Citation
Trajectories of Two Racine Birth Cohorts

Material Information

Title:
Trajectories of Two Racine Birth Cohorts A Theoretically Integrated Model for Explaining Offending
Creator:
Jennings, Wesley G
Place of Publication:
[Gainesville, Fla.]
Florida
Publisher:
University of Florida
Publication Date:
Language:
english
Physical Description:
1 online resource (181 p.)

Thesis/Dissertation Information

Degree:
Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Criminology, Law, and Society
Criminology, Law and Society
Committee Chair:
Lanza-Kaduce, Lonn M.
Committee Co-Chair:
Piquero, Alexis R.
Committee Members:
Parker, Karen F.
Stults, Brian J.
Peek, Charles W.
Graduation Date:
12/14/2007

Subjects

Subjects / Keywords:
Crime ( jstor )
Criminals ( jstor )
Criminology ( jstor )
Delinquency ( jstor )
Modeling ( jstor )
Neighborhoods ( jstor )
Police ( jstor )
Recidivism ( jstor )
Social control ( jstor )
Trajectories ( jstor )
Criminology, Law and Society -- Dissertations, Academic -- UF
multi, offending, trajectories
Genre:
bibliography ( marcgt )
theses ( marcgt )
government publication (state, provincial, terriorial, dependent) ( marcgt )
born-digital ( sobekcm )
Electronic Thesis or Dissertation
Criminology, Law, and Society thesis, Ph.D.

Notes

Abstract:
Macro-level research evidencing the relationship between communities and crime dates back to the early ecological work of Park and Burgess and later further developed by Shaw and McKay into what is commonly referred to as social disorganization theory. This research has since shown that poverty, ethnic heterogeneity, and residential mobility lead to the inability for members of the community to develop mechanisms of informal social control, which in turn leads to higher rates of crime. In contrast, micro-level theories such as Sampson and Laub?s theory argues that individual pathways of offending can be altered by life transitions such as marriage and employment. Comparatively, Moffitt?s taxonomy asserts that there are two theoretically distinct groups of offenders, where one group displays an adolescent-limited type of offending and the other group are considered to be life-course-persistent or chronic offenders. Moffitt has also argued that the life-course-persistent type is predominantly a male phenomenon. This study will provide one of the first attempts to explore the possibility of developing a theoretically integrated model for explaining group offending using trajectory analysis and relying on data from two Racine Birth Cohorts. Results from this study relying on official data yielded varying levels of support for all of the theoretical frameworks mentioned above. Specifically, all of the trajectory models discovered groups consistent with Moffitt?s typology, as well as uncovering groups not necessarily consistent with her typology. Also consistent with Moffitt, sex was shown to be a significant covariate for predicting group assignment. Bivariate support was found for Sampson and Laub's age-graded theory of informal social control. Lastly, multivariate support was revealed for the impact of living in a socially disorganized area. Yet, once mlogit models were estimated incorporating the individual and neighborhood effects there was evidence of mediation. It appeared that individual and family variables at times completely or at least partially mediated the effect of social disorganization as a distinguishing feature of group membership. Implications for the current findings and directions for future research are also discussed. ( en )
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Bibliography:
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Thesis:
Thesis (Ph.D.)--University of Florida, 2007.
Local:
Adviser: Lanza-Kaduce, Lonn M.
Local:
Co-adviser: Piquero, Alexis R.
Statement of Responsibility:
by Wesley G Jennings.

Record Information

Source Institution:
UFRGP
Rights Management:
Copyright Jennings, Wesley G. 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.
Classification:
LD1780 2007 ( lcc )

Downloads

This item has the following downloads:


Full Text





Klebanov, P.K., Brooks-Gunn, J., Chase-Landsdale, P., & Gordon, R.A. (1997). Are
neighborhood effects on young children mediated by features of the home environment?
In J. Brooks-Gunn, G.J. Duncan, & J. Aber (ed.), Neighborhood poverty: Context and
consequences for children. Sage: New York, NY.

Klein, M.W. (1971). Street gangs and street workers. Englewood Cliffs, NJ: Prentice Hall.

Kling, J.R., Liebman, J.B., & Katz, L.F. (2007). Experimental analysis of neighborhood effects.
Econometrica, 75, 83-119.

Kornhauser, R.R. (1978). Social sources of delinquency. Chicago: University of Chicago Press.

Kratzer, L., & Hodgins, S. (1999). A typology of offenders: A test of Moffitt's theory among
males and females from childhood to age 30. Criminal Behavior andMental Health, 9,
57-73.

Krivo, L.J., & Peterson, R.D. (1996). Extremely disadvantaged neighborhoods and urban crime.
Social Forces, 75, 619-650.

Kroneman, L., Loeber, R., & Hipwell, A.E. (2004). Is neighborhood context differently related
to externalizing problems and delinquency for girls compared with boys? Clinical Child
and Family Psychological Review, 7, 109-122.

Kupersmidt, J.B., Griesler, P.C., DeRozier, M.E., Patterson, C.J. & Davis, P.W. (1995).
Childhood aggression and peer relations in the context of family and neighborhood
factors. Child Development, 66, 360-375.

Lambert, D. (1992). Zero-inflated Poisson regression, with an application to defects in
manufacturing. Technometrics, 34, 1-14.

Land, K. (2000). Influence of neighborhood, peer, and family context: Trajectories of
delinquent/criminal offending across the life course, Final Report. Washington, DC:
National Institute of Justice.

Land, K.C., McCall, P.L., & Cohen, L.E. (1990). Structural covariates of homicide rates: Are
there any invariances across time and social space? American Journal of Sociology, 95,
922-963.

Land, K.C., McCall, P.L., & Nagin, D.S. (1996). A comparison of Poisson, Negative Binomial,
and Semiparametric Mixed Poisson Regression Models with empirical applications to
criminal careers research. Sociological Methods and Research, 24, 387-442.

Land, K.C., & Nagin, D.S. (1996). Micro-level models of criminal careers: A synthesis of The
criminal careers and life course approaches via Semiparametric Mixed Poisson
Regression Models, with empirical models. Journal of Quantitative Criminology, 12,
163-191.









Sampson, R.J., & Laub, J.H. (2005b). When prediction fails: From crime-prone boys to
heterogeneity in adulthood. The ANNALS of the American Academy of Political and
Social Science, 602, 73-79.

Sampson, R.J., & Laub, J.H. (2005c). A life-course view of the development of crime. The
ANNALS of the American Academy of Political and Social Science, 602, 12-45.

Sarason, I.G., & Sarason, B.R. (1981). Teaching cognitive and social skills to high school
students. Journal of Clinical and Consulting Psychology, 49, 908-918.

Schreck, C.J., Fisher, B., & Miller, J.M. (2004). The social context of delinquent peer networks
and violent victimization. Justice Quarterly, 21, 23-48.

Schuerman, L., & Kobrin, S. (1986). Community careers in crime. In A.J. Reiss & M. Tonry
(eds.), Crime and justice: A review of research. Chicago: University of Chicago Press.

Schweinhart, L.J., & Weikart, D.P. (1988). The High/Scope Perry preschool program. In R.H.
Price, E.L. Cohen, R.P. Lorion, & J. Ramos-McKay (eds.), 14 ounces ofprevention: A
casebookfor practicioners (pp. 53-65). Washington, DC: American Psychological
Association.

Seitz, V., Rosenbaum, L.K., & Apfel, N.H. (1985). Effects of family support intervention: A ten-
year follow up. ChildDevelopment, 56, 376-391.

Shannon, L.W. (1988). Criminal career continuity: Its social context. New York: Human
Sciences Press.

Shannon, L.W. (1991). Changing patterns of delinquency and crime: A longitudinal study in
Racine. Boulder: Westview Press.

Shannon, L.W. (1994). Juvenile delinquency and adult crime, 1948-1977 [Racine, Wisconsin]:
City ecological data [Computer file]. 2nd ICPSR version. Iowa City, IA: University of
Iowa, Iowa Urban Community Research Center [producer], 1980. Ann Arbor, MI:
Interuniversity Consortium for Political and Social Research [distributor], 1994.

Shavit, Y., & Rattner, A. (1988). Age, crime, and the early life course. The American Journal of
Sociology, 93, 1457-1470.

Shaw, C.D., & McKay, H.D. (1942). Juvenile delinquency and urban areas. Chicago: University
of Chicago Press.

Short, J.F. (1990). Gangs, neighborhoods, and youth crime. Criminal Justice Research Bulletin,
5, 1-11.

Silverthorn, P., & Frick, P.J. (1999). Developmental pathways to antisocial behavior: The
delayed-onset pathway in girls. Development andPsychopathology, 11, 101-126.











Table 5-25. 1949 Racine Continuous Interviewed Residents i/ i/i Block Data: Odds Ratios and
Probabilities for Macro-Level Predictors of Offending Group Membership
Adolescent-Limited/Peaked High-Rate


Variables Offenders Chronic Offenders
Odds Prob. Odds Prob.
Macro-level
Social Disorganization 1.50 <.01** 1.48 .07

Model Diagnostics
Chi-Square= 7.96*
-Log Likelihood= -207.61
McFadden's R2 .02
Nagelkerke R2 .04
p<.10 *p<.05 **p<.01 ***p<.001
Note. Reference category is group 1 (the nun-otiender.N).


Table 5-26. 1949 Racine Continuous Interviewed Residents i/ i/i Block Data: Odds Ratios and
Probabilities for Micro-Level Predictors of Offending Group Membership
Adolescent-Limited/Peaked High-Rate
Variables Offenders Chronic Offenders
Odds Prob. Odds Prob.
Micro-level
Male 2.31 <.001*** 2.40 .01*
Black 1.38 .06+ 1.20 .64
Low SES Household 1.26 .18 2.52 <.001***
Stakes in Conformity 1.09 .62 1.13 .50
Family Stability (Natural 0.64 <.001*** 0.67 .11
Parents)
Family Structure (No. of 1.36 .04* 1.61 .14
Children)
Differential association 1.25 .16 1.01 .99
Formal Social Control 1.14 .43 1.14 .64
Moved 0.89 .47 0.79 .51

Model Diagnostics
Chi-Square= 76.67***
-Log Likelihood= -169.86
McFadden's R2 .19
Nagelkerke R2= .31
p<.10 *p<.05 **p<.01 ***p<.001
Note. Reference category is group 1 (the nun-otiender.N).









