreased odds of desistance. Men who were fathers by Wave 3 had 80.7% decreased odds of desistance. Fi nally, men who had attained a high school degree by Wave 3 had 5.77 greater odds of de sistance than men who had not attained the same educational achievement. Desistance 2: Wave 2 Independent Vari ables Predicting Desistance Waves 3-4 The next three models examined desist ance from drug use from 1983 to 1992. The first model predicted desistance using va riables from Wave 2 (1975-1976). A direct logistic regression was conducted on the out come variable of Desistance and the following Wave 2 predictors: Age, School enro llment, Employment, attainment of a High School Degree, Job Satisfaction, Welfare Use, Income, Time at current Residence, Fatherhood, Fatherhood prior to age 18, and Juvenile Delinquency.

PAGE 152

139 Table 5-14 Desist 2: Wave 3 Independent Variables Predicting Desistance Wave 3 VARIABLES B S.E. Wald Sig. Exp(B) Age .153 .188 .661 .416 1.166 Employed .199 .646 .095 .758 1.220 Income -.474 .197 5.817 .016 .622 Welfare Use .096 .768 .016 .900 1.101 Job Sat. .595 .334 3.177 .075 1.813 HS Degree 1.752 .737 5.656 .017 5.766 Time at Res. -.062 .137 .203 .653 .940 Father -1.646 .761 4.677 .031 .193 Res. Father 1.439 .934 2.371 .124 4.216 Married -.853 .841 1.029 .310 .426 Bad Behav. -.144 .734 .039 .844 .866 CONSTANT -5.624 3.879 2.102 .147 .004 The test of the full model (all predictors ) against a constant only model was not significant (Chi-square = 3.94, df = 12, p > .05) indicating that the set of predictors did not reliably distinguish between participants that did and did desist. Table 5-15 presents regression coefficients where none of the tw elve Wave 2 variable s reliably predicted desistance (N = 144). Desistance 2: Wave 3 Independent Variab les Predicting Desistance Waves 3-4 Here predictors from Wave 3 (1983-1984) were used to predict desistance (or cessation of any drug use) from 1983 to 1992. A direct logistic re gression was conducted on Desistance as the outcome variable a nd the following Wave 3 predictors: Age, Employment, Income, Welfare Use, Job Sa tisfaction, attainment of a High School Degree, Time at current Re sidence, Fatherhood, Residential Fatherhood, Marital Status, and Juvenile Delinquency. The full model c ould not be run as the variable noting

PAGE 153

140 Table 5-15 Desist 2: Wave 2 Independent Variables Predicting Desistance Waves 3-4 VARIABLES B S.E. Wald Sig. Exp(B) Age .053 .192 .077 .782 1.055 In School .179 .543 .108 .742 1.196 Employed .172 .497 .119 .730 1.187 HS Degree -.137 .532 .066 .797 .872 Job Sat. .234 .240 .948 .330 1.263 Welfare Use .002 .751 .000 .998 1.002 Income -.124 .187 .441 .507 .883 Time at Res. -.060 .112 .287 .592 .942 Father -.348 .662 .276 .600 .706 Father b/f 18 .654 1.045 .392 .531 1.923 Married 1.001 1.040 .926 .336 2.721 Bad Behav. -.075 .613 .015 .903 .928 CONSTANT -2.521 4.158 .367 .544 .080 fatherhood prior to the age of 18 was problema tic. When a cross ta bulation was run, one of the cells contained zero partic ipants (None of the participants were fathers prior to the age of 18 at wave 3 who also desisted from drug use during Waves 3 and 4). The test of the full model (all predictors) against a constant only model was significant (Chi-square = 22.04, df = 11, p < .05) indicating that the set of predictors reliably distinguished participants that di d and did not desist. Table 5-16 presents regression coefficients where none of the eleven variable re liably predicted desistance (N = 144). Desistance 2: Wave 2 Independent Variab les Predicting Desistance Waves 3-5 The final set of Desistance 2 models pred icted desistance from drug use from 1982 through 1994. A direct logist ic regression was conducted on the outcome variable of Desistance and the following Wave 2 pred ictors: Age, enrollment in School,

PAGE 154

141 Table 5-16 Desist 2: Wave 3 Independent Variables Predicting Desistance Waves 3-4 VARIABLES B S.E. Wald Sig. Exp(B) Age .168 .181 .858 .354 1.183 Employed 1.237 .633 3.814 .051 3.445 Income -.395 .204 3.753 .053 .674 Welfare Use -.356 .839 .180 .671 .701 Job Sat. .408 .363 1.266 .261 1.504 HS Degree -.710 .564 1.583 .208 .492 Time at Res. .207 .157 1.751 .186 1.230 Father .407 .583 .487 .485 1.502 Res. Father -1.697 .937 3.275 .070 .183 Married -.395 .942 .176 .675 .674 Bad Behav. -.427 .638 .447 .504 .653 CONSTANT -5.784 3.975 2.117 .146 .003 Employment, attainment of a High School Degree, Job Satisfaction, Welfare Use, Income, Time at current Residence, Fath erhood, Fatherhood before age 18, and Marital Status. The full model could not be run as the variable noting juve nile delinquency (Bad Behavior Wave 1) was problem atic. When a cross tabulati on was run, one of the cells contained zero participants (N one of the participants were juvenile delinquents who also desisted from drug use from Wave 3 through Wave 5). The test of the full model (all predictors ) against the constant only model was not significant (Chi-square = 9.66, df = 11, p > .05) indicating that th e set of Wave 2 predictors did not reliably dis tinguish between participants who did and did not desist. Table 5-17 presents regression coefficients where none of the eleven Wave 2 variables reliably predicted Desistance (N = 114).

PAGE 155

142 Table 5-17 Desist 2: Wave 2 Independent Variables Predicting Desistance Waves 3-5 VARIABLE B S.E. Wald Sig. Exp(B) Age .818 .512 2.546 .111 2.265 In School -1.132 1.187 .909 .340 .322 Employed -.002 1.111 .000 .998 .998 HS Degree .449 1.191 .142 .706 1.567 Job Sat. -.795 .482 2.724 .099 .452 Welfare Use 2.487 1.715 2.102 .147 12.025 Income -.392 .356 1.215 .270 .676 Time Res. -.048 .241 .040 .842 .953 Father -1.008 1.556 .420 .517 .365 Father b/f 18 1.362 1.636 .693 .405 3.904 Married 1.117 2.030 .303 .582 3.057 CONSTANT -15.949 10.210 2.440 .118 .000 Desistance 2: Wave 3 Independent Variab les Predicting Desistance Waves 3-5 This model was dropped as all of the pred ictors, except age, were problematic. When a cross tabulation was run, one of the cel ls contained zero participants for each of the following variables: Employment, at tainment of a High School Degree, Job Satisfaction, Welfare Use, Income, Time at current Residence, Fatherhood, Fatherhood before age 18, and Marital Status. In most cas es, there failed to be a participant who both desisted and was classified as positively holdi ng the predictor status (EX: Both employed and desistant). However, an interesting categ ory that diverged from the above statement was the predictor Fatherhood. All six men who were fathers with in this sample desisted. None of these men failed to desist fr om drug use from Wave 3 through Wave 5. Desistance 3: Wave 3 Independent Vari ables Predicting Desistance Wave 4 The remaining models in this chapter disp lay the results of models that examine “Desistance 3.” These models predicted desistance beginning in Wave 4; thereby

PAGE 156

143 offenses could be committed from the be ginning of the study up through Wave 3. The first model looked at desistance from drug use through Wave 4 (1989-1994). A direct logistic regression was conducted on the out come variable of Desistance and the following Wave 3 predictors: : Age, Employ ment Status, Attainment of a High School Diploma, Job Satisfaction, Welf are Use, Income, Time at Current Residence, Fatherhood, Residential Fatherhood, Fatherhood before ag e 18, Marital Status, and Juvenile delinquency. The test of the full model (all predictors ) against a constant only model was not significant (Chi-square = 20.30, df = 12, p > .05) i ndicating that the set of predictors did not reliably distinguish between participants who did and did not desist. Table 5-18 presents the regression coefficients Income reliably predicted Desistance (N = 140). For every one unit increase in income, th e odds of desistan ce increased by 1.75. Table 5-18 Desist 3: Wave 3 Independent Variables Predicting Desistance Wave 4 VARIABLES B S.E Wald Sig. Exp(B) Age -.075 .164 .210 .647 .927 Work -.130 .506 .066 .797 .878 Income .562 .184 9.285 .002 1.754 Welfare Use -.032 .616 .003 .959 .969 Job Sat. -.084 .223 .141 .707 .920 HS Degree -.604 .522 1.341 .247 .547 Time at Res. -.062 .111 .318 .573 .940 Father -.425 .509 .700 .403 .653 Res. Father .479 .646 .550 .458 1.615 Father b/f 18 -.210 .740 .081 .776 .810 Married .716 .649 1.216 .270 2.046 Bad Behav. -.363 .518 .490 .484 .696 CONSTANT 1.119 3.272 .117 .732 3.061

PAGE 157

144 Desistance 3: Wave 4 Independent Vari ables Predicting Desistance Wave 4 This model used variables collected in Wave 4 (1989-1990) to predict desistance from drug use from 1989 to 1992. A direct logistic regression was conducted on the outcome variable of Desistance and the follo wing Wave 4 predictors: Age, Employment, Income, Welfare Use, attainment of a Hi gh School Degree, Job Satisfaction, Time at current Residence, Fatherhood, Residential Fatherhood, Fatherhood prior to age 18, Marital Status, and J uvenile delinquency. The test of the full model (all predictors ) against a constant only model was not significant (Chi-square = 17.20, df = 12, p > .05) indicating that th e set of Wave 4 predictors did not reliably dis tinguish between participants that did and did not desist. Table 5-19 displays the regr ession coefficients where Marriage and Residential Table 5-19 Desist 3: Wave 4 Independent Variables Predicting Desistance Wave 4 VARIABLES B S.E Wald Sig. Exp(B) Age .128 .154 .693 .405 1.136 Work -.368 .514 .512 .474 .692 Income -.030 .150 .040 .842 .971 Welfare Use -1.043 .696 2.245 .134 .352 HS Degree -.555 .550 1.018 .313 .574 Job Sat. -.098 .250 .154 .695 .906 Time at Res. .068 .108 .391 .532 1.070 Father -.750 .538 1.943 .163 .472 Res. Father 1.139 .556 4.190 .041 3.124 Father b/f 18 .321 .746 .185 .667 1.378 Married 1.006 .494 4.154 .042 2.735 Bad Behav. -.465 .549 .716 .397 .628 CONSTANT -2.590 5.285 .240 .624 .075

PAGE 158

145 Fatherhood reliably predicted Desistance (N = 140). Men who were residential fathers during Wave 4 had 3.12 greater odds of desist ance. Men who were married in Wave 4 had 2.74 greater odds of desistance. Desistance 3: Wave 3 Independent Variab les Predicting Desistance Wave 4-5 Measurements of various social bonds a nd other theoretically significant variables collected during 1983 and 1984 were used to pred ict desistance from drug use after Wave 3 and through wave 5 (1989-1994). A direct logistic regression was conducted on the outcome variable of Desistance and the followi ng Wave 3 predictors : Age, Employment Status, Attainment of a Hi gh School Diploma, Job Satisfaction, Welfare Use, Income, Time at Current Residence, Fatherhood, Re sidential Fatherhood, Fa therhood before age 18, Marital Status, and Juvenile delinquency. The test of the full model (all predictors ) against the constant only model was not significant (Chi-square = 11.58, df = 12, p > .05) i ndicating that the set of predictors did not reliably distinguish between participants that did and did not desist . Table 5-20 presents the regression coeffici ents where none of the twelve Wave 3 variables correctly predicted desistance (N = 90). Desistance 3: Wave 4 Independent Variab les Predicting Desistance Wave 4-5 Measurements of various social bonds a nd other theoretically significant variables collected during 1989 and 1990 were used to pred ict desistance from drug use after wave 3 and through wave 5 (19891994). A direct logistic re gression was conducted on the outcome variable of Desistance and the following Wave 4 predictors: Age, Attainment of a High School Diploma, Job Satisfaction, Income, Time at Current Residence, Fatherhood, Residential Fatherhood, Fatherhood before age 18, Marital Status, and Juvenile delinquency. The full model could not be run, as the variab les of Employment

PAGE 159

146 Table 5-20 Desist 3: Wave 3 Independent Variables Predicting Desistance Waves 4-5 VARIABLES B S.E. Wald Sig. Exp(B) Age -.172 .261 .436 .509 .842 Employed -.418 .731 .328 .567 .658 Income .328 .276 1.419 .234 1.389 Welfare Use .005 .913 .000 .995 1.005 Job Sat. -.030 .289 .001 .917 .970 HS Degree 1.149 .821 1.956 .162 3.154 Time at Res. -.173 .165 1.103 .294 .841 Father -.135 .694 .038 .846 .874 Res. Father -.395 .907 .189 .664 .674 Father b/f 18 -.449 1.130 .158 .691 .638 Married -1.138 .946 1.445 .229 .321 Bad Behav. .762 .870 .768 .381 2.142 CONSTANT 1.410 4.930 .082 .775 4.097 and Welfare Use were problematic. When a cross tabulation was run for each variable, one of the cells contained zero participants. None of the participants who used welfare during Wave 4 desisted from drug use. Additionally, none of the men who were unemployed during Wave 4 desisted from drug use. The test of the full model (all predictors ) against the constant only model was not significant (Chi-square = 16.67, df = 10, p > .05) indicating that th e set of Wave 4 predictors did not reliably di stinguish between men who did and did not desist. Table 521 presents the regression coefficients wher e Income reliably pred icted Desistance (N = 90). For every one unit increase in inco me, a participant had 2.52 greater odds of desistance.

PAGE 160

147 Table 5-21 Desist 3: Wave 4 Independent Variables Predicting Desistance Waves 4-5 VARIABLES B S.E. Wald Sig. Exp(B) Age .223 .238 .876 .349 1.250 Income .926 .460 4.049 .044 2.524 HS Degree .418 .748 .312 .577 1.519 Job Sat. -.822 .436 3.565 .059 .439 Time at Res. .373 .230 2.628 .105 1.452 Father -.698 .889 .616 .433 .498 Res. Father -.052 .918 .003 .955 .950 Father b/f 18 .090 1.139 .006 .937 1.094 Married -.703 .835 .709 .400 .495 Bad Behav. .012 .967 .000 .990 1.012 CONSTANT -12.759 8.480 2.264 .132 .000 .

PAGE 161

148 CHAPTER 6 RESULTS: ARRESTS This chapter provides an examination of pa rticipant arrests over the course of the study. A myriad of statistical models were run utilizing a vari ety of theoretically supported variables to predict arrests in each wa ve. Each model considered arrests within a specific time frame, thereby displaying the va rying influences that diverse social bonds imparted upon deviant behaviors that may have lead to arrests. The first section of this chapter examin es the various cross-sectional and lagged models that were run. The following section displays the models th at sought to predict desistance; conceptualized as a cessation of be ing arrested. Prior to each section an in depth discussion of what follows is provided. Cross-sectional and Lagged Models The following models examined arrest as a dichotomous variable wherein within a specific set of years participants either were or were not arrested. Men who were arrested were coded as ” for the dependent variab le, while men who were not arrested were coded as .” Certain models were acco mpanied by additional models that sought to examine the strength of the social bonds of marriage and father hood. Two measures of the strength of the fatherhood bond were availa ble and utilized for Wave 3 independent variables predicting Wave 3 dependent variab les. One measure of marital success was utilized in the following models: Wave 3 independent variables predicting Wave 3 dependent variables, and Wave 5 independe nt variables predicting Wave 5 dependent

PAGE 162

149 variables. Within all of these models, onl y the men who held thes e bonds (fathers or husbands) were selected as cases for analysis. Wave 1 Independent Variables Predic ting Wave 2 Dependent Variables Wave 2 independent variables were used to predict being arrested in Wave 2. More specifically, they predicted be ing arrested during the years of 1968 to 1974. During these years participants were within the ages of 12 and 22. A direct logistic regression was conducted on Wave 2 Arrests as the outcome va riable and Wave 1 predictors: Age, being in school, family income, welfare use, GPA, length of time at current residence, presence of a father in the home, and a measurement of juvenile delinquency. The test of the full model (all predictors ) against a constant only model was not significant (Chi-square = 9.21, df=8, p>.05) indica ting that the set of Wave 1 predictors did not reliably distinguish between participan ts that were arrested and those that were not. Table 6-1 presents the regression co efficients where none of the eight Wave 1 variables correctly predicte d Wave 2 Arrests (N = 277). Table 6-1 Wave 1 Independent Vari ables Predicting Wave 2 Arrests VARIABLE B S.E Wald Sig. Exp(B) Age .027 .185 .022 .882 1.028 In School 5.728 17.957 .102 .750 307.431 Income -.003 .144 .001 .981 .997 Welfare Use .881 .654 1.814 .178 2.414 GPA -.137 .406 .114 .736 .872 Time at Res. -.049 .157 .098 .754 .952 Father in home -.907 .726 1.559 .212 .404 Bad Behav. .289 .656 1.94 .660 1.335 CONSTANT -8.712 18.234 .228 .633 .000

PAGE 163

150 Wave 2 Independent Variables Predic ting Wave 2 Dependent Variables Wave 2 independent variables were used to predict being arrested in Wave 2. More specifically, they predicted arrests during the years of 1975 to 1976. During these years participants were within the ages of 18 and 23. A direct logistic regression was conducted on the outcome variable of Wa ve 2 Arrests and the following Wave 2 predictors: Age, School enrollment, Employ ment Status, Attainment of a High School Diploma, Job Satisfaction, Welf are Use, Income, Time at Current Residence, Fatherhood, Residential Fatherhood, Fatherhood prior to Age 18, Marital Status, Juvenile delinquency, and Prior Arrests (No measure of previous drug use was available for this model.) The test of the full model (all predictors ) against a constant only model was not significant (Chi-square = 13.91, df=13, p>.05) indicating that th e set of Wave 2 predictors did not reliably dis tinguish between participants th at were arrested and those that were not. Table 6-2 presents regre ssion coefficients where none of the Wave 2 variables correctly predicte d Wave 2 Arrests (N = 277). Wave 2 Independent Variables Predic ting Wave 3 Dependent Variables Wave 2 independent variables were used to predict being arrested in Wave 3. More specifically, they predict a rrests during the years of 1977 to 1982. During these years participants were within the ages of 19 and 30. A direct logistic regression was conducted on the outcome variable of Wa ve 3 Arrests and the following Wave 2 predictors: Age, School enrollment, Employ ment Status, Attainment of a High School Diploma, Job Satisfaction, Welf are Use, Income, Time at Current Residence, Fatherhood, Residential Fatherhood, Fatherhood prior to Age 18, Marital Status, Juvenile delinquency, Prior Drug Use, and Prior Arrests.

PAGE 164

151 Table 6-2 Wave 2 Independent Vari ables Predicting Wave 2 Arrests VARIABLE B S.E. Wald Sig. Exp(B) Age .100 .305 .108 .742 1.105 In School .439 1.017 .186 .666 1.552 Employed -.799 .909 .772 .380 .450 HS Degree -.593 .891 .443 .506 .553 Job Satisfaction .910 .548 2.760 .097 2.485 Welfare Use -6.874 28.774 .057 .811 .001 Income -.032 .273 .013 .908 .969 Time at Res. -.003 .204 .000 .989 .997 Father 1.316 .979 1.807 .179 3.728 Fath. b/f 18 -.103 1.123 .008 .927 .902 Married -6.068 48.197 .016 .900 .002 Bad Behav. .552 .885 .417 .518 1.737 Prior Arrests .998 1.263 .625 .429 2.714 CONSTANT -8.519 6.777 1.580 .209 .000 The test of the full model (all predictors ) against a constant only model was not significant (Chi-square = 13.25, df = 13, p >.05) indicating that the set of Wave 2 predictors did not reliably dis tinguish between participants th at were arrested and those that were not. Table 6-3 pres ents regression coefficients wh ere none of the thirteen Wave 2 variables correctly predicted Wave 3 Arrests (N = 210). Wave 3 Independent Variables Predic ting Wave 3 Dependent Variables Wave 3 independent variables were used to predict being arrested in Wave 3. More specifically, they predict a rrests during the years of 1983 to 1984. During these years participants were within the ages of 26 and 31. A direct logistic regression was conducted on Wave 3 Arrests as the outcome variable and Wave 3 predictors: Age,

PAGE 165

152 Table 6-3 Wave 2 Independent Vari ables Predicting Wave 3 Arrests VARIABLE B S.E Wald Sig. Exp(B) Age .007 .137 .003 .957 1.007 In School .258 .399 .418 .518 1.295 Employed -.645 .378 2.919 .088 .524 HS Degree -.352 .392 .805 .370 .703 Job Sat. -.058 .170 .116 .733 .944 Welfare Use -1.089 .672 2.625 .105 .337 Income .005 .129 .001 .970 1.005 Time at Res. .076 .085 .788 .375 1.079 Father .103 .496 .043 .836 1.108 Father b/f 18 .441 .754 .342 .559 1.554 Married -5.354 16.347 .107 .743 .005 Prior Drug Use .178 .340 .275 .600 1.195 Prior Arrests -.865 .836 1.069 .301 .412 CONSTANT -1.093 3.055 .128 .721 .335 employment status, income, welfare use, j ob satisfaction, attainme nt of a high school degree, time at current residence, fath erhood, marital status, residential fatherhood, fatherhood prior to age 18, prior drug use, and prior arrests . The test of the full model (all predicto rs) against the constant only model was significant (Chi-square = 38.82, df = 13, p < .001) indicating that the set of predictors reliably distinguished between participants th at were and were not arrested in Wave 3. Table 6-4 presents the regression coeffici ents where Employment, Income, Fatherhood, and Prior Arrests reliably predicted Wave 3 Arrests. The odds of arrest in Wave 3 decreased by 94.7% for men who were employe d. For every one unit increase in income the odds of arrest increased by 1.84. The odds of arrest for men who were fathers in

PAGE 166

153 Wave 3 increased by 7.54. Finally the odds of arrest during 1983 to 1984 for men who were previously arrest ed decreased by 84.2%. Table 6-4 Wave 3 Independent Vari ables Predicting Wave 3 Arrests VARIABLE B S.E Wald Sig. Exp(B) Age -.254 .265 .919 .338 .776 Employed -2.930 .941 9.701 .002 .053 Income .608 .292 4.346 .037 1.837 Welfare Use .167 .801 .044 .835 1.182 Job Sat. -.192 .306 .392 .531 .826 HS Degree -1.255 .651 3.721 .054 .285 Time at Res. -.052 .180 .082 .774 .950 Father 2.020 .844 5.730 .017 7.537 Married .361 1.048 .119 .730 1.435 Res. Father -.556 .902 .380 .537 .573 Father b/f 18 1.001 .892 1.259 .262 2.722 Prior Drug Use .572 .735 .606 .436 1.771 Prior Arrests -1.843 .880 4.383 .036 .158 CONSTANT 1.215 5.385 .051 .821 3.372 Strength of Social Bonds: Fatherhood The following model examined the strength of fathers’ social bonds with their children. The measurements of strength were conceptualized in two distinct manners. First, men were asked to rate themselves as fathers. Second, men were asked how they feel about their relationship with their children. For bot h of these variables, higher numbers indicated better situations (An in-d epth description of the coding of these variables can be found in Chapter 4: Methods ). A model was run incorporating both of these variables. The sample of this mode l was to men who were fathers by Wave 3.

PAGE 167

154 A direct logistic regre ssion was conducted on the outco me variable of Wave 3 Arrests and the following Wave 3 predictors: Age, employme nt status, income, welfare use, job satisfaction, attainme nt of a high school diploma, time at current residence, marital status, residential fa therhood, fatherhood prior to age 1 8, prior arrests, prior drug use, a rating of one’s success as a parent, and an evaluation of one’s relationship with their child(ren). The test of the full model (all predictors) against a constant only model was significant (Chi-square = 27.00, df = 14, p < .05) indicating that the set of predictors reliably distinguished between fathers who we re and were not arrested during Wave 3. Table 6-5 presents regression coefficients where Employment, Income, and Prior Arrests reliably predicted Wave 3 Arrests among fath ers (N = 106). As fathers tended to be employed during Wave 3, the odds that they were arrested during Wave 3 decreased by 94.3%. For every one unit increase in income, the odds that a participant was arrested in Wave 3 increased by 2.07. Finally, as particip ants tended to be arrested in the prior Wave 3, the odds that they would be a rrested during Wave 3 decreased by 86.1%. Strength of Social Bonds: Marriage There were not enough participants to successfully create this model. Wave 3 Independent Variables Predic ting Wave 4 Dependent Variables There are no measurements of arrest for Wave 4. Wave 4 Independent Variables Predic ting Wave 4 Dependent Variables There are no measurements of arrest for Wave 4. Wave 4 Independent Variables Predic ting Wave 5 Dependent Variables Wave 4 independent variables were used to predict being arrest ed during Wave 5. More specifically, they pred icted arrests during the years of 1991 to 1992. During these

PAGE 168

155 Table 6-5 Wave 3 Independent Variables Predicting Wave 3 Arrests among Fathers VARIABLE B S.E. Wald Sig. Exp(B) Age -.595 .342 3.031 .082 .551 Employed -2.868 1.091 6.910 .009 .057 Income .728 .369 3.894 .048 2.070 Welfare Use 1.244 1.044 1.422 .233 3.471 Job Satisfaction .114 .390 .086 .769 1.121 HS Degree -1.077 .796 1.831 .176 .340 Time at Res. -.161 .237 .460 .497 .851 Married .839 1.165 .519 .471 2.314 Res. Father -1.007 1.049 .923 .337 .365 Father b/f 18 1.046 .960 1.186 .276 2.845 Prior Arrests -1.972 .973 4.108 .043 .139 Prior Drug Use 1.300 .996 1.703 .192 3.669 Rate Self as parent -.140 .579 .059 .808 .869 Relationship w/ Child .196 .545 .130 .718 1.217 CONSTANT .8771 7.384 1.411 .235 6442.089 years participants were within the ages of 33 and 40. A direct logistic regression was conducted on Wave 5 Arrests as the outcome variable and Wave 4 predictors: Age, employment status, income, welfare use, j ob satisfaction, attainme nt of a high school degree, time at current residence, fath erhood, marital status, residential fatherhood, fatherhood prior to age 18, prior drug use, and prior arrests. The test of the full model (all predictors ) against a constant only model was not significant (Chi-square = 7.42, df = 13, p > .05) . Table 6-6 presents the regression

PAGE 169

156 coefficients where none of the Wave 4 variab les correctly predicted Wave 5 Arrests (N = 168). Table 6-6 Wave 4 Independent Vari ables Predicting Wave 5 Arrests VARIABLE B S.E. Wald Sig. Exp(B) Age -.051 .211 .059 .809 .950 Employed -.393 .725 .294 .588 .675 Income .474 .394 1.448 .229 1.606 Welfare -.151 1.300 .013 .908 .860 Job Sat. -.155 .389 .159 .690 .856 HS Degree -.509 .667 .565 .452 .601 Time at Res. -.148 .167 .785 .376 .863 Father -.351 .759 .214 .544 .704 Married .123 .821 .022 .881 1.131 Res. Father -.162 .940 .030 .864 .851 Father b/f 18 -.637 1.211 .277 .599 .529 Prior Drug Use -1.509 1.143 1.741 .187 .221 Prior Arrests 1.058 .962 1.210 .271 2.880 CONSTANT -.063 7.463 .000 .993 .939 Wave 5 Independent Variables Predic ting Wave 5 Dependent Variables Wave 5 independent variables were used to predict being arrest ed during Wave 5. More specifically, they pred icted arrests during the years of 1993 to 1994. During these years participants were within the ages of 35 and 41. A direct logistic regression was conducted on Wave 5 Arrests as the outcome variable and Wave 5 predictors: Age, employment status, income, welfare use, j ob satisfaction, attainme nt of a high school degree, time at current residence, fatherhood, marital status, residential fatherhood, prior drug use, and prior arrests.

PAGE 170

157 The test of the full model (all predictors) against a constant only model was significant (Chi-square = 23.56, df = 12, p < .05) indicating that the set of predictors reliably distinguished between participant that were and were not ar rested during Wave 5. Table 6-7 presents regression coefficients where Fatherhood and re sidential Fatherhood reliably predicted Wave 5 Arrests (N = 168). The odds of men who were fathers in Wave 5 of being arrested were 73.4% lower than men who were not fathers. Likewise, men who were fathers and lived with their children during Wave 5 ha d odds of arrest that were 74.7% lower than men who were not cl assified as residential fathers. Table 6-7 Wave 5 Independent Vari ables Predicting Wave 5 Arrests VARIABLES B S.E Wald Sig. Exp(B) Age -.179 .155 1.339 .247 .836 Employed -1.030 .553 3.473 .062 .357 Income .026 .130 .039 .844 1.026 Welfare Use .238 .538 .196 .658 1.269 HS Degree 1.346 .721 3.483 .062 3.841 Job Sat. -.332 .249 1.769 .184 .718 Time at Res. .089 .122 .528 .468 1.093 Father -1.326 .580 5.235 .022 .266 Res. Father -1.375 .605 5.166 .023 .253 Married .106 .533 .039 .843 1.111 Prior Drug Use -1.784 1.114 2.566 .109 .168 Prior Arrests 1.470 1.051 1.957 .162 4.351 CONSTANT 7.057 5.804 1.478 .224 1160.822 Strength of Social Bonds: Marriage The next model examined the strength of marital bonds. Men were asked how satisfied they were with their relationship w ith their spouse. As w ith the prior measures

PAGE 171

158 of the strength of social bonds , higher numbers were measures of higher strength or in this case satisfaction (An in-d epth description of the coding of this variable can be found in Chapter 4: Methods). Only married men in Wave 5 were included in this model. A direct logistic regression wa s conducted on the outcome variab le of Wave 5 Arrests and the following Wave 5 predictors: Age, em ployment status, income, job satisfaction, welfare use, attainment of a high school dipl oma, time at current residence, fatherhood, residential fatherhood, prior arrests, and a ra ting of one’s marital satisfaction. The full model could not be run as the prior drug us e variable was problematic. When a cross tabulation was run, one of the cells contained zero participants (None of the men who had previously used drugs were arrested during Wave 5). The test of the full model (all predictors ) against a constant only model was not significant (Chi-square = 11.95, df = 11, p > .05) i ndicating that the set of predictors did not reliably distinguish between married me n who were and were not arrested during Wave 5. Table 6-8 presents regression coefficients where Residential Fatherhood reliably predicted Wave 5 Arrests (N = 69). Married men who were residential fathers during Wave 5 were 84.2% less likely to get arrested during Wave 5 than men who were not classified as residential fathers. Desistance Models The next set of models examined desi stance. Within the following models desistance was conceptualized as not being arre sted within a specific time frame. There were two distinct time frames around which thes e models of desistance were constructed. The two larger sets of models were referred to as “Desistance 2” and “Desistance 3.” The number in these two names referred to th e last wave at which the individual was found to have engaged in the deviant act unde r consideration. For this chapter, the

PAGE 172

159 Table 6-8 Wave 5 Independent Variables Pr edicting Wave 5 Arrest s among Married Men VARIABLE B S.E. Wald Sig. Exp(B) Age -.103 .296 .122 .727 .902 Employed -3.158 1.617 3.812 .051 .043 Income .496 .389 1.454 .228 1.598 Welfare Use -.620 1.608 .148 .700 .538 Job Satisfaction -.041 .509 .007 .936 .960 HS Degree .474 1.498 .100 .752 1.606 Time at Res. .493 .369 1.791 .181 1.638 Fatherhood -1.680 1.536 1.196 .274 .186 Res. Father -1.846 .944 3.826 .050 .158 Prior Arrests 1.222 1.536 .633 .426 3.393 Marital Satisfaction -.025 .432 .003 .955 .976 CONSTANT 2.023 12.418 .027 .871 7.560 number referred to the last wave within which the participant was arrested. For Desistance 2, men who were arrested a nytime prior to wave 3 (1968-1982) yet not arrested for a specified set of years was categorized as desistant (Coded as ”). Likewise, Desistance 3 considered men who we re arrested anytime up until the end of Wave 3 (1968-1984) and who were not arrested th rough specific sets of years as desistant (Coded as ”). Desistance 2: Wave 2 Independent Vari ables Predicting Desistance Wave 3 This model predicted desistance from a rrests from 1983 to 1984. The independent variables used to predict de sistance were collected during Wave 2 (1975-1976). A direct logistic regression was conducted on the out come variable of Desistance and the following Wave 2 predictors: Age, School enro llment, Employment Status, Attainment of

PAGE 173

160 a High School Diploma, Job Satisfaction, We lfare Use, Income, Time at Current Residence, Fatherhood, Fatherhood prior to Age 18, Marital Status, and Juvenile delinquency. The test of the full model (all predictors ) against the constant only model was not significant (Chi-square = 14.19, df = 12, p > .05) indicating that th e set of Wave 2 predictors did not reliably dis tinguish between participants th at did and did not desist. Table 6-9 presents regression coefficients where none of the twelve Wave 2 variables reliably predicted Desistance (N = 172). Table 6-9 Desist 2: Wave 2 Independen t Variables Predicting Desistance Wave 3 VARIABLES B S.E. Wald Sig. Exp(B) Age -.002 /141 .000 .987 .998 In School .202 .402 .254 .615 1.224 Employed -.569 .365 2.433 .119 .566 HS Degree -.593 .386 2.359 .125 .552 Job Sat. .113 .166 .467 .495 1.120 Welfare Use -.952 .617 2.381 .123 .386 Income -.038 .137 .077 .781 .963 Time at Res. .028 .080 .121 .728 1.028 Father -.194 .467 .154 .695 .832 Father b/f 18 .250 .753 .110 .741 1.283 Married -.408 1.194 .117 .732 .665 Bad Behav. .725 .440 2.720 .099 2.066 CONSTANT -.149 3.118 .002 .962 .862 Desistance 2: Wave 3 Independent Vari ables Predicting Desistance Wave 3 The following model predicted desistance du ring the same time period as in the previous model (1983-1984). However, this model used Wave 3 independent variables (collected in 1983 and 1984) to predict desistance from arre sts. A direct logistic

PAGE 174

161 regression was conducted on the outcome variab le of Desistance and the following Wave 2 predictors: Age, Employment Status, A ttainment of a High School Diploma, Job Satisfaction, Welfare Use, Income, Time at Current Residence, Fatherhood, Residential Fatherhood, Fatherhood prior to Age 18, Mari tal Status, and Juvenile delinquency. The test of the full model (all predicto rs) against the constant only model was significant (Chi-square = 25.94, df = 12, p < .05) indicating that th e set of Wave 3 predictors reliably distinguished between participants that di d and did not desist. Table 6-10 displays the regression coefficients wher e none of the twelve variables reliably predicted Desistance (N = 172). Table 6-10 Desist 2: Wave 3 Independent Variables Predicting Desistance Wave 3 VARIABLES B S.E. Wald Sig. Exp(B) Age -.013 .132 .010 .919 .987 Employed .110 .415 .071 .791 1.116 Income -.233 .129 3.237 .072 .793 Welfare Use -.238 .517 .211 .646 .789 Job Sat. -.115 .181 .406 .524 .891 HS Degree -.678 .399 2.883 .089 .508 Time at Res. -.134 .093 2.087 .149 .874 Father .575 .433 1.760 .185 1.777 Res. Father -.447 .523 .732 .392 .639 Father b/f 18 1.207 .687 3.088 .079 3.343 Married -.402 .525 .585 .444 .669 Bad Behav. .509 .458 1.240 .266 1.664 CONSTANT 1.839 2.656 .479 .489 6.289 Desistance 2: Wave 2 Independent Variable s Predicting Desistance in Waves 3-5 The first model in this section used Wa ve 2 independent variables to predict desistance from arrests. These measures were collected in 1975 and 1976. Men who

PAGE 175

162 desisted were not arrested between the years of 1983 and 1984, as well as 1991 to 1994. A direct logistic regression was conducted on the outcome vari able of Desistance and the following Wave 2 predictors: Age, enrollmen t in School, Employment, attainment of a High School Degree, Job Satisfaction, Welfare Use, Income, Time at current Residence, Fatherhood, Fatherhood before age 18, and Juve nile Delinquency. The full model could not be run as Marital Status was problematic . When a cross tabulation was run, one of the cells contained zero participants. The cro ss tabulation revealed that none of the men who were married during Wave 2 desisted from being arrested from Wave 3 through Wave 5. The test of the full model (all predictors ) against the constant only model was not significant (Chi-square = 13.24, df = 11, p > .05) indicating that th e set of Wave 2 predictors did not reliably dis tinguish between participants th at did and did not desist. Table 6-11 presents regression coefficien ts where Employment reliably predicted Desistance (N = 114). Men who were empl oyed during Wave 2 had odds of desistance 1.09 greater than men who were unemployed during Wave 2. Desistance 2: Wave 3 Independent Variable s Predicting Desistance in Waves 3-5 This model utilized Wave 3 independent variables to predict desistance from arrests in 1983, 1984, and between 1991 and 1994. The i ndependent variables were collected during 1983 and 1984. A direct logistic re gression was conducted on the outcome variable of Desistance and the following Wave 3 predictors : Age, Employment, Income, Welfare Use, Job Satisfaction, attainment of a High School degree, Time at current Residence, Fatherhood, Residential Fath erhood, Marital Status, and Juvenile Delinquency. The full model could not be run as the variable Father before Age 18 was problematic. When a cross tabulation was run, one of the cells contained zero

PAGE 176

163 Table 6-11 Desist 2: Wave 2 Independent Variables Predicting Desistance Waves 3-5 VARIABLE B S.E. Wald Sig. Exp(B) Age .102 .202 .256 .613 1.107 In School .087 .560 .024 .876 1.091 Employed -1.034 .522 3.921 .048 .356 HS Degree -.590 .550 1.151 .283 .554 Job Sat. .2898 .248 1.364 .243 1.336 Welfare Use .131 .969 .018 .892 1.140 Income -.076 .167 .206 .650 .927 Time at Res. .100 .116 .742 .389 1.105 Father .622 .696 .799 .371 1.863 Father b/f 18 -.362 .864 .175 .676 .697 Bad Behav. -.041 .616 .005 .946 .959 CONSTANT -3.320 4.173 .633 .426 .036 participants. The cross tabulation revealed th at none of the men who were noted as being fathers before age 18 during Wave 3 failed to desist from being arrested from Wave 3 through Wave 5 (All men who fathered a child before age 18 desisted). The test of the full model (all predicto rs) against the constant only model was significant (Chi-square = 22.62, df = 11, p < .05) indicating that th e set of Wave 3 predictors reliably distinguished between participant that did and did not desist. Table 612 presents the regression coefficients wher e Employment and Time at current Residence reliably predicted Desistance (N = 114). As men were employed during Wave 3, the odds of desistance decreased by 70.5%. Fo r every one unit incr ease in Time at Residence, the odds of de sistance decreased by 29.6%. Desistance 3: Wave 3 Independent Variable s Predicting Desistance in Waves 4-5 The first “Desist 3” arrest model used Wave 3 independent variables to predict desistance from 1991 to 1994. The independent variables were collected in 1983 and

PAGE 177

164 1984. A direct logistic regression was conduc ted on the outcome variable of Desistance and the following Wave 3 predictors: Age, Employment Status, Attainment of a High School Diploma, Job Satisfaction, Welfare Us e, Income, Time at Current Residence, Fatherhood, Residential Fatherhood, Fatherhood before age 18, Marital Status, and Juvenile delinquency. Table 6-12 Desist 2: Wave 3 Independent Variables Predicting Desistance Waves 3-5 VARIABLE B S.E. Wald Sig. Exp(B) Age -.021 .198 .011 .915 .979 Employed -1.222 .610 4.013 .045 .295 Income .054 .266 .058 .810 1.056 Welfare Use 1.104 .750 2.167 .141 3.015 Job Sat. .114 .315 .132 .717 1.121 HS Degree -.630 .599 1.107 .293 .533 Time at Res. -.350 .151 5.387 .020 .704 Father -.806 .627 1.654 .198 .447 Res. Father .977 .723 1.823 .177 2.656 Married -1.095 .868 1.594 .207 .334 Bad Behav. .889 .630 2.038 .153 2.458 CONSTANT 1.732 3.770 .211 .646 5.650 The test of the full model (all predictors ) against a constant only model was not significant (Chi-square = 18.82, df = 12, p >.05) i ndicating that the set of predictors did not reliably distinguish between participants that did and did not desist . Table 6-13 presents the regression coefficients wher e attainment of a High School Degree and Fatherhood before age 18 reliably predicted Desistance (N = 90). Men who had attained a high school diploma by Wave 3 held odds of desistance that were 75.1% lower than men who had not received a diploma. Howeve r, men who had fathered a child before the

PAGE 178

165 age of 18 were more likely to desist, in th at their odds of desi stance were 22.66 greater than men who were not classified as underage fathers by Wave 3. Table 6-13 Desist 3: Wave 3 Independent Variables Predicting Desistance Waves 3-5 VARIABLES B S.E. Wald Sig. Exp(B) Age .036 .256 .020 .889 1.037 Employed -1.015 .725 1.958 .162 .362 Income -.078 .243 .103 .748 .925 Welfare Use -.295 .816 .131 .718 .744 Job Sat. .350 .320 1.193 .275 1.419 HS Degree -1.392 .689 4.082 .043 .249 Time at Res. -.341 .178 3.679 .055 .711 Father -.132 .695 .036 .849 .876 Res. Father -.126 .867 .021 .884 .881 Father b/f 18 3.121 1.378 5.127 .024 22.662 Married -1.440 .902 2.548 .110 .237 Bad Behav. .830 .879 .893 .345 2.294 CONSTANT 1.326 4924 .073 .788 3.767 Desistance 3: Wave 4 Independent Variab les Predicting Desistance in Waves 4-5 This model employed Wave 4 independent variables to predic t desistance from arrests between 1991 and 1994. The independent variables were collected during 1989 and 1990. A direct logistic regression was conducted on the outcome variable of Desistance and the following Wave 4 predicto rs: Age, Employment Status, Attainment of a High School Diploma, Job Satisfaction, Welfare Use, Income, Time at Current Residence, Fatherhood, Residential Fath erhood, Fatherhood before age 18, Marital Status, and Juvenile delinquency.

PAGE 179

166 The test of the full model (all predictors ) against the constant only model was not significant (Chi-square = 18.82, df = 12, p > .05) indicating that th e set of Wave 4 predictors did not reliably distinguish between men who did and did not desist. Table 6-14 presents the regr ession coefficients where none of the twelve variables reliably predicted Desistance (N = 90) Table 6-14 Desist 3: Wave 4 Independent Variables Predicting Desistance Waves 3-5 VARIABLES B S.E. Wald Sig. Exp(B) Age .156 .207 .566 .452 1.169 Employed .071 .726 .010 .922 1.074 Income -.432 .280 2.384 .123 .649 Welfare Use .704 1.055 .445 .505 2.022 HS Degree -.407 .611 .445 .505 .666 Job Sat. .387 .406 .906 .341 1.472 Time at Res. .135 .177 .587 .444 1.145 Father .303 .775 .153 .696 1.354 Res. Father .369 .852 .188 .664 1.447 Father b/f 18 .827 1.032 .642 .423 2.286 Married -1.050 .775 1.838 .175 .350 Bad Behav. 1.184 .836 2.005 .157 3.266 CONSTANT -5.511 7.282 .573 .449 .004 .

PAGE 180

167 CHAPTER 7 RESULTS: INCARCERATION This chapter examines the incarceration of participants over the course of the study. Statistical models were run utilizing a variety of theore tically supported variables to predict incarceration in each wave. These models predicte d incarceration within various years. This provided evidence of the varyi ng effects of social bonds through the lifecourse. The first section of this chapter examin es the various cross-sectional and lagged models that were run. The following section displays the models th at sought to predict desistance conceptualized as a cessation of bei ng incarcerated. Prior to each section an in depth discussion of what follows is provided. Cross-sectional and Lagged Models This section presents a series of models that predicted incarceration in the same wave as well as the subsequent wave. Each model examined incarceration during specific years. The years are in chronol ogical order and do not overlap in order to present the ways that social bonds influence deviant behavi ors over the life-course. Men who were incarcerated during the years of inte rest were coded as ,” while all others were coded as .” Some of the waves in cluded specific measures of the strength of social bonds, such as marital satisfaction and the perceived strength of the father-child relationship. Two measures of the streng th of the fatherhood bond were available and utilized for the following models: Wave 3 independent variables predicting Wave 3

PAGE 181

168 dependent variables, and Wave 3 independe nt variables predicting Wave 4 dependent variables. Wave 1 Independent Variables Predic ting Wave 2 Dependent Variables Wave 1 independent variables were used to predict being incarcerated in Wave 2. More specifically, they pred icted incarceration during th e years of 1968 to 1974. During these years participants were within the ages of 12 and 22. A direct logistic regression was conducted on Wave 2 Incarceration as the outcome variable and Wave 1 predictors: Age, being in school, family income, welf are use, GPA, length of time at current residence, presence of a father in the home, and a measurem ent of juvenile delinquency. The test of the full model (all predictors ) against a constant only model was not significant (Chi-square = 10.09, df = 8, p > .05) indicating that the set of Wave 1 predictors did not reliably di stinguish between participan ts who were and were not incarcerated during Wave 2. Table 7-1 presen ts the regression coefficients where Age reliably predicted Wave 2 Incarceration (N = 277). For every one unit increase in Age, a participant had 1.68 greater odds of being incarcerated during Wave 2. Wave 2 Independent Variables Predic ting Wave 2 Dependent Variables Wave 2 independent variables were used to predict being incarcerated in Wave 2 (1975 to 1976). During these years participan ts were within the ages of 18 and 23. A direct logistic regression wa s conducted on the outcome variable of Wave 2 Incarceration and the following Wave 2 predictors: Age, School enrollment, Employment Status, Attainment of a High School Diploma, Job Sa tisfaction, Welfare Use, Income, Time at Current Residence, Fatherhood, Residen tial Fatherhood, Fatherhood prior to Age 18, Marital Status, Juvenile delinquency, and Prio r Arrests (No measure of previous drug use was available for this model.)

PAGE 182

169 Table 7-1 Wave 1 Independent Variab les Predicting Wave 2 Incarceration VARIABLE B S.E Wald Sig. Exp(B) Age .519 .256 4.111 .043 1.680 In School 6.998 28.799 .059 .808 1094.241 Income -.087 .231 .144 .705 .916 Welfare Use .587 .908 .419 .518 1.799 GPA -.137 .562 .060 .807 .872 Time at Res. -.201 .230 .761 .383 .818 Father in home -1.389 1.192 1.358 .244 .249 Bad Behav. -.353 .970 .132 .716 .703 CONSTANT -16.191 29.125 .309 .578 .000 The test of the full model (all predictors ) against a constant only model was not significant (Chi-square = 13.94, df = 13, p > .05) indicating that th e set of Wave 1 predictors did not reliably di stinguish between participan ts who were and were not incarcerated. Table 7-2 presents regression coefficients wher e none of the thirteen Wave 2 variables correctly predicted Wave 2 Incarceration (N = 277). Wave 2 Independent Variables Predic ting Wave 3 Dependent Variables Wave 2 independent variables were used to predict being incarcerated in Wave 3. More specifically, they pred icted incarceration during th e years of 1977 to 1982. During these years participants were w ithin the ages of 19 and 30. A direct logistic regression was conducted on the outcome variable of Wa ve 3 Incarceration and the following Wave 2 predictors: Age, School enrollment, Employ ment Status, Attainment of a High School Diploma, Job Satisfaction, Welf are Use, Income, Time at Current Residence, Fatherhood, Residential Fatherhood, Fatherhood prior to Ag e 18, Marital Status, Prior Drug Use, and Prior Arrests.