Final Thoughts

Guerra (1997:300-301) perhaps best summarizes the direction for programmatic and policy

initiatives in the following statement presented below if we as a community have any hopes for

reducing the risk factors associated with both types of offending patterns identified by Moffitt,

but particularly for prevention/intervention with chronic offenders:

It is likely that the most effective interventions will address multiple causal factors, be

carried out in multiple contexts, and be extended in time over the course of development.

Families, peers, schools, and communities must work in concert to provide opportunities for

children to learn prosocial behavior and to create environments where prosocial behavior is

modeled and reinforced.

It is important for researchers, policymakers, and even individuals to be aware that all of

the abovementioned elements do not operate in isolation from one another. There are individual,

family, and community factors that all have influence on life-course pathways of development.

Sampson et al. (2002:474) also challenges neighborhood level researchers to move beyond mere

structural compositional measures of neighborhood effects and give greater attention toward

understanding "cultural, normative, and collective action perspectives that attach meaning to how

residents frame their commitment to places." Spatial dynamics are also important to

investigating the influence of neighborhoods on offending patterns. It is possible that while

youth may reside in a disadvantaged area, they may spend more time in an adjacent neighborhood

where there are stronger informal social ties and collective supervision of teenage peer groups.

Thus, while compositional measures of disadvantage may be related to crime rates, without

measuring the influence of spatial proximity to more advantaged areas the effect of

neighborhoods might be overestimated.










Table 5-22. 1942 Racine Continuous Interviewed Residents i/ i/i Block Data: Odds Ratios and
Probabilities for Macro-Level Predictors of Offending Group Membership
Adolescent-Limited/Peaked High-Rate


Variables Offenders Chronic Offenders
Odds Prob. Odds Prob.
Macro-level
Social Disorganization 1.97 <.001*** 2.18 <.001***

Model Diagnostics
Chi-Square= 18.30***
-Log Likelihood= -176.95
McFadden's R2= .06
Nagelkerke= .11
p<.10 *p<.05 **p<.01 ***p<.001
Note. Reference category is group 1 (the nuin-oitider.').


Table 5-23. 1942 Racine Continuous Interviewed Residents i/ i/i Block Data: Odds Ratios and
Probabilities for Micro-Level Predictors of Offending Group Membership
Adolescent-Limited/Peaked High-Rate
Variables Offenders Chronic Offenders
Odds Prob. Odds Prob.
Micro-level
Male 2.99 <.001*** 6.12 <.001***
Black 1.01 .97 1.27 .30
Low SES Household 1.38 .13 1.10 .72
Stakes in Conformity 0.90 .57 0.96 .89
Family Stability (Natural 0.62 <.01** 0.59 .05*
Parents)
Family Structure (No. of 0.96 .80 1.37 .11
Children)
Differential association 1.01 .97 0.84 .57
Formal Social Control 1.13 .55 1.12 .71
Moved 1.11 .61 2.30 <.01**

Model Diagnostics
Chi-Square= 53.14***
-Log Likelihood= -143.38
McFadden's R2= .23
Nagelkerke= .39

p<.10 *p<.05 **p<.01 ***p<.001
Note. Reference category is group 1 (the nun-otiender.N).









CHAPTER 5
RESULTS

Introduction

This chapter begins with a description of the summary statistics and mean differences

across the cohorts for all of the independent and dependent measures. Following this discussion

is a presentation of the group-based trajectories for all of the samples of the 1942 and 1949

Racine continuous residents (full sample, sample with interview data, and sample with interview

and block data). Included in the explanation of the group-based trajectory results is also a

discussion of the effect of sex (being male) as a covariate for increasing the likelihood of being

in an offender group as opposed to a non-offender group. This chapter concludes with an

ANOVA presenting the mean covariate levels across groups for all variables used in the

multivariate analysis by cohort. This is followed by an illustration of the multivariate results and

a description of the impact of macro-level and micro-level factors for distinguishing between

group-based trajectories. The multivariate analysis concludes with a presentation of the

theoretically integrated model for predicting group membership.

Descriptive Statistics

Table 5-1 below presents the descriptive statistics that correspond to all of the macro-level

and micro-level variables by cohort described previously that will be used in the mlogit analysis

that follows and the zero-order correlations among the continuous measures can be found in

Appendix A.

Independent Variables

As can be seen in Table 5-1 there was a variety of measures that significantly differed

between the 1942 and 1949 Racine continuous residents that precluded the possibility of

combining the cohorts in further analyses. Regarding the measures that were combined to create









Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

TRAJECTORIES OF TWO RACINE BIRTH COHORTS: A THEORETICALLY
INTEGRATED MODEL FOR EXPLAINING OFFENDING

By

Wesley G. Jennings

December 2007

Chair: Lonn Lanza-Kaduce
Cochair: Alex Piquero
Major: Criminology, Law, and Society

Macro-level research evidencing the relationship between communities and crime dates

back to the early ecological work of Park and Burgess and later further developed by Shaw and

McKay into what is commonly referred to as social disorganization theory. This research has

since shown that poverty, ethnic heterogeneity, and residential mobility lead to the inability for

members of the community to develop mechanisms of informal social control, which in turn

leads to higher rates of crime.

In contrast, micro-level theories such as Sampson and Laub's theory argues that individual

pathways of offending can be altered by life transitions such as marriage and employment.

Comparatively, Moffitt's taxonomy asserts that there are two theoretically distinct groups of

offenders, where one group displays an adolescent-limited type of offending and the other group

are considered to be life-course-persistent or chronic offenders. Moffitt has also argued that the

life-course-persistent type is predominantly a male phenomenon.

This study will provide one of the first attempts to explore the possibility of developing a

theoretically integrated model for explaining group offending using trajectory analysis and

relying on data from two Racine Birth Cohorts. Results from this study relying on official data









neighbors may be due to their exposure to violence and fear of crime (Roncek, 1981; Sampson,

1995). Sampson and Groves (1989) have also shown that high rates of residential instability

were associated with juvenile delinquency and crime.

Gephart (1997) reviewed the developmental impact of community effects on adverse

individual outcomes. In early childhood, children who reside in areas of extreme disadvantage

face a slightly higher risk for having a low birth weight and suffering from child abuse.

Comparatively, additional analyses have shown that the presence of affluent neighbors was

associated with higher IQs and the presence of low-income neighbors was related to a greater

likelihood of behavioral problems in early childhood.

Furthermore, research has also found that the influence of a neighborhood's socioeconomic

status on individual outcomes in early childhood and middle childhood were different across

races. At age 3, the benefit of affluent neighbors (versus middle SES neighbors) on children's

IQ scores was more salient among White children than African-American children (Brooks-

Gunn et al., 1993). Similar findings were also found among children at ages 5 and 6 for IQ

scores, verbal ability, and reading achievement scores (Chase-Lansdale et al., 1997; Duncan et

al., 1994). In addition, all of these studies revealed that the presence of low SES neighbors was

positively associated with early childhood behavior problems.

Gephart (1997) also indicated that research investigating neighborhood effects during late

childhood and early adolescence found that community impoverishment, the burden of child-

care, and residential instability were significant predictors of various forms of delinquency such

as violent crime, drug trafficking, juvenile delinquency, and teen pregnancy (see Gephart, 1997).

Two additional studies have also shown that neighborhood SES had a stronger effect for male

adolescents than female adolescents (Entwisle et al., 1994; Halpem-Felsher et al., 1997).









Moffitt and others (see Moffitt et al., 1994) have argued that female delinquency tends to be

more erratic when compared with male delinquency. More recently, Piquero and Moffitt (2005)

have contended that males do tend to show more involvement in crime/delinquency regardless of

offender group, i.e., adolescent-limited or life-course-persistent. Consequently, it is expected

that a greater proportion of males will exhibit both types of offense trajectories and more

specifically, being male will be a significantly distinguishing feature of group-based patterns of

offending.

In contrast, Sampson and Laub's age-graded theory of informal social control argues that

social bonds or stakes in conformity in adulthood affect desistance from crime. It is expected

that individuals who exhibit a virtual non-offending trajectory will also display the greatest

amount of stakes in conformity since they are likely to be already bonded as youth and

adolescence since they show no involvement in delinquency. Similarly, this assumed early

attachment to institutions such as family and school is likely to continue into adulthood in that

they will also be more likely to be married and have full-time jobs which will further restrict

their involvement in crime. Second, while social disorganization is likely to affect group

assignment to an offender group, it is believed that the experience of life transitions such as

marriage and employment will increase the likelihood that individuals will express an

adolescent-limited/peaked offense trajectory compared with a chronic offending pattern since

these social bonds are theorized to have an effect on criminal career desistance. Or in other

words, while these groups may have relatively similar patterns of offending during adolescence,

these life transitions will be related to desistance among the adolescent-limited/peaked offender

group.









CHAPTER 2
THEORETICAL PERSPECTIVES

Introduction

The origins of macro-level research, specifically social disorganization, can be traced back

to the early work in the first half of the twentieth century by Park and Burgess (1925) and the

later expansions by Shaw and McKay (1942). These researchers and others that followed in their

footsteps have formed what is commonly referred to today as the Chicago School. These

researchers, most notably Shaw and McKay, were influenced by the deep-rooted sociological

tradition of studying the ecology of crime and focused on explaining why crime rates not

individual criminality was higher in certain areas of the community. This tradition of studying

and theorizing about crime in the aggregate persisted for some time until there became an

increased interest in studying individual-level factors that affect offending, which was strongly

influenced by the increased use of survey methodology in sociology and criminology. This early

individual-level research sparked some of the most influential criminological theories that exist

today including social learning theory (Burgess & Akers, 1966; Akers et al., 1979) and social

control/bonding theory (Hirschi, 1969).