PAGE 183

170 Table 7-2 Wave 2 Independent Variab les Predicting Wave 2 Incarceration VARIABLE B S.E. Wald Sig. Exp(B) Age -.164 .302 .297 .586 .848 In School .219 1.022 .046 .830 1.245 Employed -.835 .903 .855 .355 .434 HS Degree -.445 .856 .270 .604 .641 Job Satisfaction 1.192 .617 3.737 .053 3.295 Welfare Use -7.905 46.477 .029 .865 .000 Income .047 .275 .029 .865 1.048 Time at Res. -.043 .202 .046 .831 .958 Father 1.247 .946 1.736 .188 3.479 Fath. b/f 18 -.800 1.307 .375 .540 .449 Married -7.230 78.715 .008 .927 .001 Bad Behav. .257 .908 .080 .777 1.293 Prior Arrests -7.392 58.695 .016 .900 .001 CONSTANT -3.917 6.619 .350 .554 .020 The test of the full model (all predictors) against a constant only model was significant (Chi-square = 22.34, df = 13, p = .05) indicating that th e set of Wave 2 predictors reliably distingui shed between participants who were and were not incarcerated. Table 7-3 presents the regressi on coefficients where Employment reliably predicted Wave 3 Incarceration (N = 210). As participants were employed during Wave 2, the odds that they would be incar cerated during Wave 3 decreased by 83.15%. Wave 3 Independent Variables Predic ting Wave 3 Dependent Variables Wave 3 independent variables were used to predict being incarcerated in Wave 3. More specifically, they pred icted incarceration during th e years of 1983 to 1984. During these years participants were w ithin the ages of 26 and 31. A direct logistic regression

PAGE 184

171 Table 7-3 Wave 2 Independent Variab les Predicting Wave 3 Incarceration VARIABLE B S.E Wald Sig. Exp(B) Age .036 .172 .045 .832 1.037 In School .140 .520 .073 .787 1.151 Employed -1.777 .594 8.942 .003 .169 HS Degree -.381 .506 .569 .451 .683 Job Sat. .319 .237 1.813 .178 1.375 Welfare Use -596 .705 .714 .398 .551 Income -.046 .158 .086 .769 .955 Time at Res. .046 .108 .184 .668 1.048 Father .509 .570 .798 .372 1.664 Father b/f 18 -.788 .993 .629 .428 .455 Married -4.358 16.349 .071 .790 .013 Prior Drug Use .370 .428 .745 .388 1.447 Prior Arrests -.490 .871 .317 .574 .612 CONSTANT -2.919 3.906 .559 .455 .054 was conducted on the outcome variable of Wa ve 3 Incarceration and the following Wave 2 predictors: Age, Employment Status, A ttainment of a High School Diploma, Job Satisfaction, Welfare Use, Income, Time at Current Residence, Fatherhood, Residential Fatherhood, Fatherhood prior to Age 18, Marita l Status, Prior Drug Use, and Prior Arrests. The test of the full model (all predictors) against a constant only model was significant (Chi-square = 26.50, df = 13, p < .05) indicating that th e set of Wave 3 predictors reliably distingui shed between participants who were and were not incarcerated. Table 7-4 presents the regressi on coefficients where attainment of a High School Degree reliably predicted Wave 3 Inca rceration (N = 210). The odds of Wave 3

PAGE 185

172 Incarceration were 78.4% lower for men who had attained a high school degree by Wave 3. Table 7-4 Wave 3 Independent Variab les Predicting Wave 3 Incarceration VARIABLE B S.E Wald Sig. Exp(B) Age -.052 .283 .034 .854 .949 Employed -1.611 .910 3.137 .077 .200 Income .382 .287 1.766 .184 1.465 Welfare Use 1.189 .848 1.946 .161 3.283 Job Sat. -.178 .350 .258 .611 .837 HS Degree -1.535 .733 4.378 .036 .216 Time at Res. .160 .211 .577 .447 1.174 Father 1.574 .953 2.731 .098 4.827 Married .897 1.095 .670 .413 2.451 Res. Father -.354 .924 .146 .702 .702 Father b/f 18 1.000 .939 1.135 .287 2.719 Prior Drug Use .459 .788 .340 .560 1.583 Prior Arrests -.747 .815 .841 .359 .474 CONSTANT -3.788 5.726 .438 .508 .023 Strength of Social Bonds: Fatherhood The following model examined the strength of fathers’ social bonds with their children. The measurements of strength were conceptualized in two distinct manners. First, men were asked to rate themselves as fathers. Second, men were asked how they feel about their relationship with their children. For bot h of these variables, higher numbers indicated better situations (An in-d epth description of the coding of these variables can be found in Chapter 4: Methods ). A model was run incorporating both of these variables.

PAGE 186

173 The following model used Wave 3 indepe ndent variables to predict Wave 3 incarceration among men who were fathers at Wa ve 3. A direct logistic regression was conducted on the outcome variable of Wave 3 Incarceration and the following Wave 3 predictors: Age, employment status, income, we lfare use, job satisfaction, attainment of a high school diploma, time at current reside nce, marital status, residential fatherhood, fatherhood prior to age 18, prior arrests, prio r drug use, a rating of one’s success as a parent, and a rating of one’s re lationship with their child(ren). The test of the full model (all predictors ) against a constant only model was not significant (Chi-square = 18.04, df = 14, p > .05) i ndicating that the set of predictors did not reliably distinguish between fathers who were and were not incarcerated during Wave 3. Table 7-5 presents regressi on coefficients where Welfare Use reliably predicted Wave 3 Incarceration among fathers (N = 107). As fa thers tended to use welfare during Wave 3, the odds that they were incarcerated during Wave 3 increased by 9.28. Strength of Social Bonds: Marriage There were not enough subjects who were bot h married and incarcerated at Wave 3 in order to successfully run this model. Wave 3 Independent Variables Predic ting Wave 4 Dependent Variables Wave 3 independent variables were used to predict being incarcerated in Wave 4. More specifically, they pred icted incarceration during th e years of 1985 to 1988. During these years participants were w ithin the ages of 28 and 36. A direct logistic regression was conducted on the outcome variable of Wa ve 4 Incarceration and the following Wave 3 predictors: Age, Employment Status, Income, Welfare Use, Job Satisfaction, Attainment of a High School Diploma, Time at Current Residence, Fatherhood,

PAGE 187

174 Residential Fatherhood, Fatherhood prior to Ag e 18, Marital Status, Prior Drug Use, and Prior Arrests. Table7-5 Wave 3 Independent Variables Pr edicting Wave 3 Incarceration among Fathers VARIABLE B S.E. Wald Sig. Exp(B) Age -.392 .344 1.293 .255 .676 Employed -1.153 .987 1.365 .243 .316 Income .368 .339 1.179 .278 1.445 Welfare Use 2.227 1.043 4.563 .033 9.276 Job Satisfaction -.066 .420 .025 .875 .936 HS Degree -.993 .843 1.390 .238 .370 Time at Res. .300 .267 1.262 .261 1.350 Married 1.494 1.213 1.518 .218 4.456 Res. Father -.664 .985 .427 .513 .525 Father b/f 18 1.016 .949 1.147 .284 2.763 Prior Arrests -.599 .877 .467 .494 .549 Prior Drug Use .478 .947 .255 .614 1.613 Rate Self as parent -.067 .638 .011 .916 .935 Relationship w/ Child -.703 .479 2.150 .143 .495 CONSTANT 3.144 7.436 .179 .672 23.189 The test of the full model (all predictors) against a constant only model was significant (Chi-square = 38.02, df = 13, p < .05) indicating that th e set of Wave 3 predictors reliably distingui shed between participants who were and were not incarcerated. Table 7-6 presents the regressi on coefficients where attainment of a High School Degree, Fatherhood, and Prior Arrests were reliable predictors of Wave 4

PAGE 188

175 Table 7-6 Wave 3 Independent Variab les Predicting Wave 4 Incarceration VARIABLE B S.E. Wald Sig. Exp(B) Age .313 .241 1.691 .194 1.368 Employed -.933 .737 1.602 .206 .393 Income -.337 .253 1.769 .183 .714 Welfare -.607 .896 .460 .498 .545 Job Sat. .081 .386 .044 .834 1.085 HS Degree -1.731 .714 5.876 .015 .177 Time at Res. .023 .206 .012 .911 1.023 Father 2.086 .930 5.037 .025 8.054 Married 1.068 1.073 .989 .320 2.908 Res. Father -1.331 1.024 1.688 .194 .264 Father b/f 18 .795 1.002 .628 .428 2.213 Prior Drug Use -.148 .676 .048 .827 .863 Prior Arrests 2.277 1.002 .628 .021 9.750 CONSTANT -8.031 5.231 2.357 .125 .000 Incarceration (N = 177). As pa rticipants tended to have a high school degree by Wave 3, the odds that they would be incarcerated during Wave 4 decreased by 82.3%. Men who were fathers by Wave 3 had odds of Wave 4 Incarceration that we re 8.05 greater than men who were not fathers. Finally, men who were arrested in the previous Wave had 9.75 greater odds of Wa ve 4 incarceration. Strength of Social Bonds: Fatherhood The following models examined the strength of fathers’ social bonds with their children. The measurements of strength were conceptualized in two distinct manners. First, men were asked to rate themselves as fathers. Second, men were asked how they feel about their relationship with their children. For bot h of these variables, higher

PAGE 189

176 numbers indicated better situations (An in-d epth description of the coding of these variables can be found in Chapter 4: Methods ). A model was run incorporating both of these variables. In the following model Wave 3 variables were used to predict incarceration between 1985 and 1988. A direct logistic regression was conducted on the outcome variable of Wave 4 Incarcer ation and the following Wave 3 predictors: Age, employment status, income, welfare use, j ob satisfaction, attain ment of a high school diploma, time at current residence, marital status, residen tial fatherhood, fatherhood prior to age 18, prior arrests, prior drug use, a rating of one’s success as a parent, and a rating of one’s relationship with their child(ren). The test of the full model (all predictors) against a constant only model was significant (Chi-square = 29.32, df = 14, p < .05) indicating that the set of predictors reliably distinguished between fathers who were and were not incarcerated during Wave 4. Table 7-7 presents regressi on coefficients where Prior Arre sts reliably predicted Wave 4 Incarceration (N = 83). As men were arre sted in Wave 3, the odds that they were incarcerated during Wave 4 increased by 26.85. Strength of Social Bonds: Marriage There were not enough subjects who were bot h married and incarcerated at Wave 3 in order to successfully run this model. Wave 4 Independent Variables Predic ting Wave 4 Dependent Variables Wave 4 independent variables were used to predict being incarcerated in Wave 4. More specifically, they pred icted incarceration during the years of 1989 to 1990. A direct logistic regression was conducte d on the outcome variable of Wave 4 Incarceration and the following Wave 4 predictors: Age, Empl oyment Status, Income, Welfare Use, Job

PAGE 190

177 Table 7-7 Wave 3 Independent Variables Pr edicting Wave 4 Incarce ration among Fathers VARIABLE B S.E. Wald Sig. Exp(B) Age .662 .379 3.046 .081 1.938 Employed -1.365 1.103 1.531 .216 .255 Income -.405 .318 1.615 .204 .667 Welfare Use -1.381 1.237 1.246 .264 .251 Job Satisfaction .182 .513 .125 .723 1.199 HS Degree -1.361 .939 2.101 .147 .256 Time at Res. .031 .250 .015 .903 1.031 Married 1.594 1.400 1.295 .255 4.922 Res. Father -1.344 1.288 1.088 .297 .261 Father b/f 18 1.468 1.145 1.643 .200 4.341 Prior Arrests 3.290 1.267 6.742 .009 26.854 Prior Drug Use .489 .914 .286 .593 1.631 Rate Self as parent .361 .758 .227 .634 1.435 Relationship w/ Child 1.410 1.008 1.955 .162 4.096 CONSTANT -14.949 8.494 3.097 .078 .000 Satisfaction, Attainment of a High School Diploma, Time at Current Residence, Fatherhood, Residential Fatherhood, Fatherhood pr ior to Age 18, Marital Status, Prior Drug Use, and Prior Arrests. The test of the full model (all predictors) against a constant only model was significant (Chi-square = 41.19, df = 13, p < .05) indicating that th e set of Wave 4 predictors reliably distingui shed between participants who were and were not incarcerated. Table 7-8 presents the regressi on coefficients where Em ployment and Prior

PAGE 191

178 Table 7-8 Wave 4 Independent Variab les Predicting Wave 4 Incarceration VARIABLE B S.E. Wald Sig. Exp(B) Age -.078 .191 .168 .682 .925 Employed -2.607 .739 12.451 .000 .074 Income -.043 .202 .045 .832 .958 Welfare -.336 .953 .124 .724 .715 Job Sat. -.090 .357 .064 .800 .914 HS Degree .640 .772 .687 .407 1.897 Time at Res. -.100 .152 .430 .512 .905 Father .917 .823 1.242 .265 2.503 Married -.248 .703 .125 .724 .780 Res. Father -1.289 .905 2.030 .154 .276 Father b/f 18 1.087 .873 1.549 .213 2.965 Prior Drug Use .197 .639 .095 .758 1.218 Prior Arrests 1.975 .996 3.928 .047 7.204 CONSTANT 1.417 6.771 .044 .834 4.124 Arrests reliably predicted Wave 4 Incarce ration (N = 177). Men who were employed during Wave 4 had 92.6% lower odds of incar ceration than men who were unemployed. Arrests during the previous wave indicated a 7.20 increa se in the odds of Wave 4 Incarceration. Wave 4 Independent Variables Predic ting Wave 5 Dependent Variables Wave 4 independent variables were used to predict being incarcerated in Wave 5. More specifically, they pred icted incarceration during th e years of 1991 to 1992. During these years participants were w ithin the ages of 33 and 40. A direct logistic regression was conducted on the outcome variable of Wa ve 5 Incarceration and the following Wave 4 predictors: Age, Employment Status, Income, Welfare Use, Job Satisfaction,

PAGE 192

179 Attainment of a High School Diploma, Time at Current Residence, Fatherhood, Residential Fatherhood, Fatherhood prior to Ag e 18, Marital Status, Prior Drug Use, and Prior Arrests. The test of the full model (all predictors ) against a constant only model was not significant (Chi-square = 14.78, df = 13, p > .05) indicating that th e set of Wave 4 predictors did not reliably di stinguish between participan ts who were and were not incarcerated. Table 7-9 presen ts the regression coefficients where none of the thirteen Wave 4 variables reliably predicte d Wave 5 Incarceration (N = 168). Table 7-9 Wave 4 Independent Variab les Predicting Wave 5 Incarceration VARIABLE B S.E. Wald Sig. Exp(B) Age -.173 .266 .425 .514 .841 Employed -.762 .838 .827 .363 .467 Income 1.539 1.024 2.257 .133 4.661 Welfare -8.438 41.545 .041 .839 .000 Job Sat. .499 .548 .829 .363 1.646 HS Degree -1.230 1.165 1.114 .291 .292 Time at Res. -.085 .209 .166 .683 .918 Father .249 .944 .069 .792 1.282 Married .273 .968 .079 .778 1.313 Res. Father -.449 1.082 .173 .678 .638 Father b/f 18 .150 1.299 .013 .908 1.162 Prior Drug Use -8.624 34.745 .062 .804 .000 Prior Arrests .146 1.257 .013 .908 1.157 CONSTANT -4.157 10.125 .169 .681 .016

PAGE 193

180 Wave 5 Independent Variables Predic ting Wave 5 Dependent Variables Wave 5 independent variables were used to predict being incarcerated in Wave 5 (1993 to 1994). During these years participants were within the ages of 35 and 41. A direct logistic regression wa s conducted on the outcome variable of Wave 5 Incarceration and the following Wave 5 predictors: Age, Employment Status, Income, Welfare Use, Job Satisfaction, Attainment of a High School Diploma, Time at Current Residence, Fatherhood, Residential Fatherhood, Marital Stat us, Prior Drug Use, and Prior Arrests. The test of the full model (all predicto rs) against a constant only model not significant (Chi-square = 34.39, df = 12, p > .01) indicating that th e set of Wave 5 predictors reliably distingui shed between participants who were and were not incarcerated. Table 7-10 presents the re gression coefficients where Employment, Fatherhood, and Residential Fatherhood reliably predicted Wave 5 Incarceration (N = 168). Men who were employed during Wave 5 had odds of incarceration that were 75.5% lower than men who were not employed. Fathers had odds of incarceration that were 76.1% lower than men who were not father s. Finally, residential fathers had odds of incarceration that were 81.4% lower than me n who were not classified as residential fathers. Strength of Social Bonds: Marriage There were not enough participants to successfully run this model. Desistance Models The final set of models examined desist ance measured via incarceration. Within the following models desistance was conceptu alized as not being incarcerated during a specific set of years. There were two distin ct time frames around which these models of desistance were constructed. The two larger sets of models were referred to as

PAGE 194

181 Table 7-10 Wave 5 Independent Variab les Predicting Wave 5 Incarceration VARIABLES B S.E Wald Sig. Exp(B) Age -.118 .166 .501 .479 .889 Employed -1.405 .604 5.402 .020 .245 Income -.125 .127 .971 .325 .883 Welfare Use .260 .546 .226 .635 1.296 HS Degree .683 .664 1.060 .303 1.980 Job Sat. -.232 .258 .812 .386 .793 Time at Res. -.154 .121 1.620 .203 .858 Father -1.433 .643 4.971 .026 .239 Res. Father -1.680 .667 6.351 .012 .186 Married .802 .573 1.955 .162 2.230 Prior Drug Use -1.972 1,193 2.734 .098 .139 Prior Arrests 1.464 1.085 1.821 .177 4.323 CONSTANT 6.756 6.308 1.147 .284 858.992 “Desistance 2” and “Desistance 3.” The num ber in these two names refers to the last wave within which the participant engaged in criminal behavior. For Desistance 2, men who were incarcerated anytime prior to wave 3 (1968-1982) and not incarcerated for various sets of years were cat egorized as desistant (Coded as ”). Likewise, Desistance 3 will count men who were incarcerated a nytime up until the beginning of wave 4 (19681984) and who were not incarcerated through vari ous sets of years as desistant (Coded as ”). Unlike the measures of arrests, ther e were no gaps in the data on incarceration. Desistance 2: Wave 2 Independent Vari ables Predicting Desistance Wave 3 The first two models predicted desistance from incarceration between the years of 1983 and 1988. A direct logistic regression wa s conducted on the outcome variable of Desistance and the following Wave 2 predic tors: Age, School enrollment, Employment

PAGE 195

182 Status, Attainment of a Hi gh School Diploma, Job Satisfaction, Welfare Use, Income, Time at Current Residence, Fatherhood, Fath erhood prior to Age 18, Marital Status, and Juvenile delinquency. The test of the full model (all predictors ) against the constant only model was not significant (Chi-square = 18.90, df = 12, p > .05) indicating that th e set of Wave 2 predictors did not reliably dis tinguish between participants that did and did not desist. Table 7-11 presents regression coefficien ts where Employment reliably predicted desistance (N = 172). Employe d participants had 74.3% lowe r odds of desistance than men who were not employed during Wave 2. Table 7-11 Desist 2: Wave 2 Independent Variables Predicting Desistance Wave 3 VARIABLES B S.E. Wald Sig. Exp(B) Age .100 .168 .352 .553 1.105 In School .366 .490 .560 .454 1.443 Employed -1.359 .486 7.833 .005 .257 HS Degree -.150 .473 .101 .751 .860 Job Sat. .349 .206 2.868 .090 1.418 Welfare Use -1.257 .814 2.384 .123 .285 Income -.073 .163 .200 .655 .930 Time at Res. .027 .098 .074 .786 1.027 Father .211 .541 .153 .696 1.235 Father b/f 18 .406 .854 .226 .634 1.501 Married .626 1.238 .256 .613 1.870 Bad Behav. -.614 .563 1.193 .275 .541 CONSTANT -3.452 3.758 .844 .358 .032 Desistance 2: Wave 3 Independent Vari ables Predicting Desistance Wave 3 This model examined desistance, or th e absence of incarceration between 1983 and 1988. This model used variables from Wa ve 3 (collected 1983-1984) to predict

PAGE 196

183 desistance. A direct logi stic regression was conducted on the outcome variable of Desistance and the following Wave 2 predictors : Age, Employment Status, Attainment of a High School Diploma, Job Satisfaction, We lfare Use, Income, Time at Current Residence, Fatherhood, Residential Fath erhood, Fatherhood prior to Age 18, Marital Status, and Juvenile delinquency. The test of the full model (all predicto rs) against the constant only model was significant (Chi-square = 22.21, df = 12, p < .05) indicating that the set of predictors reliably distinguishes between participants that did and did not desist. Table 7-12 presents regression coefficients where Fath erhood reliably predicted desistance (N = 172). Men who were fathers by Wave 3 held odds of desistance that were 2.97 greater than men who were not fathers. Table 7-12 Desist 2: Wave 3 Independent Variables Predicting Desistance Wave 3 VARIABLES B S.E. Wald Sig. Exp(B) Age .079 .156 .257 .612 1.082 Employed .126 .460 .075 .784 1.135 Income -.140 .146 .920 .337 .869 Welfare Use .035 .582 .004 .952 1.036 Job Sat. -.370 .206 3.217 .073 .691 HS Degree -.293 .467 .393 .531 .746 Time at Res. -.138 .105 1.733 .188 .871 Father 1.090 .493 4.894 .027 2.974 Res. Father -.701 .602 1.354 .245 .496 Father b/f 18 .723 .676 1.146 .284 2.061 Married -1.034 .681 2.307 .129 .355 Bad Behav. -.801 .588 1.856 .173 .449 CONSTANT -.767 3.103 .061 .805 .465

PAGE 197

184 Desistance 2: Wave 2 Independent Variab les Predicting Desistance Waves 3-4 The next three models examined desistan ce through a longer peri od of time. Here men who desisted had avoided incarcerati on from 1983 through 1992. A direct logistic regression was conducted on the outcome variab le of Desistance and the following Wave 2 predictors: Age, School enrollment, Employ ment, attainment of a High School Degree, Job Satisfaction, Welfare Use, Income, Time at current Residence, Fatherhood, Fatherhood prior to age 18, a nd Juvenile Delinquency. The test of the full model (all predictors ) against a constant only model was not significant (Chi-square = 10.99, df = 12, p > .05) i ndicating that the set of predictors did not reliably distinguish between participants that did and did not desist . Table 7-13 presents regression coefficients where Empl oyment reliably predic ted desistance (N = 144). Men who were employed during Wave 2 held odds of desistance that were 66.7% lower than men who were unemployed. Desistance 2: Wave 3 Independent Variab les Predicting Desistance Waves 3-4 In this model Wave 3 variables were used to predict desistance from incarceration between 1983 and 1992. A direct logistic re gression was conducted on Desistance as the outcome variable and the following Wave 3 predictors: Age, Employment, Income, Welfare Use, Job Satisfaction, attainment of a High School Degree, Time at current Residence, Fatherhood, Residential Fath erhood, Fatherhood prior to age 18, Marital Status, and Juvenile Delinquency. The test of the full model (all predictors ) against a constant only model was not significant (Chi-Square = 12.63, df = 12, p > .05) indicating that th e set of Wave 3 predictors did not reliably dis tinguish between participants w ho did and did not desist.

PAGE 198

185 Table 7-13 Desist 2: Wave 2 Independent Variables Predicting Desistance Waves 3-4 VARIABLES B S.E. Wald Sig. Exp(B) Age .117 .205 .329 .566 1.125 In School .135 .601 .050 .822 1.144 Employed -1.099 .552 3.959 .047 .333 HS Degree -.098 .566 .030 .863 .907 Job Sat. .367 .257 2.044 .153 1.444 Welfare Use -1.408 1.116 1.591 .207 .245 Income .050 .208 .057 .811 1.051 Time at Res. .123 .128 .918 .338 1.131 Father .531 .621 .731 .392 1.701 Father b/f 18 -.473 1.292 .134 .715 .623 Married 1.169 1.289 .823 .364 3.220 Bad Behav. -.174 .618 .080 .778 .840 CONSTANT -5.417 4.569 1.406 .236 .004 Table 7-14 presents regression coefficients where none of the twelve Wave 3 variables reliably predicted Desistance (N = 144). Desistance 2: Wave 2 Independent Variable s Predicting Desistance in Waves 3-5 The next set of models predicted desist ance from incarceration from 1983 to 1994. The variables from wave 2 (Collected in 1975 and 1976) were used to predict later desistance from incarceration. A direct logist ic regression was conducted on the outcome variable of Desistance and the following Wave 2 predictors : Age, enrollment in School, Employment, attainment of a High School De gree, Job Satisfaction, Income, Time at current Residence, Fatherhood, Fatherhood be fore age 18, and Juvenile Delinquency. The full model could not be run as the vari ables Welfare Use and Marital Status were problematic. When a cross tabulation was run for each variable, one of the cells contained zero participants. None of the participants who were married during Wave 2

PAGE 199

186 desisted. Additionally, none of the men who used welfare during Wave 2 desisted from Wave 3 through Wave 5. Table 7-14 Desist 2: Wave 3 Independent Variables Predicting Desistance Waves 3-4 VARIABLES B S.E. Wald Sig. Exp(B) Age .175 .189 .860 .354 1.191 Employed .595 .586 1.030 .310 1.813 Income -.023 .203 .013 .909 .977 Welfare Use .112 .775 .021 .885 1.119 Job Sat. -.353 .261 1.824 .117 .703 HS Degree -.115 .572 .040 .841 .892 Time at Res. -.236 .134 3.111 .078 .790 Father .234 .596 .155 .694 1.264 Res. Father .098 .717 .019 .891 1.103 Father b/f 18 .358 .857 .175 .676 1.431 Married -1.565 .902 3.008 .083 .209 Bad Behav. -.131 .635 .042 .837 .877 CONSTANT -3.449 3.794 .826 .363 .032 The test of the full model (all predicto rs) against the constant only model was significant (Chi-square = 25.10, df = 10, p < .05) indicating that the set of predictors reliably distinguished between participants that did and did not desist. Table 7-15 presents the regression coefficients where Employment and Fatherhood reliably predicted Desistance (N = 114). As participants were employed during Wave 2, the odds of desistance decreased by 90.4%. Men who were fathers by Wave 2 had odds of desistance that were 11.46 greater than men who were not fathers. Desistance 2: Wave 3 Independent Variable s Predicting Desistance in Waves 3-5 This model utilized measurements from wave 3 to predict desistance. Here measurements of various social bonds and ot her theoretically signi ficant variables

PAGE 200

187 Table 7-15 Desist 2: Wave 2 Independent Variables Predicting Desistance Waves 3-5 VARIABLES B S.E. Wald Sig. Exp(B) Age .207 .304 .466 .495 1.230 In School .924 .949 .950 .330 2.521 Employed -2.340 1.036 5.105 .024 .096 HS Degree -.791 .962 .676 .411 .453 Job Sat. .846 .506 2.799 .094 2.331 Income -.007 .243 .001 .978 .993 Time Res. .313 .210 2.220 .136 1.367 Father 2.439 1.122 4.725 .030 11.456 Father b/f 18 -1.517 1.375 1.217 .270 .219 Bad Behav. -2.084 1.311 2.527 .112 .124 CONSTANT -9.812 6.630 2.190 .139 .000 collected during 1983 and 1984 were used to pr edict desistance from incarceration from 1983 to 1994. A direct logistic regression was conducted on the outcome variable of Desistance and the following Wave 3 predic tors: Age, Employment, Income, Welfare Use, Job Satisfaction, attainment of a High School degree, Time at current Residence, Fatherhood, Residential Fatherhood, Father hood prior to age 18, and Juvenile Delinquency. The full model could not be run as the variable Marital Status was problematic. When a cross tabulation wa s run, one of the cells contained zero participants, as none of the participants who were married during Wave 3 desisted. The test of the full model (all predictors ) against the constant only model was not significant (Chi-square = 13.40, df = 11, p > .05) i ndicating that the set of predictors did not reliably distinguish between participants that did and did not desist . Table 7-16 presents the regression coefficients where Fatherhood prior to age 18 reliably predicted Desistance (N = 114). Men who were noted to have fathered children before they were

PAGE 201

188 age 18 during Wave 3 held odds of desistance that were 18.64 greater than men who did not father children while they were under the age of 18. Table 7-16 Desist 2: Wave 3 Independent Variables Predicting Desistance Waves 3-5 VARIABLES B S.E. Wald Sig. Exp(B) Age .326 .298 1.196 .274 1.385 Employed .477 .894 .285 .593 1.612 Income -.180 .309 .340 .560 .835 Welfare Use 1.208 .967 1.560 .212 3.347 Job Sat. -.577 .421 1.880 .170 .562 HS Degree -.698 .828 .709 .400 .498 Time Res. .133 .210 .403 .525 1.143 Father -1.092 .879 1.544 .214 .336 Res Fath. .195 .907 .046 .829 1.216 Father b/f 18 2.925 1.210 5.843 .016 18.636 Bad Behav. -1.981 1.393 2.022 .155 .138 CONSTANT -6.603 5.414 1.487 .223 .001 Desistance 3: Wave 3 Independent Variab les Predicting Desistance in Waves 4 The remaining models in this chapter ex amined desistance following Wave 3. Men who desisted in these models may commit an offense up through 1988. The first two models examined desistance through Wave 4. These models examined incarceration between 1989 and 1992. A direct logistic regression was conducted on the outcome variable of Desistance and the following Wave 3 predictors : Age, Employment Status, Attainment of a High School Diploma, Job Sa tisfaction, Welfare Use, Income, Time at Current Residence, Fatherhood, Residential Fatherhood, Fatherhood before age 18, Marital Status, and J uvenile delinquency. The test of the full model (all predictors ) against a constant only model was not significant (Chi-square = 14.32, df = 12, p > .05) indicating that th e set of Wave 3

PAGE 202

189 predictors did not reliably dis tinguish between participants w ho did and did not desist. Table 7-17 presents regression coefficients where Time at current Residence reliably predicted Desistance (N = 140). For every one unit increase in time at residence, the odds of desistance decreased by 21.9%. Table 7-17 Desist 3: Wave 3 Independent Variables Predicting Desistance Wave 4 VARIABLES B S.E Wald Sig. Exp(B) Age .186 .173 1.149 .284 1.204 Work .258 .536 .233 .629 1.295 Income -.101 .184 .299 .584 .904 Welfare Use .074 .680 .012 .913 1.077 Job Sat. -.112 .225 .249 .617 .894 HS Degree -.487 .529 .848 .357 .614 Time at Res. -.247 .116 4.539 .033 .781 Father .957 .553 2.996 .083 2.604 Res. Father -.237 .648 .134 .715 .789 Father b/f 18 -.584 .796 .538 .463 .558 Married -1.097 .678 2.620 .106 .334 Bad Behav. .087 .556 .024 .876 1.091 CONSTANT -3.399 3.466 .961 .327 .033 Desistance 3: Wave 4 Independent Variab les Predicting Desistance in Waves 4 The second model examining desistance from 1989 through 1992 used Wave 4 variables to predict desistance from incarce ration. A direct lo gistic regression was conducted on the outcome variable of Desist ance and the following Wave 4 predictors: Age, Employment, Income, Welfare Use, attainment of a High School Degree, Job Satisfaction, Time at current Residence, Fatherhood, Residential Fatherhood, Fatherhood prior to age 18, Marital Stat us, and Juvenile delinquency.

PAGE 203

190 The test of the full model (all predictors) against a constant only model was significant (Chi-square = 24.39, df = 12, p < .05) indicating that th e set of Wave 4 predictors reliably distinguished between participants that di d and did not desist. Table 7-18 displays the regression coefficients where fatherhood and residential fatherhood reliably predicted Desistance (N = 140). Men who were fathers by Wave 4 displayed 5.58 greater odds of desistance. Men who were residential fathers by Wave 4 held odds of desistance that were 74.5% lower than me n who were not classi fied as residential fathers. Table 7-18 Desist 3: Wave 4 Independent Variables Predicting Desistance Wave 4 VARIABLES B S.E Wald Sig. Exp(B) Age .245 .173 2.004 .157 1.278 Work .888 .582 2.326 .127 2.429 Income -.138 .157 .772 .380 .872 Welfare Use 1.449 .776 3.486 .062 4.260 HS Degree -1.030 .539 3.654 .056 .357 Job Sat. .162 .288 .318 .573 1.176 Time at Res. -.040 .122 .109 .742 .960 Father 1.719 .632 .7405 .007 5.579 Res. Father -1.368 .647 4.473 .034 .255 Father b/f 18 -1.116 .877 1.621 .203 .328 Married -.697 .543 1.645 .200 .498 Bad Behav. -.204 .619 .109 .742 .816 CONSTANT -9.844 6.049 2.648 .104 .000 Desistance 3: Wave 3 Independent Variable s Predicting Desistance in Waves 4-5 The final set of “Desist 3” models pred icted desistance from incarceration from 1989 through 1994. The first model used Wave 3 independent vari ables to predict desistance. Here measurements of soci al bonds and other theo retically significant

PAGE 204

191 variables that were collected in 1983 and 1984 were used to predict desistance from incarceration from 1989 to 1994. A direct logistic regression was conducted on the outcome variable of Desistance and the followi ng Wave 3 predictors : Age, Employment Status, Attainment of a Hi gh School Diploma, Job Satisfaction, Welfare Use, Income, Time at Current Residence, Fatherhood, Re sidential Fatherhood, Fa therhood before age 18, and Juvenile delinquency. The full model c ould not be run as the variable of Marital Status was problematic. When a cross tabula tion was run, one of the cells contained zero participants, as none of the participants who were married during Wave 3 desisted. The test of the full model (all predictors ) against a constant only model was not significant (Chi-square = 12.31, df = 11, p >.05) i ndicating that the set of predictors did not reliably distinguish between participants that did and did not desist . Table 7-19 presents regression coefficients where atta inment of a High School Degree reliably predicted desistance (N = 90). Participants who had acquired a high school diploma by Wave 3 held odds of desistance that were 85.8% lower than those who had not finished high school. Wave 4 Independent Variables Pred icting Desistance in Waves 4-5 This model used Wave 4 independent va riables to predict desistance. The independent variables were collected in 1989 and 1990. Men who were desistant were not incarcerated from 1989 to 1994. A direct logistic regression was conducted on the outcome variable of Desistance and the followi ng Wave 4 predictors : Age, Employment Status, Attainment of a Hi gh School Diploma, Job Satisfaction, Welfare Use, Income, Time at Current Residence, Fatherhood, Re sidential Fatherhood, Fa therhood before age 18, Marital Status, and Juvenile delinquency.

PAGE 205

192 Table 7-19 Desist 3: Wave 3 Independent Variables Predicting Desistance Waves 4-5 VARIABLES B S.E. Wald Sig. Exp(B) Age .447 .352 1.614 .204 1.564 Employed 1.553 .996 2.429 .119 4.724 Income .014 .326 .002 .966 1.014 Welfare Use 1.086 1.002 1.176 .278 2.964 Job Sat. -.513 .412 1.551 .213 .599 HS Degree -1.952 .970 4.052 .044 .142 Time at Res. -.014 .181 .006 .938 .986 Father .543 .853 .406 .524 1.722 Res. Father -1.452 .998 2.115 .146 .234 Father b/f 18 1.344 .998 1.814 .178 3.833 Bad Behav. -2.244 1.466 2.343 .126 .106 CONSTANT -9.583 6.513 2.165 .141 .000 The test of the full model (all predictors ) against the constant only model was not significant (Chi-square = 18.27, df = 12, p > .05) indicating that th e set of Wave 4 predictors did not reliably di stinguish between men who did and did not desist. Table 720 displays the regression coefficients wher e Welfare Use reliably predicted desistance (N = 90). Men who used welfare during Wa ve 4 had 32.85 greater odds of desistance than men who did not use welfare

PAGE 206

193 Table 7-20 Desist 3: Wave 4 Independent Variables Predicting Desistance Waves 4-5 VARIABLES B S.E. Wald Sig. Exp(B) Age .361 .314 1.009 .315 1.371 Employed -.168 1.094 .024 .878 .846 Income .569 .382 2.220 .136 1.766 Welfare Use 3.492 1.571 4.940 .026 32.848 HS Degree -2.190 1.163 3.544 .060 .112 Job Sat. -.885 .671 .1739 .187 .413 Time at Res. .354 .282 1.572 .210 1.424 Father -1.832 1.203 2.321 .128 .160 Res. Father -1.213 1.701 .508 .476 .297 Father b/f 18 3.622 1.874 3.734 .053 37.401 Married -2.366 1.766 1.794 .180 .094 Bad Behav. -3.932 2.599 2.290 .130 .020 CONSTANT 12.350 11.121 1.233 .267 .000 .

PAGE 207

194 CHAPTER 8 DISCUSSION Introduction This research provided the first analysis of criminal behavior and desistance among African American men who lived in Harlem from 1968 through 1994. This project addressed some of the calls of scholars in that it addressed aspects of criminal careers among an all black sample (Laub and Sampson, 2001). The results of this research presented a variety of answers about crim inal careers and desistance from criminal activity among lower income African Americans. The findings allowed for the beginnings of new sets of que stions about the various li fe-course transitions that influence African American men’s de sistance from criminal behavior. This chapter is organized as follows. First, a list of the hypotheses is provided about the three dependent variables, orga nized by groups of conceptually linked independent variables. After each set of hypot heses about a group of conceptually linked variables are provided, the result s from the previous chapters are summarized. However, unlike the previous chapters, al l three dependent variables are discussed in concert rather than separately. This provides a format fo r the discussion of the results, where the disparate influences of the independent va riables upon all three de pendent variables are compared. These sections offer insights and in terpretations about the findings and trends that the models provide. Literature and th eory that supports and opposes the findings are also discussed within these se ctions. At the end of these sections of hypotheses, an overall impression of the results is provided. Next a discussion of the limitations and

PAGE 208

195 significance of this research is provided. Finally future research questions and policy implications are discussed. Answering Research Questions Age 1. Age of Participant: a. Younger participants are expected to be more likely to use drugs/ be arrested/ be incarcerated in the same wave. b. Younger men are expected to be more li kely to use drugs/ be arrested/ be incarcerated in the subsequent wave. c. Younger men are less likely to desist from using drugs/ being arrested/ being incarcerated. Age was a significant predictor in lagged models where drug use and incarceration were the dependent variables. Age did not emer ge as a significant pr edictor of arrests. With regards to drug use, the following was found in the model where Wave 4 independent variables were used to predict Wave 5 Drug Use: For every one unit increase in age during Wave 4, the odds decreased by 34% that the participant would engage in Wave 5 drug use. In the first lagged model on incarceration, every one unit increase in Age during Wave 1 accounted for a 1.68 increase in the odds of being incarcerated during Wave 2. While these two findings revolved around two different dependent variables, they reflected expected changes in criminal beha vior in accordance with the age-crime curve, as noted within the hypotheses. By Wave 4, pa rticipants were between the ages of 31 and 38, when increasing age is correlated with a decreased likelihood of criminal behavior. Similarly, the findings within the Wave 1 independent variables predicting Wave 2 Incarceration model were not su rprising, as participants were between the ages of 12 and

PAGE 209

196 23. Many of these young men were within the ag e group that resides near the peak of the age crime curve. Another factor of note a bout the Wave 1 predicting Wave 2 models is that the only other predictor that was significant across all three models was Family Income. At this early age, these participants did not seem to hold a social bond with their familial income (this is discussed later). Pe rhaps at this young age participants had not formed connections to the institutions of so cial control that are increasingly available later in the life course. Family: Marriage 2. Marital Status: a. Married men are hypothesized to be less likely to use dr ugs/ be arrested/ be incarcerated in the same wave. b. Married men are hypothesized to be less likely to use dr ugs/ be arrested/ be incarcerated in the subsequent wave. c. Married men are hypothesized to be mo re likely desist from using drugs/ being arrested/ being incarcerated. 3. Marital Satisfaction: a. Among married men, those who are in marriages that are rated more favorably are hypothesized to be less li kely to use drugs/ be arrested/ be incarcerated in the same wave. b. Among married men, those who are in marriages that are rated more favorably are hypothesized to be less li kely to use drugs/ be arrested/ be incarcerated in the subsequent wave. c. Among married men, those who are in marriages that are rated more favorably are hypothesized to be more li kely to desist from using drugs/ being arrested/ being incarcerated. Marital satisfaction failed to emerge as a significant predictor for any of the dependent variables. Being married was si gnificant within one model that predicted desistance from drug use. Within the Desi stance-3 model that examined desistance during Wave 4, utilizing Wave 4 predictors , men who were married during Wave 4 had

PAGE 210

197 2.74 greater odds of desistance. When exam ined in terms of years, men who were married by 1989 held greater odds of refraini ng from drug use between the years of 1989 and 1992. Marital Satisfaction’s failure to reliably predict criminal activity was surprising. This finding contra sted with prior research th at directed the theoretical structure of this project (Laub, Nagin, and Sampson, 1998; Laub and Sampson, 1993). These findings did not buttress the aut hor’s hypotheses. While marriage was a reliable predictor of an incr eased likelihood of desistance, th e failure of either of these variables to emerge as significant in more than one model was surprising. Previous research has found marriage to be a signifi cant influence on criminal behavior and desistance (Farrington and West, 1995; Laub, Nagin, and Sampson, 1998; Osborn and West, 1979; Rand, 1987; Sampson and Laub, 1990; 1993; Trassler, 1979; Warr, 1998; West, 1982). While some research has f ound no support for the inhibitive effects of marriage upon drug use (Knight, Osborn, and We st, 1977), other factors may have been at work within this sample. Nielsen (1999) found that marriage held different effects for African American males than it did for white males. Another possible explanation fo r these results could be th at the type of crime under consideration may have influenced the findi ngs, as prior research found an inhibitive effect for marriage among nonwhites for nonviol ent crimes, but not for violent crimes (Piquero, MacDonald, and Parker, 2002). Th is brought to light one potential limitation that could be reflected in these results, with regards to the dependent variables of arrest and incarceration. The crimes for which the pa rticipants were arrested/incarcerated were not made available in this data set. This limited my knowledge of these men’s criminal careers in regards to the seriousness of their cr imes that composed their criminal careers.

PAGE 211

198 Perhaps the relationships displayed between marriage and crime were a reflection of the type of crime committed. Family: Fatherhood 4. Fatherhood Status: a. Fathers are predicted to be less likely to use drugs/ be arrested/ be incarcerated in the same wave. b. Fathers are predicted to be less likely to use drugs/ be arrested/ be incarcerated in the subsequent wave. c. Fathers are predicted to be more like ly to desist from using drugs/ being arrested/ being incarcerated. 5. Fatherhood Prior to the Age of 18: a. Men that father children before the age of 18 are hypothesized to be more likely to use drugs/ be arrested/ be incarcerated in the same wave. b. Men that father children before the age of 18 are hypothesized to be more likely to use drugs/ be arrested/ be incarcerated in the subsequent wave. c. Men that father children before the age of 18 are hypothesized to be less likely to desist from using drugs/ being arrested/ be ing incarcerated. 6. Residential Father: a. Among men, residence with some or a ll of their children is expected to predict a lower likelihood of drug use in the same wave. b. Among men, residence with some or a ll of their children is expected to predict a lower likelihood of dr ug use in the subsequent wave c. Among men, residence with some or a ll of their children is expected to predict a higher likeliho od of desistance from usi ng drugs/ being arrested/ being incarcerated. 7. Rate Self as Father: a. Fathers who rate themselves more favor ably as fathers ar e hypothesized to be less likely to use drugs/ be arrest ed/ be incarcerated in the same wave. b. Fathers who rate themselves more favor ably as fathers ar e hypothesized to be less likely to use drugs/ be arrest ed/ be incarcerated in the subsequent wave. 8. Evaluation of relationship with Child(ren) a. Fathers who rate their relationship with their children more favorably are hypothesized to be less likely to use drugs/ be arrested/ be incarcerated in

PAGE 212

199 the same wave. b. Fathers who rate their relationship with their children more favorably are hypothesized to be less likely to use drugs/ be arrested/ be incarcerated in the subsequent wave. Fatherhood, Residential Fatherhood, and Father hood prior to age 18 all emerged as significant predictors across all three dependent variables. Ho wever, the predictors that evaluated fatherhood success (Rate Self as Fa ther and Evaluation of relationship with Child) failed to emerge as significant. Fatherhood was a reliable predictor across a number of different models. With regards to drug use, Fatherhood was a significant predictor within one desistance model. In the Desist 2 model that used Wave 3 independent variables to predict desistance through Wave 3, men who were fathers by Wave 3 had 80.7% decreased odds of desistance from drug use. This was th e only desistance model where fatherhood predicted a decreased lik elihood of desistance. Fatherhood was a significant predictor within two cross sectiona l Arrest models. When Wave 3 independent variables were used to predict Wave 3 A rrests, men who were fathers had 7.54 greater odds of arrest. Howe ver, in the Wave 5 independent variables predicting Wave 5 Arrest model, the odds of men who were fathers in Wave 5 of being arrested were 73.4% lower than men who were not fathers. Fatherhood was a significant predictor in two cross sectional models and three desistance models where Incarceration was the dependent variable. Men who were fathers by Wave 3 had odds of Wave 4 Incar ceration that were 8.05 greater than men who were not fathers. However men who were fathers at Wave 5 had 76.1% lower odds of Wave 5 incarceration than men who were not fathers. In all three Incarceration

PAGE 213

200 desistance models Fatherhood re liably predicted a greater likelihood of desistance from incarceration. Wave 3 indepe ndent variables predicting Cr ime-2-Wave 3 desistance, Wave 2 independent variables predicting Cr ime-2-Wave 3 through Wave 5 desistance, and Wave 4 independent variables predicti ng Crime-3-Wave 4 desistance respectively noted 2.97, 11.46, and 5.58 greater odds of desi stance from incarceration for men who were fathers. Residential Fatherhood was also significant across all three depe ndent variables. The model of Wave 4 independent variab les predicting Crime-3-Wave 4 desistance revealed that Residential Fa therhood during Wave 4 reliably predicted 3.12 greater odds of desistance from drug use. Residentia l Fatherhood was significant in one cross sectional model that predicted arrests. However, Residential Fatherhood was only significant in one of the limited samples where only men who were married were included in the sample. Men who were bot h married by Wave 5 and who were fathers that lived with their childre n during Wave 5 had odds of a rrest that were 74.7% lower than men who were not classifi ed as residential fathers. Finally, Residential fatherhood emerged as a significant predictor in one cross sectional model and in one model predicting desistance from incarceration. In Wave 5, residential fathers had odds of Wave 5 incarceration that were 81.4% lowe r than men who were not classified as residential fathers. Additionally, in the model of Wave 4 independent variables predicting Crime-3-Wave 4 desistance men, who were residential fathers by Wave 4 held odds of desistance that were 74.5% lower than men who were not classi fied as residential fathers.