By the late 1970s and early 1980s individual-level research had evolved in the sociological

and criminological literature and led to what is still considered one of the most robust

relationships in the social sciences, i.e., the age-crime relationship, or the notion that crime rises

to a peak in middle to late adolescence followed by a precipitous decline in early adulthood.

Hirschi and Gottfredson (1983) argued that there is no need for any theory to account for the

relationship between age and crime because it is invariant across culture, time, and space.

Although, there has been a relatively strong amount of debate over this early assertion (which

will be returned to later), this along with evidence from Wolfgang et al.'s (1972) research on









McDermott, S., & Nagin, D.S. (2001). Same or different? Comparing offender groups and
covariates over time. Sociological Methods and Research, 29, 282-318.

McDermott, S., & Nagin, D. (1998). Same or different over time? In S.O White (ed.), Handbook
ofLaw and Social Science: Youth and Justice. New York: Plenum Press.

Meredith, W., & Tisak, J. (1990). Latent curve analysis. Psychometrika, 54, 557-585.

Messner, S., Krohn, M., & Liska, A. (1989). Theoretical integration in the study of deviance and
crime: Problems and prospects. Albany, NY: State University of Albany Press.

Moffitt, T.E. (1990). Juvenile delinquency and attention-deficit disorder: Developmental
trajectories from age 3 to 15. ChildDevelopment, 61, 893-910.

Moffitt, T.E. (1993). Adolescent-limited and life-course-persistent antisocial behavior: A
developmental taxonomy. Psychological Review, 100, 674-701.

Moffitt, T.E. (1994). Natural histories of delinquency. In E.G.M. Weitekamp & Hans-Jurgen
Kerner (eds.), Cross-national longitudinal research on human development and criminal
behavior. Dordrecht, the Netherlands: Kluwer Academic.

Moffitt, T.E. (1997a). Neuropsychology, antisocial behavior, and neighborhood context. In J.
McCord (ed.), Violence and childhood in the inner city. Cambridge, UK: University of
Cambridge Press.

Moffitt, T.E. (1997b). Adolescence-limited and life-course-persistent offending: A
complementary pair of developmental theories. In T.P Thornberry (ed.), Developmental
ithew ie\ of crime and delinquency: Advances in criminological theory, Volume Seven.
New Brunswick, NJ: Transaction Publishers.

Moffitt, T.E. (2003). Life-course-persistent and adolescent-limited antisocial behavior: A ten-
year research review and a research agenda. In B.B. Lahey, T.E. Moffitt, & A. Caspi
(eds.), Causes of conduct disorder and delinquency. New York: Guilford Press.

Moffitt, T.E. (2005). Life-course-persistent versus adolescence-limited antisocial behavior. In D.
Cicchetti & D. Cohen (eds.), Developmental psychopathology (2nd ed.). New York:
Wiley.

Moffitt, T.E., & Henry, B. (1989). Neuropsychological assessment of executive functions in self-
reported delinquents. Development and Psychopathology, 1, 105-118.

Moffitt, T.E., & Henry, B. (1991). Neuropsychological studies of juvenile delinquency and
violence: A review. In J. Milner (ed.), The neuropsychology of aggression Norwell, MA:
Kluwer Academic.

Moffitt, T.E., Lynam, D.R., & Silva, P.A. (1994). Neuropsychological tests predicting persistent
male delinquency. Criminology, 32, 277-300.









household density measure (percent overcrowding), a second measure of family dynamics was

included to represent family structure. This was a measure of the number of children within each

of the cohort member's family (including the cohort member).8

Differential Association. Differential association theory (Sutherland, 1939, 1947) has

long stressed criminal and deviant behavior is learned through exposure to and acceptance of

definitions favorable to crime. Burgess and Akers (1966), and in later reformulations by Akers

(1977), expanded on this idea by specifying the process in which learning occurs. Favorable

definitions of crime and deviance are learned in a social setting where the imitation of criminal

and deviant behavior also occurs. Delinquency, like any behavior, generates either beneficial or

adverse consequences positive or negative reinforcement. If delinquent or criminal behavior is

reinforced either socially or non-socially, the behavior is likely repeated in the presence of cues

to which it has been paired. Or in other words, Akers (1998:50) argues that the probability that

individuals will engage in deviant behavior is increased when they differentially associate with

others who commit the behavior and express favorable definitions toward it. Thus, based on

research that has produced strong evidence for what is now termed the perceived "delinquent

peer effect" (Felson & Haynie, 2002; Matsueda & Anderson, 1998; Schreck, Fisher, & Miller,

2003; Warr, 1993; Warr & Stafford, 1991), a measure was included with 1 indicating that the







8 Overall, there was very little missing data at the individual-level (less than 1% for any of the variables used in the
analyses). The only two variables that had missing data were family stability and family structure. In order to
correct for the missing values, the method of group specific means were computed. In order to do this procedure the
sample was sorted into five categories: 1) White males; 2) White females; 3) Black males; 4) Black females; and 5)
Chicano males (there were not any Chicano females in the sample). For the few cases (n=4) where a missing value
was recorded, the sex/race specific group means were imputed.









concluded that "offending may increase or decrease over time, but the worst offenders at one age

still tend to be the worst at another age." This statement is essentially the basis of Gottfredson

and Hirschi's persistent-heterogeneity interpretation (or static approach) of the relationship

between past and future offending where they state that:

The best predictor of crime is prior criminal behavior... [and] research shows that
differences between people in the likelihood that they will commit criminal acts persist
over time... and these differences appear early and remain stable over much of the life
course (1990:107-108).

Gottfredson and Hirschi (1990) expand upon this viewpoint in their discussions regarding

their theory of low self-control. According to Gottfredson and Hirschi (1990: 97), the main

underlying cause of low self-control is ineffective child-rearing. They argue that it is the primary

caregiver's (i.e., the parents) role to teach the child self-control through monitoring their

behavior, recognizing deviant behavior when it occurs, and effectively punishing the observed

deviant behavior. These practices are believed to foster the child's development of self-control

and an ability to delay gratification. The formulation of this core trait will also better enable the

child to become more sensitive to the interests and desires of others, become more independent

and willing to restrain their behaviors, and have a lesser propensity for using force or violence as

a mechanism for resolving disputes or achieving their ends.

Gottfredson and Hirschi (1990:90-91) further argue that the six primary elements that

characterize individuals who exhibit low self-control are impulsivity, insensitivity, physically as

opposed to mentally oriented, short-sightedness, risk-takers, and displaying a low tolerance for

frustration. Although these six components are known to be distinct personality and behavioral

traits, these various dimensions are all believed to be established early on in childhood and

coalesce within individuals to manifest a single construct that remains persistent throughout the

life course. The fact that these elements fuse together to generate a unitary construct, what









that the semiparametric Poisson-based regression analyses produced these three distinct classes

of offenders and that (1) the non-offenders were the most prevalent of the groups; (2) the

adolescent-limited/peaked offenders were the most prevalent of the offending groups; and (3) the

high-rate offenders were the least prevalent group of offenders but were characterized by a

higher rate of mean offending across the observation period when compared with the adolescent-

limited/peaked offender group. However, there were a few groups that were recognized in the

larger sample trajectory analyses (low-level chronic and late-onsetters) who are not consistent

with Moffitt's taxonomy but have been discussed elsewhere (see D'Unger et al., 1998; Piquero,

2007; Sampson & Laub, 1993).

Support was also shown for the second hypothesis with sex being a significant covariate

for distinguishing group-based trajectories. Specifically, in all of the six trajectory analyses,

being male significantly increased the probability that an individual would be classified as either

an adolescent-limited/peaked or chronic offender with the non-offender group serving as the

reference group. The only exception was in the 1949 continuous resident sample (N=1,297).

While there were more males in the late-onset offender group for the cohort, being male did not

significantly increase the probability of assignment to this particular group-based trajectory.

Concerning hypothesis #3 that significant life transitions including marriage and

employment can affect desistance from crime, there was limited support for Sampson and Laub's

age-graded theory of informal social control. There was descriptive evidence at the bivariate

level suggesting that individuals in either offender group (adolescent-limited/peaked or high-rate

chronic offender group) scored lower on the stakes in conformity index when compared with

those individuals assigned to the non-offender group. The results from the ANOVA with the

1949 cohort also indicated that the stakes in conformity index was statistically associated with









Reiss & Rhodes, 1961; Wikstrom, 1991a), only a few studies have investigated the influence of

other individual and community covariates on offending. Lindstrom (1996) found that the level

of family interaction was significantly associated with juvenile offending and this relationship

was the most pronounced in more socially disadvantaged neighborhoods. Kupersmidt et al.

(1995) showed that juveniles living in single-parent families in low SES neighborhoods were

more likely to be aggressive than their counterparts in high SES neighborhoods. Relying on a

comparison of two different groups of Pittsburgh neighborhoods (i.e., underclass and non-

underclass), Peeples and Loeber (1994) indicated that the relationship between race and juvenile

offending was only apparent in the underclass neighborhoods.

Even less research has targeted the integration of individual and neighborhood effects on

offending under the auspices of the criminal career paradigm (Blumstein et al., 1986) and recent

developmental frameworks. Wikstrom (1991b) conducted one of the first studies to address this

deficiency in the existing research and found that individuals that resided in the most socially

disadvantaged areas had the highest levels of both participation and frequency of offending.

Loeber and Wikstrom (1993) also showed that not only did juveniles offend at higher rates (i.e.,

greater participation and frequency) in low SES neighborhoods when compared with those from

neighborhoods with higher SES a significantly larger proportion of the offenders in the low SES

neighborhoods became serious offenders.

In a more recent study, Wikstrom and Loeber (2000) examined whether the interaction of

risk and/or protective factors and community context had an influence on early and late-onset

offending among three samples of boys in 90 Pittsburgh neighborhoods. They found that

neighborhood socioeconomic context did not have a direct effect on the prevalence of offending

for boys who had a high number of risk factors, but for those who had balanced risk-protective










Table 5-1. Descriptive Statistics for the Macro-level and Micro-level Measures by Cohort
1942 1949
Variables
Mean (SD) Mean (SD)
Independent Variables
Macro-level
Social Disorganization Index 0.00 (1.00) 0.00 (1.00)
Owner-Property Value 9374.30 (2895.29) 9196.40 (2951.55)
% Renter-Occupied Units 34.87 (20.83) 34.99 (21.43)
% Overcrowded 1.48 (3.83)* 1.89 (4.40)*

Micro-level
Male 0.47 (0.50) 0.52 (0.50)
Life Transitions
Stakes in Conformity Index 1.71 (0.45)* 1.60 (0.53)*
Employment 0.94 (0.22) 0.94 (0.23)
Marriage 0.77 (0.42)* 0.66 (0.47)*
Black 0.06 (0.24)* 0.11 (0.31)*
Low SES Household 3.61 (2.08)* 3.25 (1.91)*
Family Stability (Natural Parents) 0.84 (0.37)* 0.89 (0.31)*
Family Structure (No. of Children) 3.74 (2.21) 4.02 (2.17)
Differential association 0.22 (0.42) 0.25 (0.43)
Formal Social Control 1.24 (0.96) 1.32 (1.03)
Moved 0.61 (.0.49)* 0.49 (0.51)*

Police Contacts 1.29 (2.63)* 2.33 (6.71)*
Prevalence 37.9% 41.2%
Incidence 302* 754*


Note. The asterisk signs denotes significant mean differences between the 1942 and 1949 cohorts (p<.05). Listed
below the social disorganization and stakes in conformity indexes are the values of the measures that were used to
create each index prior to the data manipulations (principal components analysis for the social disorganization index
and the additive index for experiencing either none, 1, or 2 of the life transitions).


Table 5-2. Tabulated Bayesian Information Criteria (BIC) for 1942 Racine Continuous
Residents (N=633)
Optimal # of Groups BIC
Cubic ZIP Model
2 -2453.58
3 -2437.23
4 -2399.77
5 False Convergence









that the concentrated effects of poverty cannot be fully explained without including control

measures for both familial and individual-level characteristics (see Gephart, 1997:14). Although,

while social disorganization theory (as originally formulated) sought to explain crime at the

aggregate level, there are empirically-based reasons to expect that social disorganization may

also have an effect on individual behavior. Simcha-Fagan and Schwartz took a step toward this

possibility:

An aggregate-level theory cannot be expected to provide a comprehensive explanation of
individual delinquency or criminality, it should clarify those contextual effects which,
independently of or in interaction with the adolescent's predisposition, affect individual
behavior (1986:688).

Jencks and Mayer (1990) provided a review of the current extant knowledge on the effects

that social compositional measures (e.g., poverty) have on a series of developmental outcomes

including: educational attainment, cognitive skills, criminal activity, and economic success.

They indicated that the research is generally mixed at best regarding the influence of average

SES (in the aggregate) on the likelihood of an individual's chances for planning to go to college,

attending college, and/or graduating from college, especially for Whites. Research has shown

that youth from high SES neighborhoods are more likely to have higher educational attainment

compared with their counterparts who come from lower SES neighborhoods. In addition, Black

men's chances of competing in the job market and finding well-paid jobs were greatly reduced

when they resided in a heavily concentrated Black or welfare dependent neighborhood (see also

Aber et al., 1997).

Klebanov et al. (1997) provided an extension of the research on the influences of

neighborhood effects on crime by specifically testing whether or not these effects are mediated

and/or moderated (or interact with) by family processes such as maternal warmth, cognitive

stimulation in the home, maternal coping style, availability of social support, and the physical









Duncan, G.J., Brooks-Gunn, J.P., & Klebanov, P.K. (1994). Economic deprivation and early-
childhood development. ChildDevelopment, 65, 296-318.

Duncan, G.J., & Raudenbush, S.W. (1999). Assessing the effect of context in studies of child and
youth development. Educational Psychologist, 34, 29-41.

D'Unger, A.V., Land, K.C., McCall, P.L, & Nagin, D.S. (1998). How many latent classes of
delinquent/criminal careers? Results from mixed Poisson regression analyses. American
Journal of Sociology, 103, 1593-1630.

D'Unger, A.V., Land, K.C., & McCall, P.L. (2002). Sex differences in age patterns of
delinquent/criminal careers: Results from Poisson latent class analyses of the
Philadelphia Cohort Study. Journal of Quantitative Criminology, 18, 349-375.

Eggleston, E.P., Laub, J., & Sampson, R.J. (2004). Methodological sensitivities to latent class
analysis of long-term criminal trajectories. Journal of Quantitative Criminology, 20, 1-
26.

Elder, G.H. (1975). Age differentiation and the life course. AnnualReview of Sociology, 1, 165-
190.

Elder, G.H. (1985). Perspectives on the life course. In Glen H. Elder, Jr. (ed.), Life course
dynamics: Trajectories and transitions, 1968-1980. Ithaca: Cornell University Press.

Elliott, D.S., Ageton, S.S., Huizinga, D., Knowles, B.A., & Canter, R.J. (1983). The prevalence
and incidence of delinquent behavior: 1976-1980. Boulder, CO: Behavioral Research
Institute.

Elliot, D.S., Huizinga, D., & Morse, B. (1986). Self-reported violent offending-A descriptive
analysis of juvenile violent offenders and their offending careers. Journal of
Interpersonal Violence, 1, 472-514.

Elliot, D.S. (1994). Serious violent offenders: Onset, developmental course, and termination-
The American Society of Criminology 1993 address. Criminology, 32, 1-21.

Elliott, D.S., Wilson, W.J., Huizinga, D., Sampson, R.J., Elliott, A. & Rankin, B. (1996). The
effects of neighborhood disadvantage on adolescent development. Journal of Research in
Crime andDelinquency, 33, 389-426.

Ensminger, M.E., Lamkin, R.P., & Jacobson, N. (1996). School leaving: A longitudinal
perspective including neighborhood effects. Child Development, 67, 2400-2416.

Entwisle, D.R., Alexander, K.L., & Olson, L.S. (1994). The gender gap in math: Its possible
origins in neighborhood effects. American Sociological Review, 59, 822-838.

Farrington, D.P. (1986a). Age and crime. In M. Tonry & N. Morris (eds.), Crime and Justice,
Vol. 7. Chicago, IL: University of Chicago Press.









Table 5-3. Mean (Median) Posterior Probabilities for Group Assignments: 1942 Racine
Continuous Residents (N=633)
Group Prob Prob Prob Prob 25th 75th
(G1) (G2) (G3) (G4) Percentile Percentile
1 (n=463) .91(.99) .02(.00) .07(.01) .00(.00) .89 .99
2 (n=56) .02(.00) .84(.87) .12(.08) .03(.01) .71 .98
3 (n=82) .11(.03) .04(.00) .85(.93) .01(.00) .75 .98
4 (n=32) .00(.00) .04(.00) .05(.00) .91(.99) .91 .99
Note: Minimum and Maximum values presented for the group/prob(group) combination.


Table 5-4. Coefficient Estimates for Effect of Sex (being Male) on Trajectory Group
Membership: 1942 Racine Continuous Residents (N=633)
Groups coefficient SE p-value
Cubic ZIP Model
2 2.861 0.566 <0.001***
3 2.721 0.508 <0.001***
4 2.642 0.642 <0.001***
*p<.05 **p<.01 ***p<.001
a Group 1 (the lowest rate group) is the reference category of comparison.


Table 5-5. Tabulated Bayesian Information Criteria (BIC) for 1942 Racine Continuous
Interviewed Residents (N=333)
Optimal # of Groups BIC
Cubic ZIP Model
2 -1164.02
3 -1148.61
4 False Convergence









group begin offending relatively early in life (some even as early as age 7) and they maintain a

rather high amount of police contacts throughout the remainder of the observation period but

does decline in young adulthood. Although, it does appear that these individuals peak in late

adolescence they still exhibit high levels of offending in early adulthood. No individual in this

group accumulated less than 5 police contacts and a handful of offenders in this group had even

amassed more than 20 police contacts before reaching age 26. Overall, this small group of

offenders was responsible for 352 total police contacts between the ages of 6 and 25.

When turning to an examination of the mean posterior probabilities for group assignment

(see Table 5-3), the mean assignment probabilities for each group are noticeably high suggesting

that there was little ambiguity in the model when determining group assignments. For example,

for G1, the mean posterior probability was .91, and the next highest was G3 with a mean

probability of .07 of being assigned to G1; for group 2, the mean posterior probability was .84

(median=.87); for group 3 it was .85 (median=.93); and for group 4 it was .91 (median=.99).1

The results of this trajectory analysis provided evidence that sex was a highly significant

covariate for determining group assignment (see Table 5-4). With G1 serving as the reference

group (i.e., the non-offender group), being male significantly increased an individual's likelihood

for being assigned to any of the three offender groups (i.e., adolescent-limited/peaked, low-level

chronic, and high-level chronic).

Table 5-5 illustrates the BIC criteria for assisting in model selection for the trajectory

analysis using the reduced random sample of interviewed Racine continuous residents and the

average and predicted group-based offending trajectories are displayed in Figure 5-2. These



1 All average posterior probabilities are above the standard cut-off of .70 (Nagin, 2005).While there is always the
possibility of mis-classification of individuals to particular trajectories none of the posterior probabilities in these
data were below .80 thus indicating a high degree of discrimination between groups.









failed to find much support for the usefulness of an integrated theoretical model for explaining

group-based offending trajectories. Once the micro-level factors (particularly sex, natural two

parent household, having moved, and low SES households depending on which cohort mlogit is

being referred to) were added to the model, social disorganization failed to exert a significant

independent effect on predicting group membership. Although, the various indicators of model

fit (chi-square, log likelihood, and the psuedo R-square statistics) showed that the integrated

model was a better fit, it appeared that the influence of social disorganization on increasing an

individual's likelihood for being in either an adolescent-limited/peaked or chronic offender

group tended to be indirect in that it operated through individual and family-level factors.