PAGE 214

201 The final measure of fatherhood, Fatherhood prior to Age 18, was also a reliable predictor of all three dependent variables. As men tended to be fathers before the age of 18 (at Wave 4), the odds of Wave 5 drug us e increased by 5.96. When examining Wave 4 and 5 desistance from Arrests, following Wa ve 3 crime, and using Wave 3 predictors, men who had fathered a child before the age of 18 were more likely to desist, in that their odds of desistance were 22.66 greater than me n who were not classified as underage fathers by Wave 3. Finally, when examin ing Wave 3, 4, and 5 desistance from Incarceration, following Wave 2 crime, and us ing Wave 3 predictors , participants who were noted to have fathered children before they were ag e 18 during Wave 3 held odds of desistance that were 18.64 greater than men who did not father children while they were under the age of 18. As noted earlier, prior res earch asserted that the in fluence of social bonds on criminal behaviors may be gradual and cumulative (Laub, Nagin, and Sampson, 1998). This notion may be reflected in these result s in that within cross sectional and lagged models, Fatherhood did not hold a positive influence in these men’s lives until Wave 5. This was true for all three variables that related to fatherhood: Fatherhood (itself), Residential Fatherhood, and Fatherhood prior to age 18. Slowly, over the life course these men may have developed a more salient identity as fathers. The Fatherhood variable provided intriguing results with regards to desistance. Fatherhood was a reliable predictor of desist ance within four models (One on Drug Use and three on Incarcera tion). Even more compelling wa s that Fatherhood predicted an increased likelihood of desistance from Incar ceration and a decreas ed likelihood of Drug Use within the same time frame (Crime committed during Wave 2, desistance during

PAGE 215

202 Wave 3, with Wave 3 independent variables as predictors). This be gs the question: How could Fatherhood assert dispar ate influence upon two measures of criminal behavior? The first manner in which this question coul d be addressed, is to look at the outcome variable under consideration. While both outcome variables examine desistance, they are two different measures of desistance. Cert ainly drug use and incarceration differ in many ways. Indeed, could be seen as drug us e as secret deviance (within the Labeling perspective) or as a part of the dark figure of crime (Sutherland, 1947). While the participant’s drug use may have been disc overed by the police, incarceration is a measurement of criminal activity within which the individual has definitely been processed within the criminal justice system . The difference between these two variables could be viewed within life-course crimi nology as differences within the individual’s criminal career trajectory. Drug use and incarc eration could be conceptualized as holding two different levels of seriousness (with in carceration evaluated as more serious), hence yielding different criminal car eer trajectories. Therefore, one could assert that the fatherhood trajectory has the ability to infl uence the criminal career trajectory with regards to incarceration, but not with regards to drug use. The power of either the fatherhood trajectory, or fatherhood as a transi tion, was a reliable predictor of desistance from Incarceration within three different models. Here, the fatherhood trajectory influenced the criminal career with rega rds to crimes serious enough to warrant incarceration. The Residential Fatherhood vari able also displayed contra sting effects with regards to desistance within the same time frame. Here, Residential Fatherhood was a reliable predictor of a greater likelihood of desistance from drug us e, and a reliable predictor of a

PAGE 216

203 lower likelihood of desistance from incarcera tion. Two possible explanations follow. First, it is possible that men within this st udy were indeed less likely to use drugs within their home, due to the presence of children. Pe rhaps this same factor led other men to use drugs in public, hence leading to an in creased likelihood of apprehension and incarceration. However, it is more likely that the Residential Fatherhood variable was not a valid measure of the quality or the st rength of this social bond. While these men may have lived with their children, residenc e does not provide a measurement of the other aspects of fathering whic h contribute to the quality of this bond. More specifically, within the classic conceptualization of fath er involvement, residence only meets the demands of one of the three components (Lam b et. al., 1985, 1987). I ndeed, within this conceptualization of father involvement (with components of engagement, accessibility, and responsibility), residence only meet s the accessibility requirement. Finally, the long-term influence of Fath erhood before the age of 18 suggests that the results support the notion that social bonds such as fatherhood build strength and influence on the life course (and criminal trajectories) gradually. Men who fathered a child early in the life course saw an incr eased likelihood of desist ance from arrests and incarceration as the study progresses. Thes e men held higher likelihood of desistance over long periods of time (1991 to 1994 and 1983 to 1994). If the investment into the fatherhood bond is a gradual process, this w ould explain the positive effects that youthful fathering has upon desistance. Residential Status 9. Residential Stability a. Men who maintain the same reside nce for a longer period of time are hypothesized to be less likely to use drugs/ be arrested/ be incarcerated in

PAGE 217

204 the same wave. b. Men who maintain the same reside nce for a longer period of time are hypothesized to be less likely to use drugs/ be arrested/ be incarcerated in the subsequent wave. c. Men who maintain the same reside nce for a longer period of time are hypothesized to be more likely to desi st from using drugs/ being arrested/ being incarcerated. Residential stability was a reliable predic tor for all three depe ndent variables. However, it failed to significantly predict Arrests and Incarceration within the lagged and cross sectional models. Residential stability was a reliable predictor of drug use within two cross sectional models and one desistan ce model. Both cross sectional models involved Wave 3 independent variables predic ting Wave 3 Drug Use. The initial Wave 3 independent variables predicting Wave 3 Dr ug Use model was comprised of all Wave 3 participants. In this model, every one unit increase in Time at Residence decreased the odds of Wave 3 Drug Use by 17.6%. The remain ing model was comprised of all fathers at Wave 3, where two different evaluations of the strength of the fatherhood social bond were included as additional pred ictors. In this father “success” model, men were asked to evaluate both their success as fathers and their relationship with their children. Here, every one unit increase in Time at Residenc e decreased the odds of Wave 3 Drug Use by 32.5%. Thereby, residential stability at Wave 3 corresponded with a decreased likelihood of drug use. Interestingly, the power of reside ntial stability was great er for fathers than it was for the sample as a whole at Wave 3. Ho wever, residential stability in the previous wave (Wave 2) displayed contrasting influen ce in relation to desist ance from drug use. When Residential Stability in Wave 2 was us ed to predict Crime Wave 2/ Desistance

PAGE 218

205 Wave 3, every one unit increase in time at current residence decreased the odds of desistance by 22.5%. Residential stability displayed deleteri ous effects on desistance across all three dependent variables. As noted above, Wa ve 2 residential stability was a reliable predictor of a decreased likeli hood of desistance from drug use. The same held true for desistance from arrests and in carceration. When Wave 3 independent variables were used to predict Wave 3 through 5 desistance fr om arrests following Wave 2 crimes, every one unit increase in Time at Residence d ecreased the odds of desistance by 29.6%. Additionally, when Wave 3 i ndependent variables were used to predict Wave 4 desistance from incarceration following Wave 3 crimes, every one unit increase in Time at Residence decreased the odds of desistance by 21.9%. These findings, in particular, display the different outcomes that can be seen when cross sectional and longitudinal models are compared. Men who maintained the same residence for longer periods of time during th e years of 1983 and 1984 held lower odds of drug use. However, participants who remain ed in the same residence across a number of different years were more likely to use drugs , be arrested, or be incarcerated. These oppositional findings reflect two manners in whic h residential stability can influence the lives of men. Previous research found that men who moved away from the city where they grew up were less likely to continue to engage in criminal behavior than men who remained (Osborn, 1980; West, 1982). Other res earch noted that resi dential stability was decreased the likelihood criminal behavi or (Laub and Sampson, 1988; Sampson and Laub, 1990). The divergent results seen in previous works were not answered within this project.

PAGE 219

206 Employment 10. Employment Status a. Men who are employed are predicted to be less likely to use drugs/ be arrested/ be incarcerated in the same wave. b. Men who are employed are predicted to be less likely to use drugs/ be arrested/ be incarcerated in the subsequent wave. c. Men who are employed are predicted to be more likely to desist from using drugs/ being arrest ed/ being incarcerated. 11. Job Satisfaction a. Men with a higher level of job satisfaction are predicted to be less likely to use drugs/ be arrested/ be inca rcerated during the same wave. b. Men with a higher level of job satisfaction are predicted to be less likely to use drugs/ be arrested/ be incarcer ated during the subsequent wave. c. Men with a higher level of job satisfac tion are predicted to be more likely to desist from using drugs/ be ing arrested/ being incarcerated. Job satisfaction was signif icant in one model: Wave 5 Independent Variables Predicting Wave 5 Drug Use. Here, the odds of Wave 5 drug use decreased by 33.9% for every one unit increase in job satisfaction. Employment was a reliable predictor of Arrests and Incarceratio n. Employment failed to emerge as a significant predictor of drug use. Employment reliably predicted Arrests acr oss two desistance models and two cross sectional models. Both cross sectional mode ls used Wave 3 independent variables to predict Wave 3 Arrests. Th e initial model sampled all Wa ve 3 participants, while the other model was comprised of men who were fathers by Wave 3. These models (the initial model, and the father success model) revealed Employment to be a variable that predicted decreased odds of arrests (94.7% and 94.3%). Interestingly, Employment did not yield positive social effects within the desistance from arrest models. In two models

PAGE 220

207 that considered desistance following a crim inal act in Wave 2, employment reliably predicted decreased odds of desistance from arrest. Wave 2 employment reliably predicted 64.4% lower odds of desistance. Likewise Wave 3 employment reliably predicted 70.5% lower odds of desistance from arrests. Participants, who were employed during Wave 2, had 83.2% d ecreased odds of being incarcerated during Wave 3. Men who were employed during Wave 4 had 92.6% lower odds of incarceration than men who we re unemployed. Men who were employed during Wave 5 had odds of Wave 5 incarcerati on that were 75.5% lower than men who were not employed. Across all three mode ls, Employment corresponded with decreased odds of incarceration. However, the positive influence of Employment did not persist within three models examining desistance from incarceration. All three of these models used Wave 2 independent variables to pred ict desistance from incarceration following criminal behavior at Wave 2. When desi stance through Wave 3 was predicted, Employed participants had 74.3% lower odds of desistan ce than men who were not employed during Wave 2. When desistance in Waves 3 a nd 4 was predicted men who were employed during Wave 2 held odds of desistance th at were 66.7% lower than men who were unemployed. Finally, when desistance in Wave 3 through Wave 5 was predicted participants who were employe d during Wave 2, held odds of desistance that were 90.4% lower than men who were unemployed. Some of the results here were supported by prior research in that employment and job satisfaction were associated with lower probabilities of criminal behavior (Laub and Sampson, 1993; Sampson and Laub, 1990; 1993; Shover, 1985; 1996; Uggen, 2000). However, all of the desistance models reveal ed Employment as a f actor that decreased

PAGE 221

208 the likelihood of desistance. Prior research has found cases where employment increased the odds of criminal behavior (Horney, Osgood, and Marshall, 1995). In many cases the relationship between work and cr ime has been mediated by age (Uggen, 2000). The distinction between the cross sectional and desistance models could have been a reflection of the different ways that employ ment affected criminal behavior between subjects and within subjects. Finally, with regards to employment, this was the only variable (other than residen tial fatherhood) where the consider ation of the strength of the social bond was a significant predictor (Althoug h the validity of Residential Fatherhood as a measurement of the strength of the so cial bond was questioned above). Higher job satisfaction in Wave 5 reliably predicted a lower likelihood of Wave 5 Drug Use. Once again, we see a variable that predicted a lowe r likelihood of criminal behavior in the later stages of the data collection. When viewed in context of the four elements of social bonds (attachment, commitment, involvement, and belief), it seems that participants develop higher levels of commitment to va rious social bonds later in the life course. Finances 12. Income a. Men with higher incomes are hypothesi zed to be less likely to use drugs/ be arrested/ be incarcerat ed during the same wave. b. Men with higher incomes are hypothesi zed to be less likely to use drugs/ be arrested/ be incarcerated during the subsequent wave. c. Men with higher incomes are hypothesi zed to be more likely to desist from using drugs/ being a rrested/ being incarcerated. 13. Welfare Use a. Men who use welfare are predicted to be more likely to use drugs/ be arrested/ be incarcerated in the same wave. b. Men who use welfare are predicted to be more likely to use drugs/ be arrested/ be incarcerated in the subsequent wave.

PAGE 222

209 c. Men who use welfare are predicted to be less likely to desist from using drugs/ being arrested/ being incarcerated. Income emerged as a significant predicto r of both Drug Use and Arrests. While Income predicted Drug Use across both cros s-sectional and desist ance models, Income only predicted Arrests within cross sectional models. Welfare was a significant predictor of Incarceration. For every one unit increase in Wave 1 Fa mily Income, the odds of Wave 2 Drug Use increased by 1.14. In the models where Wave 3 independent variables were use to predict Wave 4 Drug Use, measures of “succe ssful” fatherhood were available. Hence, an additional model was run to examine the in fluence that successful or more powerful social bonds held over criminal behavior. In the intial model, every one unit increase in Income decreased the odds of Wave 4 drug use by 38%. In th e “Strength of Social Bond: Fatherhood” model, every one unit increase in Income decr eased the odds of Wave 4 drug use by 44.1%. Perhaps these disparate finding s reflected the different social spaces that the participants occupied during Wave 1 and Wave 3. During Wave 1, participants lived with their family and were between the ages of 12 and 17. The income under consideration at Wave 1 was the family income , not the income of the participant. At Wave 3, these men were between the ages of 26 and 31. The income noted was their own. The positive relationship between income and drug use at Wave 1 may have been a reflection of “disposable income,” at leas t through the eyes of the young participants. Three Drug Use desistance models revealed contrasting influence from Income. When Wave 3 independent variables were us ed to predict Wave 3 desistance from Drug Use following Wave 2 criminal behavior, every one unit increase in income decreased the

PAGE 223

210 odds of desistance from Drug Use by 37.8%. However in the Wave 3 independent variables predicting crime Wave 3-desistance Wave 4 model and the Wave 4 independent variables predicting crime Wave 3-desist ance Waves 4 and 5 model, every one unit increase in income, amounted to a 1. 75 and 2.52 increased odds of desistance respectively. These divergent results created a series of questions. A period effect was possible, as a key historical period within Harlem was the Crack Epidemic. However, this occurred during the late 1980’s and early 1990’s, and during this time period higher incomes predicted a higher likelihood of desi stance from drugs (rather than lower). The best explanation for these contrasting results is that social bonds hold different influence over individuals at different points of th e life course. From 1983 to 1988, a lower likelihood of desistance was not ed, while a higher likelihood of desistance was predicted from 1989 to 1992, as well as from 1989 to 1994. Income was significant within two cross se ctional models predicting Arrests. For every one unit increase in Wave 3 Income, th e odds of Wave 3 Arrest increased by 1.84. Income was also significant in one of the “Father Success” models. In this model every one unit increase in income increased the odds of Wave 3 Arrests by 2.07 among fathers. Both this finding and the finding discussed in the previous paragraph were confusing. One possible explanation was that the higher income could have been the result of illegal activities. However, only eight of the Wave 3 participants noted that their income was derived from illegal means. The remaining pa rticipants noted that full time jobs, part time jobs, government assistance, or family assistance were the main sources of their income. Indeed, 64.3% of the sample noted th at the main source of their income was a

PAGE 224

211 regular job, compared with 1.4% of the sample who noted that the main source of their income was from illicit means. Some of these results suppor ted prior research which f ound that higher levels of income were related to lower levels of drug use and higher levels of desistance (Pezzin, 1995; Trassler, 1979; Waldorf, Reinarman, a nd Murphy, 1991). However, three cross sectional models and one desistance model linke d higher income with criminal behavior. Perhaps the key distinction within these mode ls was that higher income resulted in positive social actions from Wave 4 thr ough the end of the study. The two cross sectional models that displayed a lower like lihood of drug use contained Wave 4 outcome variables. Likewise, the tw o desistance models where higher incomes were predicative of a higher likelihood of desistance examined desistance from drug use during Waves 4 and 5. This distinction could be indicative of the long term outcome of higher incomes. Perhaps the distinction between these models revealed that higher income may influence desistance in the same way that prior resear ch has described other social bonds such as marriage: gradually (Laub, Nagin, and Samps on, 1998). If higher levels of income are viewed as an aspect of social bonds this w ould make sense. Men who hold higher levels of income are gradually expected to act differently in accordance with their higher socioeconomic status. For these men, immersi ng oneself in the new status of a person of higher status may take time, and hence may result in a higher likelihood of desistance from criminal behavior in the long run. A dditionally, new wealth without the assumption of the status of an individu al of higher economic means could be spent on drugs for the purpose of leisure, or as symbols that repr esented status in thei r prior socioeconomic position. The positive impact of Income with in the cross sectional models could be

PAGE 225

212 viewed along parallel lines where, as noted above, soci al bonds can offer divergent influence at different stages of the life course. Welfare Use emerged as a significant pred ictor of Incarceration. While Welfare Use failed to reliably predict Incarceration wi thin the entire sample at Wave 3, Wave 3 fathers displayed influence from Welfare Use. In the “fatherhood success” model, fathers who tended to use welfare during Wave 3 held 9.28 increased odds of Wave 3 incarceration. However, Welfare Use also acted as a positive social factor when examining desistance from incarceration. W ithin the crime Wave 3, desistance Waves 4 and 5 model, men who used welfare during Wave 4 had 32.85 greater odds of desistance from incarceration than men who did not use welfare. Prior research has shown that the inability to provide for one’s children is tied to a lack of parental involvement (Achatz a nd MacAullum, 1994; Anderson, 1993; Danziger and Radin, 1990; Furstenberg, 1995). Perhaps the use of welfare amongst this sample distanced these men from their father role or identity, hence weakening or breaking a powerful social bond. The use of welfare among fathers may be an indication that these men were cut off from the informal soci al bonds provided by employment and family (via paternal particip ation). Two theoretical questions emerged from this finding. First, did the use of welfare for fathers act as a tr ansition within their life course? If so, a second question needs to be aske d: Did the transition into welfare use cause these men to diverge from their trajectory as a father, or some other social bond that had previously impeded criminal behavior? On the other hand, within the genera l sample (all cases) Welfare Use yielded positive social impact. Here, men who were able to successfully

PAGE 226

213 utilize government assistance may have develope d a small, but in th is case, substantial connection to society. Education 14. GPA a. Men who held a higher Grade Point Aver age in Wave 1 are predicted to be less likely to use drugs/ be arrested/ be incarcerated in the subsequent wave. 15. High School Degree a. Men with a high school degree or highe r are hypothesized to be less likely to use drugs/ be arrested/ be inca rcerated within the same wave. b. Men with a high school degree or highe r are hypothesized to be less likely to use drugs/ be arrested/ be incar cerated within the subsequent wave. c. Men with a high school degree or hi gher are hypothesized to be more likely to desist from using drugs/ being arrested/ be ing incarcerated. Attainment of a High School Degree re liably predicted all three dependent variables. Meanwhile, GPA and Enrollment in School did not emerge as reliable predictors within any of the models. Attain ment of a High School Degree was a reliable predictor within two cross sect ional “success” models. In the model predicting Drug Use among fathers, Wave 3 High School graduati on predicted a higher likelihood of Wave 4 Drug Use. Men who held High School Dipl omas by Wave 3 had 13.42 greater odds of Wave 4 Drug Use than men who did not gra duate. The Wave 5 independent variables predicting Wave 5 Drug Use marital success mo del revealed that married men who had attained a high school degree by Wave 5 were 92.3% less likely to us e drugs than married men who had not completed high school. Fi nally, high school graduation was also a reliable predictor of desistance from drug use within one model. Within the crime Wave

PAGE 227

214 2, desistance Wave 3 model, men who had atta ined a high school degree by Wave 3 had 5.77 greater odds of desistance. Attainment of a High School Degree was a re liable predic tor within one desistance from Arrests model and three Incarcerati on models (two cross sectional and one desistance). Men who had attained a hi gh school diploma by Wave 3 held odds of desistance from being arrested that were 75.1% lower than men who had not received a diploma within the Crimes Wave 3, Desistan ce Waves 4 and 5 model. When Wave 3 independent variables were used to predic t Wave 3 Incarceration the odds of Wave 3 Incarceration were 78.4% lower for men who had attained a high school degree by Wave 3. Additionally, when Wave 3 independent variables predicted Wave 4 Incarceration the odds of Wave 4 Incarceration were 82.3% lower for men who had attained a High School Diploma. Finally, within the one Incarc eration desistance model where High School graduation was significant, partic ipants who had engaged in cr iminal behavior in Wave 3 and had acquired a high school diploma by Wave 3 held odds of desistance through Waves 4 and 5 that were 85.8% lower than those who had not finished high school. Two potential explanations emerged for the overall negative effect of education. First, as previously noted for other vari ables, social bonds hold diverse impact on people’s lives at different stag es of the life course. This, along with the declining value of a high school diploma and the rising perc entage of the population who held college degrees, could explain why a high school educ ation was linked to a higher likelihood of desistance in Wave 3 and a lower likeli hood of desistance in Wave 4 and 5. The second explanation considered the quali ty and strength of the social bond of education for this sample. Perhaps, the quali ty of the education received by these men

PAGE 228

215 was low as they lived in a lower income neighborhood. Previous research has linked the strength of social bonds to their impact on desistance (Laub, Nagin, and Sampson, 1998; Laub and Sampson, 1993; Samspson a nd Laub, 1993; Sampson and Laub, 1990). Likewise, poor academic performances and low levels of attachment to school have also been common in both young and adult offenders , revealing educationa l opportunities as a factor in criminal involvement (Samps on and Laub, 1993; Hagan and McCarthy, 1997; Wolfgang et al., 1972). This could be seen as a two-way street where these young men had a lack of access to quality education whic h either minimized their connection to this bond or caused them to obtain a connection to a social institution that had no positive influence on their lives. Prior models were r un using the level of e ducation rather than a dichotomous measure of a high school degree . In these models, higher levels of education always predicted lower levels of desistance, and higher levels of criminal behavior. Prior Deviance 16. Bad Behavior a. Men who engaged in bad behavior in Wave 1 are hypothesized to be more likely to use drugs/ be arrested/ be incarcerated in the subsequent wave. b. Men who engaged in bad behavior in Wave 1 are hypothesi zed to be less likely to desist from using drugs/ being arrested/ be ing incarcerated. 17. Drug Use a. Men who previously used drugs are hypothesized to be more likely to use drugs/ be arrested/ be incarc erated in the same wave. b. Men who previously used drugs are hypothesized to be more likely to use drugs/ be arrested/ be incarcer ated in the subsequent wave. c. Men who previously used drugs are hypothe sized to be less likely to desist from using drugs/ being a rrested/ being incarcerated. 18. Arrests a. Men who were previously arrested ar e hypothesized to be more likely to

PAGE 229

216 use drugs/ be arrested/ be incarcerated in the same wave. b. Men who were previously arrested ar e hypothesized to be more likely to use drugs/ be arrested/ be incar cerated in the subsequent wave. c. Men who were previously arrested are hypothesized to be less likely to desist from using drugs/ bei ng arrested/ being incarcerated. Bad Behavior failed to reliably predict a ny of the outcome variables. Prior Drug Use predicted drug use in cross sectional a nd lagged models. Prior Arrests predicted future arrests in cross sectional and lagge d models. Finally, Prior Incarceration was a reliable predictor of Drug Use, Arre sts, and Incarceration in Wave 5. Participants who used drugs during Wave 2 had 4.98 greater odds of Wave 3 drug use than men who did not engage in previous drug use. Similarly, men who previously used drugs had 4.67 greater odds of Wave 3 drug use than men who did not engage in prior drug use. These findings also held tr ue when Wave 3 independent variables were used to predict Wave 3 Drug Use among fathers. Fathers who previously used drugs held odds of Wave 3 Drug Use that were 4.58 highe r than fathers who refrained from prior drug use. This trend continue d in the two models that used Wave 3 independent variables to predict Wave 4 Drug Use (among all part icipants and among fathers respectively) where prior drug users held 4.58 and 4.55 greater odds of Drug Use. Prior Drug Use as noted in Wave 4 also was a reliable predic tor of Wave 4 Drug Use. Here, men who had previously used drugs had 4.24 greater odds of Wave 4 Drug Use. Prior arrests, as noted during Wave 3 in terviews, reliably predicted a decreased likelihood of future arrests, both within th e entire sample and amongst the sample of fathers. The odds of men who were arrest ed from 1977 to 1982 of being arrested during Wave 3 (1983 to 1984) decreased by 84.2%, as compared with men who were not

PAGE 230

217 previously arrested. Within the fatherhood model, as participants tend ed to be previously arrested, the odds that they would be arre sted during Wave 3 decreased by 86.1%. Prior arrests also reliably predicted Incarceration. In the Wave 3 independent variables predicting Wave 4 Incarceration mode l, men who were arrested in the previous wave had 9.75 greater odds of Wave 4 Incarcerat ion. The same trend held true within the model that included only men who were fath ers within the sample. Men who were fathers by Wave 3 that had previously been arrested had 26.85 great er odds of Wave 4 Incarceration. Finally, within the Wave 4 independent va riables predicting Wave 4 Arrests model (full sample), men who were previously arrested had 7.20 greater odds of Wave 4 Incarceration. As noted within Chapter 4 (Methods), a seri es of models were run to examine the impact that all three dependent variables could hold as predictors of future criminal behavior. The models where Incarceration was a significant predictor are discussed here. All three models were composed of Wave 5 predictors and Wave 5 outcome variables. Men who were previously incarcerated were 81.6% less likely to us e drugs than men who had not been previously incarcerated dur ing the previous Wave. Men who were previously incarcerated had odds of arrest th at were 11.11 greater than men who were not previously incarcerated. Finally, men who we re previously incarcerated had odds of Wave 5 incarceration that were 533.29 great er than men who were not previously incarcerated. This abnormally large number could have been a reflection of Dr Brunswick’s tenacity in retain ing participants. The primar y investigator of this study would search police and probation records to find her participants. This number may have reflected that some of the men from w ho were incarcerated in the previous wave

PAGE 231

218 were still incarcerated for the same crime. Interestingly, incarceration seemed to have a deterrent effect on Drug Use, while increasi ng the likelihood of futu re involvement with the criminal justice system. Three main findings emerged with regards to these variables. First, as Bad Behavior failed to emerge as significant, th ese findings reminded us that most individuals that offend as children do not offend as a dults (Gove, 1985). Second, sometimes official involvement with the criminal justice system acts as a deterrent to future criminal behavior (Nagin and Patternoste r, 1991). Finally, prior criminal behavior was a strong predictor of future criminal behavior (B lumstein et al., 1985; et al., 1993; Nagin and Paternoster, 1991; 2000). Theoretical Applications Interpretations of the findings provided thre e broad theoretical insights. First, the manner in which bonds function over the life c ourse was revealed. Second, the ways in which the strength, quality, or su ccess of social bonds operate s was revealed. Finally, the manner in which social bonds function am ong African American men was displayed. Overall, the findings in this research pr esented the manner in which social bonds are acquired, are cultivated, and yield influen ce on the lives of African American men. Many of the social bonds under considera tion revealed disparate influence at different stages of the life course. Fath erhood (along with Fatherhood before Age 18 and Residential Fatherhood), Income, Educati on, and Employment/Job satisfaction all displayed varying influence during different wa ves. All of these so cial bonds displayed positive social influence during later stages of the life course. The men who comprised the sample of this research acquired a broa d array of social bo nds which took time to