Overall, these findings tend to mirror those produced in prior multi-level studies that have shown

that the addition of neighborhood level processes typically only increase the variance explained

by 1% to 4% (see Cheong & Raudenbush, 2000; Duncan & Raudenbush, 1999; Elliot et al.,

1996; Simcha-Fagan & Schwartz, 1986), although other studies have begun to find more support

for school contextual effects ranging from a 6% to 7% increase in the variance explained

(Anderson, 2002; Welsh et al., 1999).

Limitations

As with all research, there are a series of data limitations that are worth mentioning. First,

the most apparent limitation relates to the era wherein these individuals were raised. These

individuals were either born during WWII (1942 cohort) or baby-boomers (i.e., 1949 cohort)

who grew up in the immediate post-WWII era. They were living their early years during the

1960s, which is a time characterized by large social change including the Civil Rights and

Women's Movement, Vietnam, riots, etc. Thus, it is likely that some of these findings may not

necessarily generalize to individuals who were socialized in different periods of time, say the

1980s or 1990s. Second, the area of Racine itself was a Midwestern industrial city with a









Research has also shown that living in a high SES neighborhood increased older

adolescents' chances of graduating high school and attending college (Brooks-Gunn et al., 1993;

Duncan, 1994; Ensminger et al., 1996; Halpern-Felsher et al., 1997). All of these studies

suggested that the protective effect of living in a higher SES neighborhood was more salient for

males. Crane (1991) has also provided evidence that males had the worst outcomes when living

among low SES neighbors. In addition, a series of studies reviewed earlier have all consistently

shown that living in a low SES neighborhood was positively associated with higher rates of

criminal and delinquent behavior (Braithwaite, 1979; Lindstrom, 1996; Loeber & Wikstrom,

1993; Lynam et al., 2000; Peeples & Loeber, 1994; Piquero et al., 2005; Reiss & Rhodes, 1961;

Sampson & Groves, 1989; Wikstrom, 1991a, 1991b; Wikstrom & Loeber, 2000).

The U.S. Department of Housing and Urban Development (HUD) initiated what is known

as the Moving to Opportunity for Fair Housing Demonstration (or MTO) in 1994 motivated by

the success of the Gautreaux Program that has been in operation in Chicago for almost three

decades. Rosenbaum (1995; see also Rosenbaum & Popkin, 1991) presented quasi-experimental

evidence that Gautreaux youth who moved to the suburbs greatly increased their educational

attainment and job opportunities when compared with those that remained in the poor urban

areas. Intrigued by these findings and the idea that residential mobility may benefit individuals

from disadvantaged communities, the MTO program became operational in five selected sites

(Baltimore, Boston, Chicago, Los Angeles, and New York City).

The MTO program involved randomly providing families vouchers allowing some families

with children in high poverty areas to move to less poor areas with the hope that this relocation

would increase their educational and employment opportunities. Eligibility for the program was

determined by the families that lived in Section 8 public housing units with at least a 40%









Pratt, T.C., Turner, M., & Piquero, A.R. (2005). Parental socialization and community context: A
longitudinal analysis of the structural sources of low self-control. Journal of Research in
Crime andDelinquency, 41, 219-243.

Pratt, T.C., & Cullen, F.T. (2005). Assessing macro-level predictors of theories of crime: A
meta-analysis. In M. Tonry (ed.), Crime and Justice (vol. 32). Chicago: University of
Chicago Press.

Provence, S., & Naylor, A. (1983). Wo king ni ith disadvantaged parents and children: Scientific
issues and practice. New Haven, CT: Yale University Press.

Raudenbush, S.W. (2005). How do we study "what happens next"? The ANNALS of the
American Academy of Political and Social Science, 602, 131-144.

Reiss, A.J., & Lewis, A.L. (1961). The distribution of juvenile delinquency in the social class
structure. American Sociological Review, 26, 720-732.

Reiss, A.J. (1986). Why are communities important in understanding crime? In A.J. Reiss & M.
Tonry (eds.), Communities and Crime. Chicago: University of Chicago Press.

Reiss, A.J. (1989). Ending criminal careers. Prepared for the final report of the "Desistance-
Persistance Working Group" of the Program on Human Development and Criminal
Behavior. MacArthur Foundation and National Institute of Justice.

Roncek, D.W. (1981). Dangerous places: Crime and residential environment. Social Forces, 60,
74-96.

Rosenbaum, J.E. (1995). Changing the geography of opportunity by expanding residential
choice: Lessons from the Gautreaux Program. Housing Policy Debate, 6, 231-269.

Rosenbaum, J.E., & Popkin, S. (1991). Employment and earnings of low-income Blacks who
move to middle-class suburbs. In C. Jencks & P. Peterson (eds.), The urban underclass.
Washington, DC: Brookings Institution.

Rowe, D.C., Vazsonyi, A.T., & Flannery, D.J. (1995). Sex differences in crime: Do means and
within-sex variation have similar causes? Journal ofResearch in Crime and Delinquency,
32, 84-100.

Rutter, M. (1983). Developmental neuropsychiatry. New York, NY: Guilford.

Rutter, M. (1989). Pathways from childhood to adult life. Journal of Child Psychology and
Psychiatry, 30, 23-51.

Sampson, R.J. (1987). Urban black violence: The effect of male joblessness and family
disruption. American Journal of Sociology, 93, 348-382.

Sampson, R.J. (1995). The community. In J.Q. Wilson & J. Petersilia (eds.), Crime. San
Francisco, CA: ICS, Inc









In short, recent theoretical and methodological advancements in criminological research

permits me to examine the possible generalizability of formerly discussed offender groups (non-

offenders, adolescent-limited/peaked offenders, and chronic offenders) and exploring a potential

integration of macro-level and micro-level theories for explaining offending. These issues along

with a description of the methodology and analytic procedures that will be utilized in this current

study will be the focal point of the following chapter (Chapter 4) and the results chapter (Chapter

5) will involve a systematic exploration into the several key hypotheses mentioned below.

Hypotheses and Expectations


Hypothesis #1: According to Moffitt's (1993) taxonomy and prior applications of trajectory
analysis (for review, see Piquero, 2007), it is expected that the semiparametric Poisson-based
regression analyses will produce three distinct classes of offenders including: 1) non-offenders;
2) adolescent-limited; and 3) chronic offenders.

Hypothesis #2: In accordance with the theoretical extensions provided by Moffitt (1994) and
Moffitt et al. (1994), it is expected that sex with be a significant covariate for distinguishing
group-based trajectories or in other words, males will be more likely to be in offending groups
than females.

Hypothesis #3: In line with Sampson and Laub's age-graded theory of informal social control, it
is hypothesized that individuals who self-report having experienced these key life events (or
stakes in conformity) will be less likely to be chronic offenders.

Hypothesis #4: It is expected that individuals whose neighborhood of origin is characterized by
poverty/low neighborhood socioeconomic status and residential instability (i.e., social
disorganization) will be more likely to be chronic offenders than adolescent-limited offenders or
non-offenders.

Hypothesis #5: Macro-level and micro-level effects on the participation and frequency of
offending will still be observed net of the effects of each other-or in other words, both macro-
level and micro-level effects will be significantly related to latent class membership. However,
the possibility of mediation will also be explored.









unsupervised teenage peer groups, and low organizational participation) mediated a rather

significant portion of the direct effect of family disruption on crime/delinquency rates as well as

mediating the effects of other structural characteristics such as classical social disorganization

measures (low SES, residential mobility, and ethnic heterogeneity).

In sum, the social disorganization framework provides a backdrop for explaining why

crime rates may be higher in communities that are characterized by high population turnover,

ethnic/racial heterogeneity, and residential instability, but just how social disorganization can

also be expected to have an affect on individual outcomes will become clearer in the next

Chapter (Ch. 3). The next section (along with the remaining sections in Chapter 2) illustrates a

complete shift in theoretical framework by not focusing on explanations of crime in the

aggregate but by presenting a series of criminological paradigms and micro-level developmental

theories that have been offered to explain individual variation in crime.

Age-Crime Relationship and Criminal Careers

Perhaps the most commonly accepted fact about crime is the association between age and

crime (Cohen & Land, 1987; Hirschi & Gottfredson, 1983; Hirschi & Gottfredson, 1985;

Farrington, 1986a; Greenberg, 1985; Obrien, Stockard, & Isaacson, 1999; Sampson & Laub,

1990, 1992; Shavit & Rattner, 1988; Steffensmeier, Allan, Harer, & Streifel, 1989;

Steffensmeier, Streifel, & Shihadeh, 1992). The progression of what has been referred to in the

literature as the "age-crime curve" is that the crime rate, as measured by arrest, typically

increases from the minimum age of criminal responsibility and peaks in the teenage years

followed by a decline in early adulthood (Farrington, 1986a). Although this general statement

accurately characterizes the age-crime curve, in order to fully understand the complexity the age-

crime relationship it is necessary to revisit Hirschi and Gottfredson's (1983) original article

discussing this phenomenon entitled Age and the Explanation of Crime.









conviction records. Comparatively, it may also prove fruitful to examine the possible multi-level

effects on life-course patterns of offending based on self-reported criminal involvement.

Arguably, the best test would be to include some type of composite measure that represented a

combination of official and self-reported offending.

As far as the relatively weak found support for the macro-level effects, it would likely be

the case that future studies accounting for the processes identified by prior researchers (see

Bursik, 1988; Sampson, 1997; Sampson & Groves, 1989; Sampson et al., 1997) such as family

disruption (percentage of divorced or single-parent households), collective efficacy, social

capital, supervision of teenage peer groups, and local participation in voluntary organizations

may yield stronger support for macro-level influences on offending patterns and especially net of

the effects of micro-level factors. Future mutli-level studies should also make use of other

macro-level theoretical constructs to investigate their relative impact on trajectories of criminal

behavior including routine activities theory (see Cohen & Felson, 1979; Kennedy & Baron,

1993) and urban disadvantage perspectives (see Krivo & Peterson, 1996; Parker et al., 2005).

With attention to these aforementioned suggestions, it would seem that perhaps the ideal

research design for best examining the hypotheses highlighted in this research would be a

prospective longitudinal design which would identify individuals at birth in different social

contexts and follow them throughout their life-course and well into adulthood. Not only would

official and self-report data be gathered on the individuals at various periods over their life-

course, the optimal design would include both parent, teacher, and peer surveys throughout

different developmental periods. Additionally, surveys administered to community residents

over time would also allow for measuring the informal networks of social control, collective

efficacy, and social capital. Finally, if spousal surveys in addition to self-reports were also









2
1.5
1
0.5
0


6 7 8 9 10 11 12 13 14 15 16 17 18 19 2021 22 23 24 25
6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25


Age


--G1 pred -- G2 pred


G3 pred


G4 pred


-* G1 amg


--G2 avg


- G3 avg G4 avg


Figure 5-1. 1942 Racine Continuous Residents Group-Based Trajectories (N=633)









posited as a precursor of social disorganization. Dwelling units that are over-crowded also may

lead to sense of "anonymity" between the residents because of their failure to interact with their

neighbors (see Roncek, 1981; Sampson, 1995).

The results from principal components analysis confirmed that these three variables were

measuring a similar construct (i.e., social disorganization) across the two cohorts as presented

below in Table 4-3; therefore, the corresponding component loadings were utilized in order to

construct the appropriate indexes.5 This procedure allows for preserving the integrity of the data

by removing the potential of a partialing fallacy (Gordon, 1968) and reducing the potential for

multicollinearity which is common in macro-level data (Land, McCall, & Cohen, 1990). Thus,

positive z-scores were indicative of more socially disorganized areas.

Micro-level independent variables

Stakes in Conformity. Sampson and Laub (2005c) argue that involvement in institutions

such as marriage and work provide both short-term and long-benefits for individuals. These life

transitions can change an individual's motivation for crime by re-focusing their attention toward

spending time with their spouse and/or investing time in their job. These events are believed to

foster an individual's development of prosocial adult bonds. Prior research has shown support

for the impact that these transitions can have on desistance from crime (Sampson & Laub, 1993;

Warr, 1998) and other research has shown that the effect of these life transitions are age-graded

(see Uggen, 2000).

Therefore, in accordance with Sampson and Laub's (1993) theory and prior research (see

Piquero et al., 2002) a stakes in conformity index was created for the two local life circumstances



5 This procedure involves multiplying the raw variable scores by the standardized weights obtained as a result of the
principal components analyses and then summing these values together to produce a single measure (see Kim &
Mueller, 1985).









typologies; however, this does not mean that Nagin's and Land's latest methodological technique

has been insulated from theoretical scrutiny. The group-based modeling technique has been the

source of recent criticisms and counterpoints between its originators and two prominent life-

course theorists, Sampson and Laub.

This debate began with the findings from an earlier study by Sampson and Laub (2003)

questioning the stability of offending patterns over the life course and more specifically with an

article published in the Journal of Quantitative Criminology, entitled "Methodological

sensitivities to latent class analysis of long-term criminal trajectories." The authors (Eggleston,

Laub, & Sampson, 2004) argued that the "enthusiasm" for the use of Nagin and Land's (1993)

semiparametric group-based models has not been equally supplemented with robustness and

sensitivity analyses that are often inherent in offending data.

They proceed by summarizing the obvious advantages of these models (see also Nagin,

2005) that include (1) the fact that they are designed for longitudinal data; (2) they lead to the

ability to distinguish groups of offenders that are homogenous within their trajectory but distinct

from other trajectories; (3) they provide an objective systematic criterion for establishing

offending groups by using the Bayesian Information Criterion (BIC) rather than presupposing

subjective sub-groups of offenders; (4) posterior probabilities are used to assess the precision of

group assignment; and (5) this method accommodates missing data, whereby offenders that have

unavailable information for several time periods are not excluded from the analysis (Eggleston et

al., 2004:2).

Relying on 500 juvenile delinquents' data from the archives of the classic Gluecks'

Unraveling Juvenile Delinquency research (1950, 1968), Eggleston et al. (2004:23) concluded

that the length of follow-up can affect the trajectory patterns observed and argue that data that









REFERENCES


Aber, J.L., Gephart, M.A., Brooks-Gunn, J., & Connell, J.P. (1997). Development in context:
Implications for studying neighborhood effects. In J. Brooks-Gunn, G. Duncan, & J.L.
Aber (eds.), Neighborhoodpoverty: Context and consequences for children. New York:
Sage Foundation.

Aguilar, B., Sroufe, L.A., Egeland, B., & Carlson, E. (2000). Distinguishing the early-onset-
persistent and adolescent-onset antisocial behavior types: From birth to 16 years.
Developmental and Psychopathology, 12, 109-132.

Akers, R.L. (1985). Deviant behavior: A social learning approach. Belmont, CA: Wadsworth.

Akers, R.L., Krohn, M.D., Lanza-Kaduce, L., & Radosevich, M. (1979). Social learning and
deviant behavior: A specific test of a general theory. American Sociological Review, 44,
635-655.

Akers, R.L. (1977). Deviant behavior: A social learning approach. 2nd edition. Belmont, CA:
Wadsworth.

Allison, P.D. (1999). Multiple regression: A primer. Thousand Oaks, CA: Sage Publications.

Anderson, E. (1990). Streetwise: race, class, and change in an urban community. Chicago:
University of Chicago Press.

Anderson, A.L. (2002). Individual and contextual influences on delinquency: The role of the
single-parent family. Journal of Criminal Justice 30, 575-87.

Agresti, A. (1990). Categorical data analysis. New York: John Wiley and Sons.

Bandura, A. (1982). The psychology of chance encounters and life paths. American
Psychologists, 37, 747-755.

Barrett, A.E., & White, H.R. (2002). Trajectories of gender role orientations in adolescence and
early adulthood: A prospective study of mental health effects of masculinity and
femininity. Journal ofHealth and Social Behavior, 43, 451-468.

Bartusch, D.R., Lynam, D.R., Moffitt, T.E., Silva, P.A. (1997). Is age important? Testing a
general versus a developmental theory of antisocial behavior. Criminology, 35, 13-48.

Bellair, P.E. (1997). Social interaction and community crime: Examining the importance of
neighbor networks. Criminology, 35: 677-703.

Berreuta-Clement, J.R., Schweinhart, L.J., Barnett, W.S., Epstein, A.S., Weikart, D.P. (1984).
Changed lives: The effects of the Perry Preschool Program ony ,inhi\ through age 19.
Ypsilanti, MI: High/Scope Press.

Berry, B.J., & Kasarda, J.D. (1977). Contemporary urban ecology. New York: Macmillan.









TRAJECTORIES OF TWO RACINE BIRTH COHORTS: A THEORETICALLY
INTEGRATED MODEL FOR EXPLAINING OFFENDING




















By

WESLEY G. JENNINGS


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA

2007









Silverthorn, P., Frick, P.J., & Reynolds, R. (2001). Timing of onset and correlates of severe
conduct problems in adjudicated girls and boys. Journal ofPsychopathology and
Behavioral Assessment, 23, 171-181.

Simcha-Fagan, O., & Schwartz, J.E. (1986). Neighborhood and delinquency: An assessment of
contextual effects. Criminology, 24, 667-703.

Singer, J.E., Westphal, M., & Niswander, K.R. (1968). Sex differences in the incidence of
neonatal abnormalities and abnormal performance in early childhood. Child
Development, 39, 103-112.

Skogan, W.G. (1990). Disorder and decline. New York: Free Press.

Snyder, H. (1988). Court careers ofjuvenile offenders. Washington, DC: U.S. Department of
Justice, Office of Juvenile Justice and Delinquency Prevention.

Snyder, J., & Patterson, G. (1987). Family interaction and delinquent behavior. In H. Quay (ed.),
Handbook ofjuvenile delinquency. New York: Wiley.

Steffensmeier, D.J., Allan, E.A., Harer, M.D., & Streifel, C. (1989). Age and the distribution of
crime. The American Journal of Sociology, 94, 803-831.

Steffensmeier, D.J., Streifel, C., & Shihadeh, E.S. (1992). Cohort size and arrest rates over the
life course: The Easterlin Hypothesis revisited. American Sociological Review, 57, 306-
314.

Sutherland, E.H. (1947). Principles of criminology. 4th ed. Philadelphia: Lippincott. Sutherland,
E.H. (1939). Principles of criminology. 3rd edition. Philadelphia: Lippincott.

Taylor, R.B. (1995). The impact of crime on communities. Annals of the American Academy of
Political and Social Science, 539, 28-45.

Taylor, R.B., & Covington, J. (1988). Neighborhood changes in ecology and violence.
Criminology, 26, 553-589.

Thornberry, T.P., & Krohn, M.D. (2003). The self-report method for measuring delinquency and
crime. In D. Duffee (ed.), Criminal justice 2000, volume 4, measurement and analysis of
crime. Washington, DC: National Institute of Justice.

Tibbetts, S.G., & Piquero, A.R. (1999). The influence of gender, low birth weight and
disadvantaged environment in predicting early-onset of offending: A test of Moffitt's
interactional hypothesis. Criminology, 37, 843-878.

Tittle, C.R. (1988). Two empirical regularities (maybe) in search of explanation: Commentary on
the age-crime debate. Criminology, 26, 75-86.

Tonry, M., Ohlin, L.E., & Farrington, D.P. (1991). Human development and criminal behavior.
New York: Springer-Verlag.









occurred in the first place (see Piquero et al., 2002). The differential association measure is also

noticeably weak.

Regarding the dependent variable (police contacts), there are also well known strengths

and weaknesses from using either official data (police contact, arrest, conviction) or self-report

data to measure criminal behavior. Under-reporting is one of the principal limitations of official

statistics because the majority of police activity is a direct result of the willingness of citizens to

report crimes (see Hindelang et al., 1981; Thornberry & Krohn, 2003). However, there are also

extensive disparities in victim reporting based on various factors including: requirements for

insurance companies, confidence or lack of confidence in police departments, etc. Furthermore,

there are many offenders who will never come to the attention of law enforcement, such as

shoplifters, burglars, or white-collar criminals (Junger-Tas & Marshall, 1999). This result might

be due to the fact that more skillful and shrewd offenders are less likely of being arrested and/or

may reflect the differences in police discretion and priorities across regions/agencies (Blumstein

& Cohen, 1987).