PAGE 232

219 influence desistance from criminal behavior. While a critique of these results might claim that these men simply aged out of crime; two issues would refute that notion. First, age was only significant within two models. Second, the stage at when the social bonds yield influence on desistance and cessation (cross sectional models) from crime is well beyond the peak of the age-crime curve. Overa ll, these results fall more in line with the DLC notions that social bonds de velop gradually and that elements operate differently at various stages of the life course. These results were indicative of the conn ection between social control theory and life course criminology within the DLC theory which guided this research: Sampson and Laub’s Age Graded Informal Social Control Th eory. More specifically the social timing element of trajectories and the gradual a nd cumulative effects of social bonds were evident. Social timing considers what happens, when it happens, and for how long (Elder, 1994). The age-graded expectations that accompany certain roles and life events intertwine with the development of interd ependencies which give social bonds their strength (Elder, 1994; Laub, Nagin, and Sampson, 1998; Laub and Sampson, 1993; Nagin and Paternoster, 1994). As people age in to adulthood they develop social ties that impose negative sanctions on behaviors whic h violate role expectations (Laub and Sampson, 1993). The contrasting influence of Income and the Fatherhood variables displayed how the same bonds yi elded different influence at different stages of the life course. Another way in which these results were interpreted was through an examination of the gradual process through which bonds develo p. Bonds do not simply materialize, as they take time to develop (Laub, Nagin, a nd Sampson, 1998). Over time, the investment

PAGE 233

220 that individuals make into so cial bonds grows, as does the incentive for avoiding specific behaviors (Laub, Nagin, and Sampson, 1998; Laub and Sampson, 1993; Nagin and Paternoster, 1994). The emergence of social bonds should therefore be viewed as an investment process, whereby “enduring attach ments” to a specific role, or a specific relationship, influences the manner in wh ich one acts (Laub, Nagin, and Sampson, 1998; Laub and Sampson, 1993; Nagin and Paternoster, 1994). The men in this survey displayed evidence of these “enduring attachme nts” specifically with regards to their roles as fathers. A series of variables were included in order to examine participants’ commitment to their social bonds and the success along th eir trajectories. Th e variables which found support for the hypotheses regarding the success of trajectories were Job Satisfaction and Residential Fatherhood. Residential Father hood displayed contrasting influence at different stages, and may not have been a valid measurement of success within the fatherhood trajectory. Future research needs to examine whether or not residence is an acceptable conceptualization of strength. Indeed, within the frequently utilized conception of father involvement (with co mponents of engagement, accessibility, and responsibility), residence only meets the acc essibility requirement (Lamb et. al., 1985, 1987). Additionally, domestic violence reveals that residence is not the sole component within the familial bond. The variables of Job Satisf action and Marital Satisfac tion also revealed much insight. Job satisfaction only emerged as si gnificant within one model that examined drug use at the end of the survey (Wave 5). This finding could be linked to two prior notes. First, social timing may have played a role here, causing J ob Satisfaction to hold

PAGE 234

221 greater influence at a later stage within these men’s lives. Second, the potential social bonds available may have influenced this variab le. If the jobs available were unlikely to either yield satisfaction or to be jobs which were worthy of becoming careers, it would not be surprising that this va riable was infrequently significant. Marital satisfaction’s failure to emerge as a significant predic tor was surprising. These findings may have reflected prior research in that marriage held different effects for African American males than it did for white males (Nielsen, 1999; Piquero, MacDonald, and Parker, 2002). As many prior studies have focused on white samples, this research addresses a gap in research through the examination of an all African American sample (Laub and Sampson, 2001). More specifically, this research ed examined life course transitions with an all black sample that lived in a povert y stricken neighborhood. The late movement towards desistance, the positive impact of fatherhood during their 30’s, the non-existant impact of marriage, and the ne gative impact of education ar e all findings which should be examined in future research. As some of th ese findings violate prio r assertions on social bonds, the manner in which bonds function as deterrents to criminal behavior among people of various races and socioeconomic group s needs to be investigated. This will be covered in greater detail when po licy implications are discussed. Limitations and Significance Within this section some of the potential limitations and the significance of this study were addressed. First, the sample was not highly generalizable, as it consisted of African American males who were between 12 and 17 in 1968, and who lived in Central Harlem. However, the sample within this rese arch may have also been one of its greatest contributions to the field. As surveys using general youth samples have been critiqued

PAGE 235

222 due to either under-sampling, or entirely missing out on sampling, chronic, or more serious, offenders (Brame and Piquer o, 2003; Cernkovich, Giardano, and Pugh, 1985; Piquero, Farrington, and Blumst ein, 2003) the over sampling of a low income population may have provided a sample that was more lik ely to contain serious offenders (Piquero, Farrington, and Blumstein, 2003). Additionally, as this research used a survey that contained a general population sample, participation rather than frequency was used to examine criminal behavior (Piquero, Farri ngton, and Blumstein, 2003). In this manner the outcome variables were appropriate for the sample. Additionally, the concern of under sampling chronic offenders was met th rough the populations which were included in the sample. Another problem with the sample was the le vel of attrition from Wave 1 to Wave 5. Over the course of the study more than half of the original subjects (N=183) were lost to attrition. However, most of the loss occurr ed randomly. Only 3 variables across all waves displayed significant patterns of attri tion. Three variables (Wave 3 Arrests, Wave 4 Age, and Wave 4 Income) suffered from non -random attrition were adjusted. As noted earlier, the pattern of random attrition early on negates some of the concerns of attrition (Goodwin, 2001). However, future uses of th is data set by the author will contain substitutes of a scale mean for the missing items to further address this issue (Brame and Piquero, 2003; Menard, 2002). Left-hand censoring is a problem as ch anges in offending and social bonds that occurred prior to Wave 1 are not known. Mo re measurements of juvenile delinquency would have been helpful in this project. The fact that the Bad Behavior (juvenile delinquency) variable did not emerge as sign ificant revealed that a more sound measure

PAGE 236

223 of juvenile delinquency was needed. Meanwhile , the variable of desistance faced similar problems at the end of the study. While desist ance was conceptualized as periods of nonoffending in this project, engagement in deviant behavior following Wave 5, which would signify non-desistance, was not measured and therefore remained unknown. This related directly to the concern over the definition of desistan ce. The lasting influence of the social bonds within this survey will neve r be known. Indeed, all that can be construed from this project is that cer tain social bonds did influence periods of desistance between the years of 1968 and 1994. The final set of problems faced in this research had to do with the variables within the data set. First, not all variables were consistently included in all waves. Both independent and dependent variables were mi ssing from some of the waves. Second, the ways in which some variables were measur ed were not conducive to answering the research questions proposed. Most of these problems have been dealt with via recoding. However, some variable were not measured across all waves, while others must be measured differently in various waves. Questions also remain about the variables used in this research. There were two core problems. First, did th e dependent variables provide adequate measures of crime? Second, did the independent variables provide valid measures? The main problem with the dependent variable could be better understo od when they are couched in terms of the criminal career. While the type of sample within this survey lent itself to an examination of participation within crime rather th an frequency and seriousness, a deeper understanding of the roles that social bonds pl ayed in the lives of these men would have been furthered by more detailed information a bout participation in cr ime. While future

PAGE 237

224 research utilizing this data set can expand on this issue, the options are limited. The drug use measures have the greatest potential for an examination of frequency and seriousness as more detailed information on frequency and th e type of drugs used are available. Data on incarceration and arrests are di verse and limited. However, as with much research the questions answered here lead to more que stions. With regards to the independent variables, the ability of these measures to cap ture the strength or quality of social bonds was concerning. Perhaps the failure of many of these variables to arise as significant is an indication of problems as measurements. However, it is also possible that these aspects of social bonds did not hold influence over the lives of the participants as well. Implications for Policy and Future Research Policy Implications Finally, this research lead s to a variety of recommendations for research beyond this data set and for future policy. This res earch informs policy makers in that desistance should not be viewed as a short term event. Many of the positive influences of social bonds appeared in later years, or were th e most prevalent over longer periods of time rather than shorter periods of time. Additi onally, the expected infl uence of social bonds via government programs and policies should be examined and administered with an eye towards racial and socioeconomic differences. First, further longitudinal, life course research needs to be done on poor, African American male samples before major limitations are placed on the implications of this research. Indeed, prior resear ch has found that some of the main divergences found here are indicative of the ways th at social bonds operate differe ntly for people of different races (Nielsen, 1999; Piquero, MacDonald, and Parker, 2002). However, an examination

PAGE 238

225 of the influence of social bonds upon the fre quency and seriousness of criminal activity amongst this sample would be recommended before this work is used as a tool for policy makers. Fathers who are within the criminal jus tice system are a population that needs special attention. The long term impact of fa therhood revealed that these men are open to change and that they are worth the effort to rehabilitate. Despite current sentencing practices, consideration of fatherhood (especially among men who can prove high levels of attachment and generative behaviors with their children) would seem to be a policy that would assist a group that seem more open to life course changes than other groups of offenders (Edin et al., 2004; Tripp, 2001). Research on the various dimensions of th e criminal career needs to be continued. This type of research needs to be supported by universities and federal grants, as it can assist policy makers and law enforcement in making difficult and ch allenging decisions within the criminal justice system. For exampl e, this research revealed that periods of desistance at various stages of the life course can be linked to a variet y of different social bonds. Policy makers should pay attention to the ways that a br oad array of social institutions can lead to desistance from crim e. Additionally, the diverse experiences of Americans by race, gender, and socioeconomic status could be taken as an important point from this research. As marriage wa s not influential of desistance among poor African American men, perhaps government programs built to foster marriage as a panacea to social problems may not be a heal all, given the diverse population of the United States.

PAGE 239

226 Future Research This project represents the tip of the iceberg as far as the knowledge that can be obtained from this data about criminal participation and de sistance among Harlem residents from 1968 through 1994. Future research on this data set could follow a myriad of directions. This research was limited to an examination of the influence of social bonds on criminal behavior among men. Th ere were 317 women in Wave 1. The research questions addressed in this proj ect could be applied to women, followed by comparisons of the two samples. This would be especially applicab le to the questions on drug use as the female participants were more highly represented among the drug users than among those arrested or incarcerated. Beyond the use of the female participan ts, there are many more avenues of exploration available within this data set. First, wh ile this project focused on participation in criminal behaviors, it is possible to examine the frequency of these behaviors in various manners. As noted above , the measures of arre sts and incarceration vary through the project, and because of these differences it was easier to examine participation. However, there are various counts of the frequency of arrests and incarceration. Through a series of years participants were aske d if they were incarcerated or arrested either: Zero times, once, or mo re than once. While limited, these measures could be used to examine the frequency and seriousness dimensions of the participants’ criminal careers. Another r oute through which both the seve rity and the frequency of criminal behavior could be examined is th rough the widespread measures of drug use. The frequency of drug use could be measured in days, weeks, or years in multiple waves. While this project examined drug use as a dichotomous variable, there are numerous ways in which drug use could be accounted for. Another means through which drug use

PAGE 240

227 could be investigated would be to examin e the various drugs included in the data separately. While this project grouped all of the drugs into one category, an investigation of different drugs would offer a great deal of insight, especially in regards to desistance and seriousness. Desistance was conceptualized as non-use in this pr oject. Desistance could be concpetualized as a change, or de-e scalation of the severi ty of the drug used (LeBalnc and Loeber, 1998). Future research could focus on changes in social bonds from different angles. This project looked at the possession of social bonds, and the quality of the some of the bonds. The loss of such bonds is one manner in wh ich this research could be expanded. Measures of divorce, job loss, and residential patterns in different years could allow for various models that examine the impact of th e loss of social bonds on criminal behavior. Research questions regarding th e impact of divorce, job loss, and the changing residential patterns of men as husband, fathers and partne rs could be examined. Additionally, as mentioned above, the times at which men attain ed, maintained, and/or lost such social bonds could be explored as infl uential factors on the criminal behaviors of interest. The influence of marriage, fatherhood, and co-resi dence at specific ages could provide new insights into the criminal behaviors of these men. The length of relationships such as ma rriages, co-residence, and fatherhood are various ways in which the strength of social bonds could be further explored. Another manner in which the strength of social bonds could be expanded in this research is through various measures of employment and income. In this project employment was a dichotomous measure of whether or not the participant was employed at the time of his interview. Future research could look at th e different influences th at various levels of

PAGE 241

228 work hold for these men. These men were also asked if they worked full or part-time. Other items asked the participants how many hours per week and how many hours per month they worked. These measures of employment could be compared with models that used employment (dichotomous) and job satisf action to examine if the amount of work each of the men completed (according to self -reports) during this study would influence criminal behaviors in different manner than the previously used measures. Criminal career and desistance research need s more in-depth qualitative studies that take what has been learned in quantita tive studies and expand that knowledge into questions that ask how these results come about . The meanings that offenders attach to the various dimensions of the criminal career would be an excellent place to start. How do these men describe their “onset,” or their “e scalation?” In regard s to the social bonds that have been found to influence desistan ce, how do these men describe these bonds and these relationships? One dire ction in which desistance res earch should move is towards an examination of the people in desistor’s lives. Just as co-offending is examined, perhaps co-desistors should be examined. How do the wives or partners of these men describe desistance and the mechanisms that they believe have influenced changes in these men’s life courses? These are a fe w of the questions that could begin an informative line of study on the desistance from criminal behavior. The finally, research utilizing this data should examine cumulative disadvantage among the participants. Cumulative disadvant age refers to the manner in which antisocial behavior has a “systematic attenuating effect on the social and institutional bonds linking adults to society.” (Sampson and Laub, 1997: 144). Anti-so cial behavior can remove people from holding and developing th e types of informal social bonds that

PAGE 242

229 influence desistance from criminal behavior. Future research could examine the ways that the various forms of anti-social behavior captured in this data set disrupted the creation and maintenance of positive social bonds such as marriage and employment.

PAGE 243

230 APPENDIX A direct logistic regressi on was conducted on the outcome variable of Wave 5 Drug Use and the following Wave 5 predictors: Ag e, Employment Status, Income, Welfare Use, Job Satisfaction, Attainment of a High Sc hool Diploma, Time at Current Residence, Fatherhood, Residential Fatherhood, Marital Status, Prior Drug Use, and Prior Incarceration. The test of the full model (all predictors) against a constant only model was significant (Chi-square = 25.46, df = 12, p < .05) indicating that th e set of Wave 5 predictors reliably distinguish ed between participants who did and did not use drugs. Table A-1 presents the regression coeffi cients where Job Satisfaction and Prior Incarceration predicted Wave 5 Drug Use (N = 168). Men who were previously incarcerated were 81.6% less likely to use drugs than men who had not been incarcerated during the previous Wave. For every one unit increase in Job Satisfaction, a participant was 33.5% less likely to use drugs during Wave 5. A direct logistic regre ssion was conducted on the outco me variable of Wave 5 Arrests and the following Wave 5 predictors: Age, Employme nt Status, Income, Welfare Use, Job Satisfaction, Attainment of a High Sc hool Diploma, Time at Current Residence, Fatherhood, Residential Fatherhood, Marital Status, Prior Drug Use, and Prior Incarceration. The test of the full model (all predictors) against a constant only model was significant (Chi-square = 36.53, df = 12, p < .05) indicating that th e set of Wave 5

PAGE 244

231 predictors reliably distinguished between participants who we re and were not arrested. Table A-2 presents the re gression coefficients wher e Fatherhood, and Previous Incarceration reliably predicted Wave 5 A rrests (N = 168). Fathers had odds of incarceration that were 72.9% lower than me n who were not fathers. Men who were previously incarcerated had odds of arrest th at were 11.11 greater than men who were not previously incarcerated. Table A-1 Wave 5 Independent Vari ables Predicting Wave 5 Drug Use VARIABLES B S.E Wald Sig. Exp(B) Age -.125 .125 .996 .318 .883 Employed -.619 .479 .1670 .196 .538 Income .069 .122 .326 .568 1.072 Welfare Use .475 .467 1.037 .308 1.608 HS Degree -.854 .509 2.817 .093 .426 Job Sat. -.439 .209 4.402 .036 .645 Time at Res. .005 .103 .003 .958 1.005 Father .289 .443 .427 .514 1.335 Res. Father -.852 .499 2.913 .088 .427 Married .000 .448 .000 1.000 1.000 Prior Drug Use .343 .589 .340 .560 1.409 Prior Incarceration -1.692 .725 5.441 .020 .108 CONSTANT 5.720 4.734 1.460 .227 304.812 A direct logistic regre ssion was conducted on the outco me variable of Wave 5 Incarceration and the following Wave 5 predictors: Age, Employment Status, Income, Welfare Use, Job Satisfaction, Attainment of a High School Diploma, Time at Current

PAGE 245

232 Residence, Fatherhood, Residential Fatherhood, Ma rital Status, Prior Drug Use, and Prior Incarceration. Table A-2 Wave 5 Independent Vari ables Predicting Wave 5 Arrests VARIABLES B S.E Wald Sig. Exp(B) Age -.149 .165 .811 .368 .862 Employed -.842 .565 2.220 .136 .431 Income .156 .151 1.058 .304 1.169 Welfare Use -.028 .559 .003 .960 .972 HS Degree 1.471 .779 3.565 .059 4.355 Job Sat. -.351 .281 1.563 .211 .704 Time at Res. .185 .133 1.946 .163 1.203 Father -1.307 .623 4.406 .036 .271 Res. Father -1.250 .663 3.555 .059 .286 Married -.148 .552 .071 .789 .863 Prior Drug Use -1.295 1.112 1.355 .244 .274 Prior Incarceration 2.408 .673 12.793 .000 11.111 CONSTANT 4.378 6.279 .486 .486 79.718 The test of the full model (all predictors) against a constant only model was significant (Chi-square = 88.75, df = 12, p < .05) indicating that th e set of Wave 5 predictors reliably distingui shed between participants who were and were not incarcerated. Table A-3 presents the regres sion coefficients where Prior Incarceration reliably predicted Wave 5 Incarceration (N = 168). Men who were previously incarcerated had odds of Wave 5 incarcerati on that were 533.29 gr eater than men who were not previously incarcerated.

PAGE 246

233 Table A-3 Wave 5 Independent Variab les Predicting Wave 5 Incarceration VARIABLES B S.E Wald Sig. Exp(B) Age -.114 .237 .233 .630 .892 Employed -1.404 .825 2.895 .089 .246 Income -.002 .209 .000 .992 .998 Welfare Use -.963 .931 1.069 .301 .382 HS Degree 2.455 1.404 3.057 .080 11.647 Job Sat. -.412 .430 .919 .338 .662 Time at Res. -.001 .179 .000 .995 .999 Father -1.945 1.009 3.714 .054 .143 Res. Father -1.672 1.021 2.686 .101 .188 Married .378 .805 .220 .639 1.459 Prior Drug Use -.873 1.213 .518 .472 .418 Prior Incarceration 6.279 1.444 18.902 .000 533.285 CONSTANT 4.283 9.208 .216 .642 72.483

PAGE 247

234 LIST OF REFERENCES Achatz, M., and MacAullum, C. 1994. Young Unwed Fathers, Report from the Field . Philadelphia, PA: Public/Private Ventures. Agnew, R. 1985. A revised st rain theory of delinquency. Social Forces , 64: 151-167. Akers, R.L. 1997. Criminological Theories: Introduction and Evaluation . Second Edition. Los Angeles, CA: Roxbury Publishing Company. Akers, R.L., Krohn, M.D., Lanza-Kaduce, L.L., and Radosevich, M. 1979. Social Learning and Deviant Behavior: A Speci fic Test of a General Theory. American Sociological Review , 44: 636-655. Anderson, E. 1993. Sex Codes and Family Life among Poor Inner-city Youths. In R. Lerman and T Ooms (eds) Young Unwed Fathers: Changing Roles and Emerging Policies . Philadelphia, PA: Temple University Press. Arditti, J. 2002. Ecological nightmares of the worst kind: Families and Incarceration. Paper presented at the National Confer ence on Family Relations, Houston, TX. Arditti, J., Lambert-Shute, J., and Joest, K. 2002. Saturday morning at the jail: Implications of incarceration for familie s and children. (Manuscript under review) Austin, J., and Irwin, J. 2001. It’s About Time: America’s Imprisonment Binge . Second Ed. Belmont, CA: Wadsworth. Bachman, J.G. 1986. Effects of Earl y Marriage on Substance Abuse. Medical Aspects of Human Sexuality , 20 (10), 15. Baldwin, P. 1990. The Politics of Social Solidarity: Class Bases of the European Welfare State . Cambridge: Cambridge University Press. Barnett, A., Blumstein, A., and Farrington, D.P. 1987. Probabilistic Models of Youthful Criminal Careers. Criminology , 25: 83-107. Barnett, A., Blumstein, A., and Farrington, D.P. 1989. A Prospective Test of a Criminal Career Model. Criminology , 27: 373-388. Beck, U. 1992. Risk Society: Towards a New Modernity . London: Sage. Becker, H.S. 1963. Outsiders: Studies in the Sociology of Deviance . New York: Free Press.

PAGE 248

235 Beckett, K. 1997 . Making Crime Pay: Law and Order in Contemporary American Politics. New York: Oxford University Press. Belsky, J., and Miller, B.C. 1986. Adolsecent Fatherhood in the Contxt of the Transition to Parenthood. In A.B. Elster and M. Lamb (esd) Adolescent Fatherhood . Hillsdale, NJ: Lawrence Erlbaum Associates. Bilick, S., Seymour, C., and Kreisher, K. 2001. Parents in prison. Corrections Today , 63: 108-114. Blankenhorn, D. 1995. Fatherless America: Confrontin g our most urgent social problem . New York: Basic Books. Bloom, B. 1995. Public policy and the children of incarcerated parents. In K. Gabel and D. Johnson (eds) Children of Incarcerated Parents . New York: Lexington. Blumstien, A., and Beck, A. 1999. Populati on Growth in U.S. Prisons,1980-1996. In M. Tonry and J Petersilia (ed) Prisons (Crime and Justice: A Review of the Research , vol 26. Chicago: University of Chicago Press. Blumstein, A., and Cohen, J. 1979. Estimation of Individual Crime rates from Arrest Records. Journal of Criminal Law and Criminology , 70: 561-585. Blumstein A., Cohen, J., Roth, J., and Visher, J. (eds.). 1986. Criminal Careers and Career Criminals . Washington D.C.: National Academy of Sciences. Blumstein, A., Cohen, J., and Farrington, D.P. 1988a. Criminal career research: Its value for criminology. Criminology , 26, 1: 1-36. Blumstein A., Cohen, J., and Farrington, D.P. 1988b. Longitudinal and criminal career research: Further clarifications. Criminology , 26, 1: 57-74. Bottoms, A. 1995. The philosophy and politic s of punishment and sentencing. In C. Clarkson and R. Morgan, The Politics of Sentencing Reform . New York: Oxford University Press. Bound, J., and Freeman, R.B. 1992. What went Wrong? The Erosion of Relative Earnings and Employment among Young Black Men in the 1980’s. Quarterly Journal of Economics , 107: 201-232. Bourgois, P.I. 1995. In Search of respect: Selling Crack in El Barrio . New York: Cambridge University Press. Bozett, F.W. 1985. Male Development and Fathering throughout the Life Cycle. American Behavioral Scientist , 29 (1): 41-54. Brame, R. and Piquero, A.R. 2003. Selective Attrition and the Age-Crime Relationship. Journal of Quantitative Criminology , 19, 2.

PAGE 249

236 Braithwaite, J. 1989. Crime, Shame, and Reintegration . Cambridge: Cambridge University Press. Brenner, T. 1992. The legalization of drugs : Why prolong the inevitable? In R. Evans and I. Berent (eds) Drug Legalization: For and Against . Chicago: Open Court. Brodsky, S.L. (1975). Family and friends of men in prison . Lexington, MA: Lexington Books. Brunswick, A.F. 1976. Indicators of H ealth Status in Adolescence. International Journal of Health Services , 6 (3): 475-492. Brunswick, A.F. 1980. Social Meanings a nd Developmental Needs: Perspectives on Black Youths’s Drug Abuse. Youth and Society , 16 (1): 3-28. Brunswick, A.F. 1984. Health Consequences on Drug Use: A Longitudinal Study of Urban Black Youth. In S.A.Mednick, M. Haraway, & K.M. Finello (Eds) Handbook of :Longitudinal Research Vol. 2, Teenage and Adult Cohorts . Brunswick, A.F. and Boyle, J. 1979. Patte rns of Drug Involvement: Development in Secular Influences on age at Inititiation. Youth and Society , 11 (2): 139-162. Brunswick, A.F. and Messeri, P. 1986. Path ways to Heroin Abstinence: A Longitudinal Study of Urban Black Youth. Advances in Alcohol and Substance Abuse , 5: 111135. Brunswick, A.F., Merzel, C. and Messeri, P. 1985. Drug Initiation Among Urban Black Youth. A Seven Year Follow Up of De velopmental and Secular Influences. Youth and Society , 17 (2): 198-216. Burchell, G., Gordon, C., and Miller, P. 1991. The Foucualt Effect in Studies of Governmentality . Chicage: University of Chicago Press. Bushway, S.D., Piquero, A.P., Broidy, L.M., Caufmann, E., and Mazerolle, P. 2001. An Empirical Framework for Studying Desistance as a Process. Criminology , 39: 491-515. Caplow, T., and Simon, J. 1999. Understand ing prison policy and population trends. In M. Tonry and J Petersilia (ed) Prisons (Crime and Justice: A Review of the Research , vol 26. Chicago: University of Chicago Press. Carlos, B.S., and Cevera, N. (1991). Inmates and their families. Criminal Justice and Behavior , 13, 318-339. Carrabine, E., and Longhurst, B. 1998. Gender and Prison Organization: Some Comments on Masculinities and Prison Management. Howard Journal of Criminal Justice , 37 (2): 161-176.