One key advantage of self-report measures is that they often offer more comprehensive

information than official data. They provide background and sociodemographic data on

individual offenders, as well as personal characteristics. Research has indicated that self-reports

are effective in answering questions of causality, motivation, change, and stability in criminal

behavior (Junger-Tas & Marshall, 1999:296). Self-reports also tend to include both a broader

range of deviant behavior and less serious types of crimes than official police records indicate

(Blumstein, Cohen, Roth & Visher, 1986:32). Some examples of the types of crimes that are

likely to be found in self-reports but not in official data are prostitution, drug abuse, public order

crimes, and drunk driving.









Moffitt, T.E., & Caspi, A. (2001). Childhood predictors differentiate life-course persistent and
adolescent-limited antisocial pathways among males and females. Developmental and
Psychopathology, 13, 355-375.

Moffitt, T.E., Caspi, A, Rutter, M., & Silva, P.A. (2001). Sex differences in antisocial behavior:
Conduct disorder, delinquency, and violence in the Dunedin Longitudinal Study.
Cambridge, UK: Cambridge University Press.

Nagin, D.S. (1999). Analyzing developmental trajectories: A semiparametric, group-based
approach. Psychological Methods, 4, 139-157.

Nagin, D.S. (2004). Response to "Methodological sensitivities to latent class analysis of long-
term criminal career trajectories." Journal of Quantitative Criminology, 20, 27-35.

Nagin, D.S. (2005). Group-based modeling of development. Cambridge, MA: University of
Harvard Press.

Nagin, D.S., & Paternoster, R. (1991). On the relationship of past to future delinquency.
Criminology, 29, 163-189.

Nagin, D.S., & Farrington, D.P. (1992). The onset and persistence of offending. Criminology, 30,
301-523.

Nagin, D.S., & Land, K.C. (1993). Age, criminal Careers, and population heterogeneity:
Specification and estimation of a nonparametric, mixed Poisson model. Criminology, 31,
327-362.

Nagin, D.S., Farrington, D.P., & Moffitt, T.E. (1995). Life-Course trajectories of different types
of offenders. Criminology, 33, 111-140.

Nagin, D.S., & Waldfogel, J. (1998). The effect of conviction on income through the life cycle.
International Review ofLaw and Economics, 18, 25-40.

Nagin, D.S., & Tremblay, R.E. (2001). Analyzing developmental trajectories of distinct but
related behaviors: A group-based method. Psychological Methods, 6, 18-34.

Nagin, D.S., & Tremblay, R.E. (2005a). Developmental trajectory groups: Fact or a useful
statistical fiction. Criminology, 43, 873-904.

Nagin, D.S., & Tremblay, R.E. (2005b). From seduction to passion: A response to Sampson and
Laub. Criminology, 43, 915-918.

Nagin, D.S., & Tremblay, R.E. (2005c). What has been learned from group-based trajectory
modeling? Examples from physical aggression and other problem behaviors. The
ANNALS of the American Academy of Political and Social Science, 602, 82-117.









third building block, i.e., the importance of adult social bonds. They propose that social bonds

developed in adulthood (employment, marriage) can explain involvement/lack of involvement in

crime regardless of any underlying individual criminal propensity. In addition, they emphasize

that it is not only the mere exposure of individuals to these bonds, the quality and intensity of

these bonds is critical. Drawing on the work of Elder (1975, 1985), Sampson and Laub also

hypothesize that different adaptations to key life events can affect an individual's trajectory, and

crucial turning points such as divorce or death of a love one can redirect life trajectories.

Two additional important concepts in Sampson and Laub's age-graded theory of informal

social control include the processes of self-selection and cumulative continuity. Self-selection is

the classic "birds of a feather flock together" argument, and is based on the belief that

individuals with low self-control resulting from ineffective parenting systematically sort

themselves into criminogenic environments. For example, delinquent youth will tend to

gravitate toward delinquent peers in adolescence, and similarly, will find delinquent spouses in

adulthood and continue their offending behavior (Gottfredson & Hirschi, 1990).

In comparison, cumulative continuity (Moffitt, 1993) integrated with the concept of state-

dependence (Nagin & Paternoster, 1991) refers to the idea that involvement in delinquency has

an incremental effect on the interpersonal social bonds formed in adulthood (i.e., labor force

attachment, marital cohesion). For instance, an arrest and subsequent incarceration in

adolescence may lead the youth toward dropping out of school. This practice would logically

affect their future job prospects and thus result in the failure of the individual to develop strong

adulthood bonds to the labor force, which inevitably increasing their likelihood of involvement

in crime (Tittle, 1988).









poverty rate according to 1990 census tract data. Prior MTO evaluations have indicated that

residential relocation significantly contributed to a reduction in violent arrests for males and

better overall health among the children that moved from the public housing units (Katz et al.,

2001; Ludwig et al., 2001). Katz et al. (2001) also noted significant reductions in criminal

victimization among individuals that moved out of the Section 8 housing.

Leventhal and Brooks-Gunn (2003) noted that the MTO program had profound effects on

mental health outcomes for the individuals and families that relocated. They reported that the

parents who moved reported less distress than did those that remained and the boys that moved

(especially those between ages 8 and 13) reported significantly less depression, anxiety, and/or

dependency problems when compared with their counterparts who were still in public housing.

Most recently, Kling et al. (2007) noted the substantial mental health and education benefits for

the adults and females, but commented on the adverse effects associated with the males who

received vouchers (for an extended review of MTO impact evaluations, see also Orr et al., 2003).

Overall, there is reason to believe that structural factors including those highlighted in

social disorganization theory can have an effect on individuals' outcomes and perhaps influence

their pathway of offending. It is likely that living in areas characterized by social disorganization

will increase an individual's likelihood of being a chronic offender. This concept of chronic

offending is the topic of the discussion presented in the following section.

Prevalence and Incidence of Chronic Offending

Relying on data from the well-known Philadelphia Birth Cohort Study, Wolfgang, Figlio,

and Sellin (1972) provided some of the earliest evidence of the existence of chronic offenders.

Wolfgang et al. found that there were a group of delinquents that represented only 6% of the

cohort and 18% of the cohort's delinquents but were responsible for committing roughly half of

all the offenses and about two-thirds of all of the violent offenses. In a ten-year follow-up study









disaggregated trajectories and then estimating mlogit models to see whether or not the macro-

level and micro-level factors, such as those examined here, significantly differ across sex.

Unfortunately, there were not enough individuals in the analyses relied on in this study to permit

the estimation of sex-specific trajectories and make use of all of the risk/protective factors in the

mlogit models. Further studies employing considerably larger samples would permit this type of

analysis. It is worth noting that there are other theories that may account for sex differences as

well. One such example may be power-control theory (Hagan et al., 1985). It is possible that

parents and society socializes males and females differently and this may have an impact on sex-

specific pathways of offending. This possibility should be further explored.

Larger samples with multi-level data would also allow future researchers to examine these

effects on distinguishing between offender groups for groups that were uncovered in some of the

trajectory analyses that were not consistent with Moffitt's typology such as the low-level

chronic and the late-onset offenders. It may be the case that social disorganization matters more

so for influencing low-level chronic offenders with low yet relatively stable offending rates over

the life-course. Similarly, it seems reasonable to assume that the absence of the various adult

social bonds identified by Sampson and Laub (1993) or a low level of attachment to the

institutions of marriage and employment may serve as explanations for a late-onset pattern of

offending. Additionally, future studies investigating the chronic patterns of offending should

rely on data from longer observational periods particularly through middle and late adulthood to

see if these group-based offending patterns identified in this study continue beyond age 25.

This research relied solely on an official measure of crime to estimate offense trajectories

and more specifically, police contacts. It would be interesting if future studies would attempt to

reproduce these analyses using alternative official measures of offending such as arrest or









Blackwell, B.S., & Piquero, A.R. (2005). On the relationships between gender, power control,
self-control, and crime. Journal of Criminal Justice, 33, 1-17.

Blumstein, A., & Cohen, J. (1979). Estimation of individual crime rates from arrest records.
Journal of Criminal Law and Criminology, 70, 561-585.

Blumstein, A., Cohen, J., & Hsieh, P. (1982). The duration of adult criminal careers. Final
Report submitted to NIJ. Pittsburgh, PA: School of Urban and Public Affairs, Carnegie-
Mellon University.

Blumstein, A., Cohen, J., Roth, J.A., & Visher, C.A (eds.). (1986). Criminal careers and "career
criminals, Vol 1. Report of the Panel on Criminal Careers, National Research Council.
Washington, D.C.: National Academy of Sciences.

Blumstein, A. & Cohen, J. (1987). Characterizing criminal careers. Science, 237, 985-991.

Blumstein, A., Cohen, J., & Farrington, D.P. (1988). Criminal career research: Its value for
criminology. Criminology, 26, 1-35.

Braithwaite, J. (1979). Inequality, crime, andpublic policy. London: Routledge and Kegan Paul.

Brame, B., Nagin, D., & Tremblay, R. (2001). Developmental trajectories of physical aggression
from school entry to late adolescence. Journal of Child Psychology andPsychiatry, 42,
503-513.

Broidy, L.M., Nagin, D.S., Tremblay, R.E., Bates, J.E., Brame, R., Dodge, K.A., Fergusson,
D.M., Horwood, J.L., Loeber, R., Laird, R., Lynam, D.L., Moffitt, T.E., Petit, G.S., &
Vitaro, F. (2003). Developmental trajectories of childhood disruptive behaviors and
adolescent delinquency: a six-site, cross-national study. Developmental Psychology, 39,
222-245.

Bronfenbrenner, U., Moen, P., & Garbarino, J. (1984). Child, family, and community. In R.
Parke (ed.), Review of child developmental research: Vol. 7: The family. Chicago:
University of Chicago Press.