PAGE 250

237 Caspi, A., and Moffitt, T. 1993. Continuity Amidst Change: A Paradoxical Theory of Personality Coherence. Psychological Inquiry . Caspi, A., and Moffitt, T. 1995. The continuity of maladaptive behavior: From description to understanding in the study of antisocial behavior. In D. Cicchetti and D. Cohen (eds.) Manual of Developmental Psychopathology . Catalano, R.F., and Hawkins, J.D. 1996. The Social Developmental Model: A Theory of Antisocial Behavior. In J.D. Hawkins (ed) Delinquency and Crime: Current Theories . Cambridge: Cambridge University Press. Cernkovich, SA, Giordano, P.C., and Pugh, M. D. 1985. Chronic Offenders: The Missing cases in Self-Report Delinquency Research. The Journal of Criminology and Law 76, 3. Chambliss, W. 1999. Power, Politics, and Crime . Boulder, CO: Westview Press. Chaiken, J.M., and Chaiken, M.R. 1982. Varieties of Criminal Behavior . Santa Monica, CA: RAND. Report Number: R-2814-NIJ. Chen, K., and Kandel, D.B. 1998. The Natu ral History of Drug Use from Adolescence to the Mid-thirties in a General Population sample. American Journal of Public Health , 85: 41-47. Christmon, K., and Luckey, I. 1994. Is Ea rly Fatherhood Associated with Alcohol and Other Drug Use? Journal of Substance Abuse , 6 (3): 337-343. Clear, T. 1994. Harm in American Penology: Offenders, Victims, and their Communities . New York: State University of New York Press. Clemmer, D. 1940. The Prison Community . New York: Holt, Rinehart, and Winston. Cline, H.F. 1980. Criminal Behavior over the Life Span. In O.G. Brim and J Kagan (eds) Constancy and Change in Human Development . Cambridge, MA: Harvard University Press. Cloward, R.A., and Ohlin, L.E. 1960. Delinquency and Opportunity: A Theory of Delinquent Gangs . Glencoe, IL: Free Press. Crutchfield, R.D., and Pitchford, S.R. 1997. Work and Crime: The Effects of Labor Stratification. Social Forces , 76: 93-118. Currie, E. 1985. Confronting Crime: An American Challenge . New York: Pantheon. Cordray, S., and Polk, K. 1993. The Implica tions of Respondent Loss in Panel Studies of Deviant Behavior. Journal of Research on Crime and Delinquency , 20: 214242.

PAGE 251

238 Cohen, A.K. 1955. Delinquent Boys . Glencoe, IL: FreePress Cohen, P.N. 1999. Racial-Ethnic and Gender Di fferences in returns to Cohabitation and Marriage: Evidence from the Current Population Survey. U.S. Census Bureau, Population Division Working Paper no. 35 . Washington, D.C.: U.S. Census Bureau. Coleman, J.S. 1988. Social Capital in the Creation of Human Capital. American Journal of Sociology , 94: 95-120. Coleman J.S. 1990. Foundations of Social Theory . Cambridge, MA: Harvard University Press. Cooley, C.H. 1902. Human Nature and Social Interaction . New Brunswick, N.J.: Transaction Books. Cooley, C.H. 1909. Social Organization: A Study of the Larger Mind . New York: C. Scribner’s Sons. Coombs, R. 1991. Marital Status and Pe rsonal Wellbeing: A Literature Review. Family Relations , 40: 97-102. Cowan, C.P., and Cowan, P.A. 1992 . When Partners become Parents: The Big Life Change for Couples . New York: Basic Books. Cowan, C.P., Cowan, P.A., Heming, G., Garrett, E., Coyish, W.S., Curtis-Boles, H., and Boles, A.J. 1985. Transition to Parenthood: His, Hers, and Theirs. Journal of Family Issues , 6: 451-481. Danziger, S.K., and Radin, N. 1990. Absent does not Equal Uninvolved: Predictors of Fathering in Teen Mother Families. Journal of Marriage and Family , 52: 636642. Dearden, K.A., Hale, C.B., and Woolley, T. 1995. The Antecedents of Teen Fatherhood: A Retrospective Case-control St udy of Great Britain Youth. American Journal of Public Health , 85: 551-554. Donziger, S. (ed) 1996. The Real War on Crime: The Report of the National Criminal Justice Commision . New York: Harper Perrinial. Douglas, M. and Wildavsky, A. 1982. Risk and Culture . Berkely: University of California Press. Doyle, J.M., Ahmed, E., and Horn, R.N. 1999. The Effects of Labor Markets and Income Inequality on Crime: Evidence from Panel Data. Southern Economic Journal , 65 (4): 717-731.

PAGE 252

239 Drucker, E. 2002. Population Impact of Mass Incarceration under New York’s Rockefeller Drug Laws: An Anal ysis of Years of Life Lost . Journal of Urban Health: Bulletin of the New York Academy of Medicine, 79 (3): 1-9. Durkheim, E. 1893. The Division of Labor in Society . New York: Free Press. Durkheim, E. 1895. The Rules of Sociological Methods . New York: Free Press. Durkheim, E. 1897. Suicide . New York: Free Press. Durkheim, E. 1918. Montesquieu and Rousseau: Forerunners of Sociology . AnnArbor: University of Michigan. Duster, T. 1996. Pattern, Purpose, and Race in the Drug War: The Crisis of Credibility in Criminal Justice. In C. Reinarman and H.G. Levine (eds) Crack in America: Demon Drugs and Social Justice . Berkeley, CA: Universi ty of California Press. Ebbsen, F., and Elliott, D.S. 1994 . Conti nuity and Discontinuity in Illicit Drug Use: Patterns and Antecedents. Journal of Drug Issues , 24: 75-97. Edin, K., Nelson, T.J., and Paranal, R. 2004. Fatherhood and Incarceration as Potential Turning Points in the Criminal Careers of Men. Patillo, M., Weinman, D., and Western, B. Imprisoning America: The Social Effects of Mass Incarceration . New York, NY: Russel Sage Foundation. Eggebeen, D. 2002. The Changing Course of Fatherhood: Men’ s Experiences with Children in Demographic Perspective. Journal of Family Issues , 23 (4): 486-506. Ekland-Olson, S., Supanic, M., Campbell, J., and Lenihan, K.J. 1983. Postrelease Depressions and the Importance of Familial Support. Criminology , 21: 253-275. Elder, G. 1985. Life Course Dynamics . Ithaca, NY: Cornell University Press Elder, G. 1986. Military times and turning points in men’s lives. Developmental Psychology , 22: 233-245. Elder, G. 1994. Time, human agency, and social change. Social Psychology Quarterly , 57, 1: 4-15. Elder, 2000. The life-course. In E.F. Borgatta and R. Montgomery (eds.) The Encyclopedia of Sociology . Second Edition. New York: Macmillan Reference USA. Elder, G., Modell, J., and Parke, R. (eds.). 1993. Children in Time and Place: Developmental and Historical Insights . New York: Cambridge University Press. Elliott, D.S. 1994. Serious Violent Offenders: Onset, Developmental Course, and Termination. Criminology , 32: 1-22.

PAGE 253

240 Elliott, D.S., Ageton, S.S., Huzinga, D., Knowles, B., and Canter, R. 1983. The Prevalence and Incidence of Delinquent Behavior 1976-1980. National Youth Survey, Report no. 26 . Boulder, CO: Behavioral Research Institute. Elliott, D.S., and Ageton, S.S. 1980. Reconc illing Race and Class Differences in SelfReported and Official Es timates of Delinquency. American Sociological Review , 45 (1): 95-110. Elliott, D.S., Huzinga, D., and Ageton, S.S. 1985. Explaining Delinquency and Drug Use . Beverly Hills, CA: Sage. Elliott D.S. and Voss, H.L. 1974. Delinquency and Dropout . Lexington, MA: D.C. Health. Elster, A.B. and Lamb, M.E. 1987. Adolescent Fatherhood . Hillsdale, NJ: Erlbaum. Elster, A.B., Lamb, M.E., and Tavare, J. 1987. Association between Behavioral and School Problems and Fatherhood in a Nati onal Sample of Adolescent Youths. Journal of Pediatrics , 111: 932-936. Fagan, J. 1989. Cessation of Family Violence : Detterence and Dissuation. In L. Ohlin and M. Tonry (eds) Family Violence, Vol. 11 of Crime and Justice: Review of Research . Chicago, IL: Universi ty of Chicago Press. Farrington, D.P. 1979. Longitudinal Research on Crime and Delinquency. In N. Morris and M. Tonry (eds) Crime and Justice: An Annual Review of Research, Vol. 1 . Chicago, IL: University of Chicago Press. Farrington, D.P. 1986. Age and crime. In M . Tonry and N. Morris (eds) Crime and Justice: An Annual Review of Research, vol 7. Chicago:University of Chicago Press. Farrington, D.P. 1987. Predic ting Individual Crime Rates. In D. Gottfredson and M. Tonry (eds) Prediction and Classification: Crim inal Justice Decision Making Vol. 9 . Chicago, IL: University of Chicago Press. Farrington, D.P. 1988. Studying changes within i ndividuals: The causes of offending. In M. Rutter (ed.) The Power of Longitudinal Data . Cambridge: Cambridge University Press. Farrington, D.P. 2000. Expl aining and Preventing Crim e: The Globalization of Knowledge-the American Society of Criminology 1999 Presidential Address. Criminology , 38: 1-24. Farrington, D.P. 2003. Developmental and Life-Course Criminology: Key Theoretical and Empirical Issues – The 2002 Sutherland Award Address. Criminology 41, 2: 221-256.

PAGE 254

241 Farrington, D.P. and Hawkins, J.D. 1991. Predicting Participation, Early Onset, and Later Persistence in Officially Recorded Offending. Criminal Behavior and Mental Health , 1: 1-33. Farrington, D.P., and West, D.J. 1995. “Eff ects of Marriage, Separation, and Children on Offending by Adult Males.” In Z.B. Smith and J. Hagan (eds.) Current Perspectives on Aging and the Life-Course, vol. 4 . Greenwich, CT: JAI Press. Farrington, D.P., Ohlin, L.E ., and Wilson, J.Q. 1986. Understanding and Controlling Crime . New York: Springer-Verlag. Farrington, D.P., and Tarling, R. 1985. Prediction in Criminology . Albany: State University of New York Press. Feeley, M., and Simon, J. 1992. The new pe nology: Notes on the emerging strategy of corrections and its implications. Criminology , 30: 449-474. Foucault, M. 1977. Discipline and Punish: The Birth of Prison . New York: Pantheon. Freeman, R.B. 1995. The Limits of Wage Flexibility to Curing Unemployment. Oxford Review of Economic Policy , 11:63-72. Furstenberg, F.F. Jr. 1995. Fathering in the I nner City: Paternal Par ticipation and Public Policy. In W. Marsiglio (ed) Fatherhood: Contemporary Theory, Research, and Social Policy . Thousand Oaks, CA: Sage. Furstenburg, F.F. Jr. and Cherlin, A. 1991. Divided families . Cambridge, MA: Harvard University Press Gadsen, V.L., and Smith, R.R. 1994. African American Males and Fatherhood: Issues in Research and Practice. Journal of Negro Education , 63, 4: 634-648. Garfinkle, I., and McLanahan, S. 1986. Single Mothers and their Children: A New American Dilemma . Washington, D.C.: Urban Institute. Garland, D. 1990. Punishment in Modern Society . Chicago: University of Chicago Press. Garland, D. 2001. Mass Imprisonment: Social Causes and Consequences. London, UK: Sage Publications Geronimus, A.T. 1991. Teenage Chil dbearing and Social and Reproductive Disadvantage: The Evolution of Complex Questions and the Demise of Simple Answers. Family Relations , 40: 463-471. Gibbens, T.C.N. 1984. Borstal boys after 25 years. British Journal of Criminology , 24: 49-62.

PAGE 255

242 Gibbons, D.C. 1979. The Criminological Enterprise: Theories and Perspectives . Upper Saddle River, NJ: Prentice Hall. Giddens, A. 1991. The Consequences of Modernity . Stanford, CA: Stanford University Press. Giele, J.A., and Elder Jr., G.H. 1998. Methods of Life-Course Research: Qualitative and Quantitative Approaches . Thousand Oaks, CA: Sage. Glueck, S., and Glueck, E. 1968. Delinquents and Nondelinquents in Perspective . Cambridge: Harvard University Press. Goffman, E. 1963. Behavior in Public Places: Notes on the Social Organization of Gatherings . New York: Free Press. Goffman, E. 1963. Stigma: Notes on the Managem ent of Spoiled Identity . Englewood Cliffs, NJ: Prentice-Hall. Gordon, D. 1994. The Return of the Dangerous Cl asses: Drug Prohibition and Policy Politics . New Brunswick, NJ: Rutgers University Press. Goodwin, C.J. 2001. Research in Psychology, 3rd edition. New York: John Wiley & Sons, Inc. Goring, C. 1913. The English Convict . Montclair, NJ: Patterson Smith. Gottfredson, M., and Gottfredson, D. 1988. Decisionmaking in Criminal Justice (2nd ed.). Cambridge, MA: Ballinger. Gottfredson, M., and Hirschi, T. 1986. The true value of lambda would appear to be zero: An essay on career criminals, cr iminal careers, selective incapacitation, cohort studies, and related topics. Criminology , 24: 213-234. Gottfredson, M., and Hirschi, T. 1987. The methodology of longitudinal research on crime. Criminology , 25, 3. Gottfredson, M., and Hirschi, T. 1988. Scie nce, public policy, and the career paradigm. Criminology , 26, 1: 37-56. Gottfredson, M., and Hirschi, T. 1990. A General Theory of Crime . Stanford: Stanford University Press. Gove, W.R. 1985. The Effect of Ag e and Gender on Deviant Behavior: A Biopsychological Perspective. In A.S. Rossi (ed) Gender and the Life Course . New York: Aldine. Gove, W., Style, C., and Hughes. M. 1990. The effect of marriag e on the well being of adults. Journal of Family Issues , 11, 1: 4-35.

PAGE 256

243 Grogger, J. 1995. The Effects of Arrests on the Employment and Earnings of Young Men. Quartely Journal of Economics , 110: 51-71. Guttman, L. 1977. What is not in Statistics. The Statistician , 26: 81-107. Hagan, J. 1991. Destiny and Drift: Subcultura l Preferences, Status Attainments, and the Risks and Rewards of Youth. American Sociological Review , 96: 265-299. Hagan J., and Palloni, A. 1988. Crimes as so cial events in the life course: Reconcieving a criminological controversy. Criminology , 26, 1: 87-100. Hagan, J., and Wheaton, B. 1993. The Search for Adolescent Role Exits and the Transition to Adulthood. Social Forces , 71 (4): 980-995. Hagan, J., and McCarthy, B. 1997. Mean Streets: Youth, Crime, and Homelessness . New York: Cambridge University Press. Hardy, J.E., and Duggan, A.K. 1988. Teenag e Fathers and the Fathers of Infants of Urban, Teenage Mothers. American Journal of Public Health , 78: 919-922. Hawkins J.D., and Catalano, R.F. 1992. Communities that Care: Action for Drug Abuse Prevention . San Francisco, CA: Jossey-Bass. Heath, D.H. 1978. What Meaning and Effect does Fatherhood have for the Maturing of Professional Men? Merrill-Palmer Quarterly , 24 (2): 265-278. Heath, D.H., and Heath, H.E. 1991. Fulfilling Lives: Paths to Maturity and Success . San Francisco: Jossey-Bass. Hindelang, M.J. 1981. Variations in Se x-Race-Age-Specific Incidence Rates of Offending. American Sociological Review , 46: 461-474. Hindelang, M.J., Hirschi, T., and Weis, J. 1979. Correlates of delinquency: The illusion of discrepancy between self re ports and official measures. American Sociological Review , 44: 995-1014. Hindelang, M.J., Hirschi, T., and Weis, J. 1981. Measuring Delinquency . Beverly Hills, CA: Sage. Hirschi, T. 1969. Causes of Delinquency . Berkeley, CA: Univers ity of California Press. Hirschi, T. 1983. Crime and the Family. In J.Q. Wilson (ed) Crime and Public Policy . San Fransico: Institute fo r Contemporary Studies. Hirschi, T., and Gottfredson, M. 1983. Age a nd the explanation of crime. American Journal of Sociology , 89: 552-584.

PAGE 257

244 Hirschi T., and Gottfredson, M. 1986. The dis tinction between crime and criminality. In T. Hartnagel and R. Silverman (eds.), Critique and Explanation . New Brunswick, NJ: Transaction Books. Hochschild, J. 1995. Facing up to the American Dream: Ra ce, Class, and the Soul of the Nation. Princeton, NJ: Princeton University Press. Horney, J.D., Osgood, D.W., and Marshall, I. H. 1993. Variablility in Crime and Local Life Circumstances. American Sociological Review , 60 (5): 655-673. Howel, F.E., and Frese, W. 1982. Adult Role Transitions, Parental Influence, and Status Aspirations Early in the Life Course. Journal of Marriage and the Family , 44 (1): 35-49. Huber, J., and Spitze, G. 1988. Trends in Family Sociology. In N.J. Smelser (ed) Handbook of Sociology . Newbury Park, CA: Sage. Hughes, M. 1998. Turning Points in th e Lives of Young Inner-City Men Forgoing Destructive Criminal Behavi ors: A Qualitatative Study. Social Work Research , 22: 143-151. Huizinga, D., and Elliot, D.S. 1984. Reassessing the reliability and validity of self report delinquency measures. Journal of Quantitative Criminology , 2: 293-327. Huzinga, D. and Elliot, D.S. 1986. Re-assessin g the Reliability and Validity of SelfReport Delinquency Measures. Journal of Quantitative Criminology , 2: 293-327. Ingram, H. 1993. Social construction of targ et populations: Implicat ions for politics and policy. American Sociological Review , 87, 2: 334-348. Irwin, J. and Austin, J. 1997. Its about Time: Americ a’s Imprisonment Binge . Belmont, CA: Wadsworth. Jackson, R. 1990. Children as Ethnographers. In R. Jackson and D. Starking (eds.) The Junior RE Handbook . Cheltenham: Stanley-Thornes. Jancovic, I. 1977. Labor Markets and Imprisonment. Crime and Social Justice , 8: 1731. Jackman, M. 1994. The Velvet Glove: Paternalism and Conflict in Gender, Class, and Race Relations . Berkeley: University of California Press. Jang, S.J. 1999. Different definitions, Di fferent Modeling Decisions, and Different Interpretations: A Rejoinder to Lauritsen. Criminology , 37: 695-702. Janowitz, M. 1975. Sociological Th eory and Social Control. American Journal of Sociology , 81: 82-108.

PAGE 258

245 Jenkins, P. 1998. Moral Panic: Changing Concepts of the Child Molester in Modern America . New Haven, CN: Yale University Press. Johnson, D., and Gabel, K. 1995. Incarcerated pa rents. In K. Gabel and D. Johnson (eds) Children of Incarcerated Parents . New York: Lexington. Johnson, R.E. 1979. Juvenile Delinquency in its Origin s: An Integrated Theoretical Approach . New York: Cambridge University Press. Jones, R. 2003. Excon: Managing a spoiled id entity. In J. Ross and S. Richards (eds) Convict Criminology . Belmont, CA: Thomas-Wadsworth. Jordan, W.J. 1996. Role Transitions: A Revi ew of the Literature [LR-RT-96-06]. Philadelphia: National Center on Father s and Families, Graduate School of Education, University of Pennsylvania. Junger-Tas, J., and Marshall, I.H. 1999. The Self-report Methodology in Crime Research. In M. Tonry (ed) Crime and Justice: An A nnual Review of research, Vol. 25 . Chicago: University of Chicago Press. Ketterlinus, R.D., Lamb, M.E., and Nitz, K. 1992. Adolescent Non-sexual and sexrelated Problem Behaviors. Journal of Adolescent Research , 7: 431-456. King, A.E.O. (1993). The impact of incarceration on African American families: Implications for practice. The Journal of Contemporary Human Services , Spring, 1993 Klein, B. 1985. Fathering: The First Time. Social Work , 30 (3): 264-267. Klein, M.W. 1989. Cross-National Research in Se lf-Reported Crime and Delinquency . Los Angeles: Kluwer Academic Publishers. Knight, B.J., Osborn, S.G., and West, D. 1977. Early marriages and criminal tendency in males. British Journal of Criminology , 17: 348-360. Korenman, S., and Neumark, D. 1991. Does Marriage Really Make Men more Productive? Journal of Human Resources , 26: 282-307. Kornhauser, R. 1978. Social Sources of Delinquency . Chicago: University of Chicago Press. Krohn, M.D., and Thornberry, T.P. 1999. Reten tion of Minority Populations in Panel Studies of Drug Use. Drugs and Society , 14: 185-207. Labouvie, E. 1996. Maturing Out of Substan ce Use: Selection and Self Correction. Journal of Drug Issues 26: 457-476.

PAGE 259

246 LaFree,G., and Drass, K.A. 1990. The Eff ect of Changes in Tntraracial Income Inequality and Educational Attainment on Changes in Arrest Rates for African Americans and Whites, 1957 to 1990. American Sociological Review , 61: 614634. Lamb, M.E., Pleck, J.H., Charnov, E.L., & Levi ne, J.A. (1985). Paternal Behavior in Humans. American Zoologist , 25, 884-894. Lamb, M.E., Pleck, J.H., Charnov, E.L., & Levi ne, J.A. (1987). A Biosocial Perspective on Paternal Behavior and Involvement. In Lancaster, J.B., Altmann, J., Rossi, A.S., & Sherrod, L.R . Parenting Across The Lifespan: Biosocial Dimensions . Hawthorne, NY: Aldine. Lanier, Jr., C.S. 1993. Affective states of fathers in prison. Justice Quarterly , 10, 49-65. Laub, J. H., D. S. Nagin, and Sampson, R.J. 1998. Trajectories of Change in Criminal Offending: Good Marriages and the Desistance Process. American Sociological Review 63:225. Laub, J.H, and Sampson, R.J. 1988. Unraveling Families and Delinquency: A Reanalysis of the Gluecks’ Data. Criminology 26, 3: 355-380. Laub, J.H, and Sampson, R.J. 1993. Turning Points in the Life-Course: Why Change Matters to the study of Crime. Criminology 31, 3: 301-326. Laub, J.H, and Sampson, R.J. 2001. Understand ing Decisitance for Crime. In M. Tonry (ed) Crime and Justice: An Annual Review of research, Vol. 28 . Chicago: University of Chicago Press. Lauritsen, J. 1998. The Age-Crime Debate: As sessing the Limits of Longitudinal SelfReport Data. Social Forces , 77: 127-155. LeBlanc, M. 1997. A Generic Control Theory of the Criminal Phenomenon: The Structural and Dynamic Statements of an Integrated Multilayered Control Theory. In T.P. Thornberry (ed) Developmental Theories of Crime and Delinquency, Advances in Criminological Theory, Vol. 7 . New Brunswick, NJ: Transaction. LeBlanc, M.1989. Designing a Self-Report Instrucemtn for the Study of the Development of Offending from Childhood to Adulthood: Issues and Problems. In M.W. Klein (ed) Cross-National Research in Self-Reported Crime and Delinquency . Los Angeles: Kluwer Academic Publishers. LeBlanc, M. and Frechette, M. 1989. Male Criminal Activity from Childhood through Youth: Multilevel and Developmental Perspectives . New York: Springer. LeBlanc, M. and Loeber, R. 1998. Devel opmental Criminology Updated. In M. Tonry (ed) Crime and Justice: An Annual Review of research, Vol. 23 . Chicago: University of Chicago Press.

PAGE 260

247 Lemert, E.M. 1951. Social Pathology: A Systematic Approach to the Theory of Sociopathic Behavior . Lerman, R. 1993. A National Pr ofile of Young, Unwed Fathers. In R.I. Lerman and T.J. Ooms (eds) Young Unwed Fathers: Changing Roles and Emerging Policies . Philadelphia, PA: Temple University Press. Lerman, R. 1996. The Impact of the Cha nging US Family Structure on Child Poverty and Income Inequality. Economica , 63: 119-139 Lerman, R., and Sorenson, E. 2000. Father Invo lvement with their Nonmarital Children: Patterns, Determinants, and Effects on thei r Earnings. In Peters, H.E., Peterson, G.W., Steinmetz, S.K., and Day, R.D. (eds) Fatherhood: Research, Intervention, and Policies . Binghampton, NY: Haworth Press. Levine, J.A., and Pitt, E.W. 1995. New Expectations: Community Strategies for Responsible Fatherhood . New York: Families and Work Institute. Lichter, D.T. 1988. Racial Differences in Underemployment in American Cities. American Journal of Sociology , 93 (4): 771-792. Lipton, D., Martinson, R., and Wilks, J. 1975. The Effectiveness of Correctional Treatment . New York: Praeger. Little, R.J.A. 1995. Modeling the Drop-out Me chanism in Repeated-Measure Studies. Journal of the American Statistical Association , 90: 1112-1121. Loeber, R. 1982. The stability of anti-social child behavior: A review . Child Development , 53: 1431-1446. Loeber, R. and Farrington, D.P. 1998. Seri ous and Violent Juvenile Offenders: Risk Factors and Successful Interventi ons. Thousand Oaks, CA: Sage. Loeber, R. and LeBlanc, M. 1990. Toward a Developmental Criminology. In M. Tonry and N. Morris (eds) Crime and Justice: An Annual Review of research, Vol. 12. Chicago: University of Chicago Press. Loeber R., and Stouthamer-Loeber, M. 1986. Family Factors as Correlates and Predictors of Juvenile Conduct Problems and Delinquency. In M. Tonry and N. Morris (eds) Crime and Justice: An Annual Review of Research, vol. 7 . Chicago: University of Chicago Press. Loeber, R., and Stouthamer-Loeber, M. 1987. Prediction. In H.C. Quay (ed) Handbook of Juvenile Delinquency . NewYork: Wiley. Loeber, R., Stouthamer-Loeber, M., Van Ka mmen, W.B., and Farrington, D.P. 1991. Initiation, Escalation, and Desistance in J uvenille Offending and their Correlates. Journal of Criminal Law and Criminology , 82: 36-82.

PAGE 261

248 Lott, J.R. 1990. The Effect of Convicti on on the Legitimate Income of Criminals. Economics Letters , 34 (4): 381-385. Loury, G.C. 1987. The Family as Contex t for Delinquent Prevention: Demographic Trends and Political Realities. In J.Q. Wilson and G.C. Loury (eds) From Children to Citizens: Families, Schools, and Delinquency, vol. 3 . New York: Springer-Verlag. Luhman, N. 1985. Ecological Communication . Chicago: University of Chicago Press. Manski, C.F. 1995. Identification Problems in the Social Sciences . New York: Cambridge University Press. Marsiglio, W. 1998. Procreative Man . New York: New York University Press. Marsiglio, W. 2002. Studying Fathering Traj ectories: In-depth Interviewing and Sensitizing Concepts. In R. Day and M.E. Lamb, Measuring Father Involvement . Mahwah, N.J.: Lawrence Erlbaum Associates. Martinson, R. 1974. What works? Questi ons and answers about prison reform. Public Interest , 35: 22-45. Maruna, S. 2001. Making Good: How Ex-Convicts Reform and Rebuild Their Lives . Washington D.C.: American Psychological Association. Massey, D., and Denton, S. 1993. American Apartheid: The Making of the Underclass . Cambridge, MA: Harvard University Press. Mauer, M. 1994. Americans Behind Bars: The Inter national Use of Incarceration . Washington, D.C.: The Sentencing Project. Mauer, M. 1999. Race to Incarcerate . NewYork: The Free Press. McCord, J. 1979. Some Child-rearing Anteceden ts of Criminal Behavior in Adult Men. Journal of Personality and Social Psychology , 9: 1,477-1,486. McLanahan, S. 1998. Growing up wit hout a father. In C. Daniels (ed.) Lost Fathers: The Politics of Fatherlessness in America . New York: St. Martin’s Press. McLanahan, S., Garfinkle, I ., Reichman, N.E., Teitler, J ., Carlson, M., and Audinger, C.N. 2001. The Fragile Families and Ch ild Wellbeing Study: National Baseline Report. Princeton, NJ: Bendheim-Thoma n Center for Research on Child Wellbeing. McLanahan, S., and Sandefur, G. 1994. Growing up with a Single Parent . Cambridge, MA: Harvard University Press. Mead, G.H. 1934. Mind, Self, and Society . Chicago: University of Chicago Press.

PAGE 262

249 Menard, S. 2002. Longitudinal Research. Thousand Oaks, CA: Sage. Merton, R. 1968. Social Structure and Social Action . New York: Free Press. Messerschmidt, J.W. 1993. Masculinities a nd Crime: Reconceptual ization of Theory. Lanham, MD: Rowman and Littlefield Publishers. Messner, S.F., Raffalovich, L.E., and McMi llan, R. 2001. Economic Depravation and Changes in Homicide Arrests for White and Black Youths. Criminology , 39: 591613. Moffitt, T. 1993. Adolescence-limited and li fe-course-persistent anti-social behavior: A developmental taxonomy. Psychological Review , 100: 647-701. Mosher C. J., Miethe, T.D., and Phillips, D.M. 2002. The Mismeasure of Crime . Thousand Oaks, CA: Sage. Moss, P. and Tilly, C. 1991. Why Black Men are Doing Worse in the Labor Market: A Review of Supply Side a nd Demand-side Explanations . New York: Social Science Research Council. Mott, F. 1990. When is Father Really Gone ? Paternal-child Contact in Father Asbent Homes. Demography , 27: 399-518. Mumola, C.J. 2000. Incarcerated Parents and th eir Children. Bureau of Justice Statistics. Nagin, D.S., Farrington, D.P., and Moffitt, T. 1995. Life-Course Trajectories of Different Types of Offenders. Criminology , 33: 111-139. Nagin, D.S. and Paternoster, D.A. 1991. On the Relationship of Past and Future Participation in Delinquency. Criminology , 29: 163-190. Nagin, D.S. and Paternoster, D.A. 1994. Pe rsonal Capital and Social Control: The Detterence Implications of a Theory of Individual Differences in Criminal Offending. Criminology , 32 (4): 581-606. Nagin, D.S. and Paternoster, D.A. 2000. P opulation Heterogenity and State Dependence of the Evidence and Directi ons for Future Research. Journal of Quantitative Criminology , 6: 335-356. Nagin, D., and Waldfogel, J. 1993. The E ffects of Criminality and Conviction on the Labour Market Status of Young British Offenders. Unpublished Manuscript. Carnegie Melon University. Neufeldt, V. and Sparks, A. 1990. Webster’s New World Dictionary . New York: Warner Books.

PAGE 263

250 Newton, C. 1994. Gender Theory and Pr ison Sociology: Using Theories of Masculinities to Interpret the Sociology of Prisons for Men. The Howard Journal of Criminal Justice , 33 (3): 193-202. Nielsen, A.L. 1999. Testing Sampson a nd Laub’s Life Course Theory: Age, Race/Ethnicity, and Drunkeness. Deviant Behavior: An In terdisciplinary Journal , 20: 129-151. Nock, S. 1998. Marriage in Men’s Lives . New York: Oxford University Press. Nurse, A.M. 2000. Coming Home: The Transitio n From Incarcerated to Paroled Young Father. Families, Crime, and Criminal Justice , 2, 281-308. Nurse, A.M. 2001. Fatherhood Arrested: Parenting from within the Juvenile Justice System . Nashville: Vanderbilt University Press. Nye, F.I. 1958. Family Relationships and Delinquent Behavior . New York: Wiley. Oliver, M., and Shapiro, D. 1997. Black/White Wealth: A New Perspective on Racial Inequality . Routledge Inc. Osborn, S.G. and West, D.J. 1979. Marr iage and Delinquency: A Postscript. British Journal of Criminology , 19: 254-256. Osgood, D.W., O’Malley, P.M., Bachman, J., a nd Johnston, L.D. 1989. Time Trends and Age Trends in Arrests in Se lf-Reported Illegal Behavior. Criminology , 32: 517-554. Patterson, G.R. 1993. Orderly Change in a Stable World: The Antisocial Trait as a Chimera. Journal of Consulting and Clinical Psychology , 61: 911-919. Pasternoster, R., Dean, C., Piquero, A., Mazero lle, P., and R. Brame. 1997. Generality, continuity, and change in offending. Journal of Quantitative Criminology , 13, 3: 231-266. Petersilia, J. 1995. A crime control rationale for reinvesting in community corrections. The Prison Journal , 75, 4: 479-496. Petite, B., and B. Western. 2004. Mass Impr isonment and the Life-course: Race and Class Inequality in U.S. Incarceration. American Sociological Review , 69: 151169. Palm, G. 1996. Understanding the Parent E ducation Needs of Incarcerated Fathers. Paper presented at the Annual Conferen ce of the National Council of Family Relations (Kansas City, MO). Pezzin, L. 1995. Earning Prospects, Matching Effects, and the Decision to Terminate a Criminal Career. Journal of Quantitative Criminology , 11: 29-50

PAGE 264

251 Piquero, A.R., Farrington, D.P., and Blumstein, A. 2003. The Criminal Career Paradigm. In M. Tonry (ed.), Crime and Justice, Vol. 30 . Chicago: University of Chicago Press. Piquero, A.R., MacDonald, J., and Parker, K.F. 2002. Race, Local Life Circumstances, and Criminal Activity over the Life-Course. Social Science Quartlerly , 83: 654670. Piquero, A.R., MacIntosh, R., and Hickman, M. 2002. The Validity of a Delinquency Scale. Sociological Methods and Research , 30: 492-529. Piquero, A.R. and Mazzerole, P. (eds.) 2001. Life-Course Criminology: Contemporary and Classic Readings . Belmont, CA: Wadsworth. Pleck, J.H., Sonestein, F.L., and Ku, L. 1995. Correlates of Pregnancy Experience in Young Males’ Subsequent Fertility Behavior . Urbanna-Champain: University of Illinois. Porterfield, A.L. 1943. Delinquency a nd Outcome in Court and College. American Journal of Sociology , 49: 199-208. Porterfield, A.L. 1946. Youth in Trouble . Fort Worth: Leo Potishman Foundation. Popenoe, D. 1993. American family declin e, 1960-1990: A review and appraisal. Journal of Marriage and the Family , 55, 3: 527-542. Popenoe, D. 1996. Life without Father: Compelling new Evidence that Fatherhood and Marriage are Indispensable fo r the Good of Children and Soc iety. New York: Free Press Rand, A. 1987. Transitional Life Events a nd Desistance from Delinquency and Crime. In From Boy to Man, from Delinquency to Crime , M. Wolfgang, T.P. Thornberry, and R. Figlio (Eds). Chicago: University of Chicago Press. Rowe, A.R., and Tittle, C.H. 1977. Life Cycle Changes and Criminal Propensity. Sociological Quarterly , 18: 223-236. Reinman, J. 1979. The Rich get Richer and the Poor get Prison: Ideology, Class, and Criminal Justice . New York: Wiley and Sons. Reiss, A.J. 1951. Delinquency as the failu re of personal and social controls. American Sociological Review , 16: 196-207. Resnick, M.D. Chambliss, S.A., and Blum, R. W. 1993. Health and Risk Behaviors of Urban Adolescent Males i nvolved in Pregnancy. Family Sociology , 74: 366-374. Richards, S. 1995. The Structure of Prison Release . New York: McGraw-Hill.

PAGE 265

252 Richards, S. 1998. Critical and radical pe rspectives on community punishment: Lessons from the darkness. In J. Ross (ed) Cutting the Edge: Current Perspectives in Radical/Critical Criminology and Criminal Justice . Westport, CT: Praeger. Richards, S. and Jones, R. 1997. Perpet ual incarceration m achine: Structural impediments to postprison success. Journal of Contemporary Criminal Justice , 13, 1: 4-22. Riveland, C. 1999. Prison managemnent trends : 1975-2025. In M. Tonry and J Petersilia (ed) Prisons (Crime and Justice: A Review of the Research , vol 26. Chicago: University of Chicago Press. Ross J., and Richards, S. 2002. Introductio n: What is the new school of convict criminology? In J. Ross and S. Richards (eds) Convict Criminology . Belmont, CA: Thomas-Wadsworth Rothman, D. 1972. The Diversity of the Asylum . Boston: Little. Roy, K. 2003. “Nobody can be a Father in Here ”: Identitity Construc tion and Institutional Constraints on Incarcerated Fatherhood. Paper Presented at National Council on Family Relations Annual Conference, Vancouver, British Columbia, Canada. Rutter, G.R. and Giller, H. 1983. Juvenile Delinquency: Trends and Perspectives . New York: Guilford Press. Rutter, M., Quinton, D., and Hill, J. 1990. A dult outcomes of institution reared children: Males and females compared. In L. Robins and M. Rutter (eds.) Straight and Devious Pathways from Childhood to Adulthood . Cambridge: Cambridge University Press. Rutter, M., and Rutter, M. 1993. Developing Minds: Challenges and Continuity across the Life-span . New York: Basic Books. Sabo, D., Kupers, T., and London, W. 2001. Ge nder and the politics of punishment. In D. Sabo, T. Kupers, and W. London (eds) Prison Masculinities . Philidelphia: Temple University Press. Sampson, R.J., and Laub, J.H. 1990. Crime and deviance over the life course: The salience of adult social bonds. American Sociological Review , 55: 609-627. Sampson, R.J., and Laub, J.H. 1992. Crime and deviance in the life-course. Annual Review of Sociology , 18: 63-84. Sampson, R.J., and Laub, J.H. 1993. Crime in the Making: Pathways and Turning Points through Life . Cambridge, MA: Harvard University Press.

PAGE 266

253 Sampson, R.J., and Laub, J.H. 1995. Understa nding variability in lives through time: Contributions of life-course criminology. Studies on crime and Crime Prevention , 4, 2: 143-158. Sampson, R.J., and Laub, J.H.. 1997. A Life-Course Theory of Cumulative Disadvantage and the Stability of De linquency. In T.P. Thornberry (ed) Developmental Theories of Crime and Delinquency, Vol. 7 Advances in Criminological Theory . New Brunswick, NJ: Transaction. Sampson R.J., and Lauritsen, J.L. 1997. R acial and Ethnic Disparities in Crime and Criminal Justice in the United States. Crime and Justice , 21: 311-374. Savelsberg, J. 1994. Knowledge, domin ation, and criminal punishment. American Journal of Sociology, 99: 911-943. Schlosser, E. 1998. The prison-industrial complex. The Atlantic Monthly , 282, 6: 51-77. Sellin, T. 1931. The basis of a crime index. Journal of the American Institute of Criminal Law and Criminology , 22: 335-356. Seltzer, J.A. 1991. Relationships between Fathers and Children who Live Apart: The Father’s Role after Separation. Journal of Marriage and the Family , 53 (1): 179101. Shihadeh, E.S., and Ousey, G.C. 1998. Indus trial Restructuring and Violence: The Link Between Entry-Level Jobs, Economic Deprivation, and Black and White Homicide. Social Forces , 77 (1): 185-206. Short, J.F., Jr., and Nye, F.I. 1957. Repor ted Behavior as a Cr iterion of Delinquent Behavior. Social Problems , 5: 207-213. Short, J.F., Jr., and Nye, F.I. 1958. Exte nt of Unrecorded Juvenile Delinquency: Tentative Conclusions. Journal of Criminal Law and Criminology , 49: 296-302. Shover, N. 1985. Aging Criminals . Beverly Hills: Sage. Shover, N. 1996. Great Pretenders: Pursuits and careers of Persistent Thieves . Boulder, CO: Westview Press. Simon, J. and Feeley, M. 1995. True crim e: The new penology and public discourse on crime. In T. Blomberg and S. Cohen (eds) Punishment and Social Control: essays in honor of Sheldon Messinger . New York: Aldien de Gruyter. Sonenstein, F.L., Pleck, J.H. and Ku, L. C. 1993. Paternity Risks among Adolescent Males. In R.I. Lerman and T.J. Ooms (eds) Young Unwed Fathers: Changing Roles and Emerging Policies . Philadelphia, PA: Temple University Press.

PAGE 267

254 Springarn, R.W., and DuRant, R. 1996. Male Adolescents Involved in Pregnancy: Associated Health Risks and Problem Behaviors. Pediatrics , 98: 262-268. Stacey, J. 1998. Dada-ism in the 1990’s: Get ting past baby talk about fatherlessness. In C. Daniels (ed.) Lost Fathers: The Politics of Fatherlessness in America . New York: St. Martin’s Press. Staples, R., and Johnson, L.B. 1993. Black Families at the Crossroads . San Francisco: Jossey-Bass. Statistical Abstract of the U.S., 1995. Washington, D.C.: U.S. Bureau of the Census. Stitt, B.G., and Giacopassi, D.J. 1992. Trends in the connectivity of theory and research in criminology. The Criminologist , 17, 1: 3-6. Stone, L. 1989. The road to polygamy. New York Review of Books , March: 14. Stouthamer-Loeber, M. and Wei, E.H. 1998. The Precursors of Young Fatherhood and its Effect on Deliquency of Teenage Males. Journal of Adolescent Health , 22 (1): 56-65. Sutherland, E.H. 1949. White-Collar Crime . New York: Holt, Rinehart, and Winston. Sutherland, E.H. 1947. Principles of Criminology 4th edition. Philadelphia: Lippencourt. Stattin, H., Magnusson, D., and Reichel, H. 1989. Criminal activity at Different Ages: A Study Based on a Swedish Longitudinal Research Population. British Journal of Criminology , 29: 368-385. Sykes, G., and Matza, D. 1957. Techniques of neutralization: A theory of delinquency. American Journal of Sociology , 22: 664-670. Tasker, F. and Golombok, S. 1997. Growing up in a Lesbian Family: Effects on Child Development . New York: Guilford Press Tabachnick, B.G., and Fidell, L.S. 2001. Using Multivariate Statistics 4th edition. Boston: Allyn and Bacon. Terry, C. 2000. Beyond punishment: Perp etuating difference from the prison experience. Humanity and Society , 24, 2: 108-135. Thornberry, T.P. 1987. Toward an In teractional Theory of Delinquency. Criminology , 25: 863-891. Thornberry, T.P., and Krohn, M.D. 2001. The Development of Delinquency: An International Perspective. In S.O. White (ed) Handbook of Youth and Justice . New York: Plenum.

PAGE 268

255 Thornberry, T.P. 1989. Panel Effects and the Use of Self-reported Measures of Delinquency in Longitudinal Stud ies. In M.W. Klein (ed) Cross-National Research in Self-Reported Crime and Delinquency . Dordrecht: Kluwer Academic. Tittle, C.R. 1988. Two Empirical Regularities in Search of an Explanation: Commentary on the Age/Crime Debate. Criminology , 26: 75-85. Toch, H. 1992. Mosaic of Despair: Human Breakdowns in Prison . Washington, D.C.: American Psychological Association. Tonry, M. 1995. Malign Neglect: Race, Crime, and Punishment in America . New York: Oxford University Press. Trassler, G. 1979. Delinquency, Recidivism, and desistance. British Journal of Criminology , 19: 314-322. Travis, J., Solomon, A., and Waul, M. 2001 . From Prison to Home: The Dimensions and Consequences of Prisoner Reentry . Washington D.C.: The Urban Institute. Tripp, B. 2001. Incarcerated African American Fathers: Exploring Changes in Family Relationships and Father Identity. The Journal of African American Men , 6, 1, 13-29. Thornberry, T. 1997. Introducti on: Some advantages of de velopmental and life-course perspectives for the study of crime a nd delinquency. In T. Thornberry (ed.) Developmental Theories of Crime and Deli nquency: Advances in Criminological Theory . New Brunswick, NJ: Transaction Publishers. Thornberry, T.,and Krohn, M. 2000. “The Self-Report Method of Measuring Delinquency and Crime.” In Measurement and Analysis of Crime and Justice , vol. 4 of Criminal Justice 2000 . Washington D.C.: Nationa l Institute of Justice, Office of Justice Programs, U.S. Department of Justice. Tracy, P., Wolfgang, M., and Figlio, R. 1990. Delinquency Careers in Two Birth Cohorts . New York: Plenum. Uggen, C. 2000. “Work as a Turning Point in the Life Course of Criminals: A Duration Model of Age, Employment, and Recidivism.” American Sociological Review 67:529. US Census Bureau. 2002. Current Populat ions Survey Report P20-547. Table C3, Living Arrangements of Children under 18 Years of Age by Marital Status and Selected Characteristics of Parents: March 2002 (pp. 36-55). Venkatesh, S.A., and Levitt, S.D. 2000. “Are we a Family or a Business?” History and Disjuncture in the Urban American Street Gang. Theory and Society , 29: 427462.

PAGE 269

256 Visher, C.A. 1986. The Rand Inmate Survey: A Re-Analysis. 1986. In A. Blumstein Blumstein A., Cohen, J., Roth, J., and J. Visher (eds) Criminal Careers and Career Criminals . Washington D.C.: National Academy of Sciences. Visher, C.A., and Roth, J.A. 1986. Participati on in Criminal Careers. In Blumstein A., Cohen, J., Roth, J.A., and J. Visher (eds) Criminal Careers and Career Criminals . Washington D.C.: National Academy of Sciences. Waite, L.J. 1995. Does Marriage Matter? Demography , 32 (4): 483-507. Waldfogel, J. 1994. Does Conviction have a Persistent Effect on Income and Employment?. International Review of Law and Economics , 14: 103-119. Waldorf, D., Reinarman, C., and Murphy, S. 1991. Cocaine Changes: The Experiences of Using and Quitting . Philadelphia: Temple University Press. Warr, M. 1998. Life-Course Transiti ons and Desistance from Crime. Criminology 36, 2: 183-216. Wente, A.S., and Crockenberg, S.B. 1976. Transition to Fatherhood: Lamaze Preparation, Adjustment Difficulty, a nd the Husband-Wife Relationship. The Family Coordinator : 351-357. West, D.J. 1982. Delinquency: It’s Roots, Careers, and Prospects . Cambridge, MA: Harvard University Press. Western, B., and Beckett, K. 1999. How Un regulated is the U.S. Labor Market? The Penal System as a Labor Market Institution. The American Journal of Sociology , 104, 1: 1030-1060. Western, B., and Petit, B. 2000. Incarcer ation and Racial In equality in Men’s Employment . Industrial and Labor Relations Review , 54 (1): 3-16. Western, B., and McLanahan, S. 2000. Fathers Behind Bars: The Impact of Incarceration on Family Formation. Families, Crime, and Criminal Justice , 2, 309-324. Wheeler, S. Socialization in Correctional Communities. American Sociological Review, 26: 699-712 White, H.R., and Bates, M.E. 1995. Cessation from Cocaine Use. Addiction , 90: 947957. Williams, J.R. and Gold, M. 1972. From Deli nquent Behavior to Official Delinquency. Social Problems , 20: 209-229.

PAGE 270

257 Wilson, F.D., Tienda, M., and Wu, L. 1995. Race and Unemployment: Labor Market Experiences of Black and White Men, 1968-1988. Work and Occupations , 22 (3): 245-270. Wilson, J.Q. 1983. Raising Kids. The Atlantic Monthly . October: 45-56. Wilson, W.J. 1987. The Truly Disadvantaged . Chicago: University of Chicago Press. Witte, A.D., and Reid, P. 1980. An Explorat ion of Labor Market Performance fpr Prison Releases. Journal of Urban Economics , 8: 313-339. Wolfgang, M., Figlio, R., and Sellin, T. 1972. Delinquency in a Birth Cohort . Chicago, IL: University of Chicago Press. Woltman, H., and Bushery, J. 1984. Summar y of Results from the National Crime Survey panel bias. In R. Lehnen and W. Skogan (eds) The National Crime Survey Working Papers, Vol. 2: Methodological Studies . Washington, D.C.: U.S. Government Printing Office. Yamaguchi, K. and Kandel, D.B. 1985. On the Resolution of Role Incompatibility: A Life Event History Analysis of fa mily Roles and Marijuana Use. American Journal of Sociology , 90: 1284-1325. Zimring, F., and Hawkins, G. 1991. The Scale of Punishment . Chicago: University of Chicago Press.

PAGE 271

258 BIOGRAPHICAL SKETCH Bradley G. Tripp was born in Parma, Ohio. He spent most of his life in Altamonte Springs, Florida. He attended both Wake Forest University a nd the University of Florida as an undergraduate. He attended the Universi ty of Florida for his Master’s and Doctoral work. He now lives in Fort Mill, SC with hi s wife, Kristen, and three cats; Brodie, Suzie, and Munchie. He is a professor at Winthrop University.