Brooks-Gunn, J., Duncan, G.J., Klebanov, P.K., & Sealand, N. (1993). Do neighborhoods
influence child and adolescent development? American Journal of Sociology, 99, 353-
395.

Bursik, R.J. (1984). Urban dynamics and ecological studies of delinquency. Social Forces, 63,
393-413.

Bursik, R.J. (1986). Ecological stability and dynamics of delinquency. In A.J. Reiss & M. Tonry
(eds.), Crime and Justice (vol. 66). Chicago: University of Chicago Press.

Bursik, R.J., & Grasmick, H.G. (1993). Neighborhoods and crime: The dimensions of effective
community control. New York: Lexington Books.









this continuous distribution, but rather assumes that there exist distinct clusters of group-based

trajectories that exhibit different etiological processes, and the methodology approximates the

continuous distribution by identifying points of support (i.e., distinct groups or trajectories)

(Nagin, 2005). These trajectory models are also based on a maximum likelihood (ML) function,

which of course permits them to display the qualities of ML parameter estimates-they are

consistent and asymptotically normally distributed (Cramer, 1946; Greene, 1990). Given the

assumptions and statistical properties, this method assumes that individual differences in

behavioral trajectories can be described as polynomial functions of age or time. The following

formula presents the probability for belonging to a particular group:

J
P (Y,) = Z P(Y),


where P(Y,) is the unconditional probability of observing individual i's longitudinal sequence of

police contacts, Y, over time. It is equivalent to the sum across J groups of the probability of Y,

given i's membership in group weighted by the probability of membership in group (see

Nagin, 2005:25).

There are various forms and derivatives of the trajectory model, but the two most common

parametric forms that are likely to apply to the data used for this study are the censored normal

model (CNORM) and the zero-inflated Poisson model (ZIP). The CNORM model is appropriate

for data where the outcome variable of interest is normally distributed but there exists clustering

of the data at either the minimum or maximum (which typically occurs with offending data).

This model can also be applied to data without censoring by assigning the scale maximum and

minimum values to values that are respectively smaller and larger than any value observed in the

data. The equations listed below correspond to the statistical derivations of the cases when (1)











Table 5-27. 1949 Racine Continuous Interviewed Residents n/ ith Block Data: Odds Ratios and
Probabilities for Macro-level and Micro-level Predictors of Offending Group
Membership
Adolescent-Limited/Peaked High-Rate
Variables Offenders Chronic Offenders
Odds Prob. Odds Prob.
Macro-level
Social Disorganization 1.19 .28 1.09 .54

Micro-level
Male 2.33 <.001*** 2.36 .02**
Black 1.30 .13 1.20 .64
Low SES Household 1.25 .21 2.53 <.001***
Stakes in Conformity 1.10 .60 0.79 .53
Family Stability (Natural 0.63 <.001*** 0.68 .15
Parents)
Family Structure (No. of 1.33 .05* 1.62 .13
Children)
Differential association 1.24 .19 1.01 .99
Formal Social Control 1.10 .58 1.17 .64
Moved 0.86 .37 0.78 .53

Model Diagnostics
Chi-Square= 79.26***
-Log Likelihood= -169.32
McFadden's R2 .20
Nagelkerke R= .32
p<.10 *p<.05 **p<.01 ***p<.001. Note. Reference category is group 1 (the noui-cllkder.N).









Notwithstanding the mentioned benefits, there are several limitations to the self-report

method. Under-reporting and poor recall of criminal activity are disadvantages because self-

reports tend to be retrospective in nature (as was the case with the interviews in these data). In

addition, some respondents may even inflate their actual level of criminal activity. Cross-study

comparisons of self-reports may also be hindered due to differences in the wording of the

questions, differences in the administration of the survey, or differences in time periods

(Blumstein et al., 1986:33). Taking into account the noted limitations of police records and self-

reported measures of offending, this study chose to measure police contacts because they are a

(1) more accurate official measure of "true" offending than arrests and convictions, and (2)

typically closer in time to the actual criminal act than self-reported offending which sometimes

asks individuals to recall their past behavior in their lifetime, last twelve months, last six months,

etc. (see Hindelang et al., 1981; Thornberry & Krohn, 2003).

Within the context of all of the above mentioned limitations surrounding the data, time

period, samples, and the location along with the measurement shortcomings, the following

section highlights the group-based methodology that will be used to estimate the group

trajectories.

Methodology

Group-Based Modeling

Longitudinal research in its most basic form (as applied to the study of crime), involves the

identification of a cohort at birth and following that group of individuals throughout their life, or

at least long enough to capture the termination of the majority of criminal careers (Blumstein et

al., 1986:104). The longitudinal method allows for a specified cohort's level of involvement in

crime to be traced as the cohort, itself, matures (Greenberg, 1985). This knowledge is crucial for

detecting causal factors or sequences that associate individual characteristics or certain events to









criminal careers are brief and end in their teenage years and others continue to commit crimes up

to and beyond middle age.

Piquero, Brame, and Lynam (2004) have argued that the relative inattention of research

regarding the duration of criminal careers is due to the difficulty in determining the true end of a

criminal career. This difficulty has been addressed by Blumstein et al. (1986:18) where they

stated the importance of distinguishing total career length from residual career length, wherein

the later refers to the estimated time remaining in an offender's criminal career given the specific

observational period. Attention to this element is what Piquero et al. (2003) referred to as

subsequently calculating estimates of career termination probabilities after an offender's arrest.

Numerous studies (for review, see Piquero et al., 2003) have attempted to measure offender

career lengths in actual number of years producing varying results ranging from a low of five

years (Blumstein, Cohen, & Hsieh, 1982) to a relative high of 17.27 years (Piquero, Brame, &

Lynam, 2004).

Most recently, Kazemian and Farrington (2006) found evidence of a general decline in

residual career length and offending frequency over time based on data from the Cambridge

Study in Delinquent Development. The decline in residual career length was observed in the

original South London male sample and among their sons, although offending frequency did not

decline as a result of conviction for the fathers. In addition, age of onset was associated with

both residual career length and offending frequency, but the relationship was noticeably weaker

among the fathers.

The seriousness dimension of criminal career research reflects both the number of offenses

committed, as well as the particular patterns of switching (i.e., versatility) among offenses

(Blumstein et al., 1986:13). There are numerous types of offenses in which criminals can engage









commit disproportionate amounts of crime; (3) life transitions can alter individual offending

pathways; and (4) Moffitt's taxonomy can be used to explain sex differences in offending,

particularly the findings that males are more likely to be found among the life-course-persistent

offender group than the females. Noting this evidence, the following section directs attention

toward examining the stability of offender groups across recent prior applications of Nagin and

Land's group-based modeling technique, along with a discussion of the factors that have been

found to be associated with offending group membership.

Prior Applications of Group-Based Models of Offending

Similar to the theoretical developments in criminal career and developmental research,

modern innovations in methodological approaches to studying group-based modeling,

specifically the technique developed by Nagin and Land (1993) (i.e., what has been referred to

by many names including trajectory analysis, finite mixture models, or semiparametric mixed

Poisson-based regression models). The technique permits disaggregating aggregate offending

patterns into group-based offending trajectories of individuals that exhibit distinct pathways over

time (Nagin & Land, 1993; Land, McCall, & Nagin, 1996; Land & Nagin, 1996; Nagin, 2005).

This method for modeling multiple pathways of offending provides a mechanism to identify and

distinguish between certain types of behavioral trajectories. It also enables the use of regression

models such as multinomial regression in order to examine which covariates are best able to

predict which offending group that an individual is assigned.

Nagin and Land (1993) provided the first application of this methodology based on data

from the well known South London cohort of 411 males (West & Farrington, 1973, 1977).

Relying on convictions as their measure of offending, they found evidence of four groups of

offenders including a non-offender group, an adolescent-limited group, and two groups of

chronic offenders with different degrees of high-rate offending.









Table 4-1. Resident Samples by Cohort, Size, Data Sources, and Offender Groups.
Continuous-Resident Samples Sample Size (n)


1942 COHORT
Continuous-Residents
Continuous-Residents (w/ Interview Data)
Continuous-Residents (w/ Individual & Block Data)

1949 COHORT
Continuous-Residents
Continuous-Residents (w/ Interview Data)
Continuous-Residents (w/ Individual & Block Data)


633
333
235


1297
556
323


Table 4-2. Observational Periods of Data Sources by Cohort
Data Sources Years 1942 Cohort 1949 Cohort
Police Contact Data
1942 Cohort 1948-1967 6 to 25 years of age 6 to 25 years of age
1949 Cohort 1955-1974 6 to 25 years of age 6 to 25 years of age
Interview Data 1976 34 years old 27 years old
Census Data 1950 8 years old 1 year old


Table 4-3. Social Disorganization Index Factor Loadings by Cohort
1950 Census Measures 1942 1949
Owner-Property Value -.774 -.818
% Renter-Occupied Units .794 .794
% Overcrowded .612 .665

Eigenvalue 1.61 1.75
% Variance Explained 53.47 58.04
N 235 323









The initial trajectory models will be estimated separately for the 1942 and 1949 Racine

continuous-residents including sex as a covariate in the trajectory estimation (Models 1 and 2). 15

This will be followed by two models estimated similarly by cohort with sex as a covariate but

only using the random subsamples of interviewed Racine continuous-residents from the 1942

and 1949 cohorts (Models 3 and 4). These models will allow a preliminary examination of the

stability of offender groups across the cohorts and by data source. In other words, they offer a

type of sensitivity analysis for the stability of group identification that Nagin and Tremblay

(2005a) have argued has rarely been done in the trajectory literature (for exceptions see D'Unger

et al., 1998; Sampson et al., 2004).

Finally, because the purpose of this study is to examine the importance of macro-level

and micro-level factors in predicting longitudinal trajectories of offending, the 1942 and 1949

subsamples will be merged respectively with the sparse 1950 Census block data and an

additional model will be estimated for each cohort with sex as a covariate (Models 5 and 6).

These final trajectory models with the 1942 and the 1949 continuous-resident subsamples

(merged with the interview and block data) are the basis for assessing the possible independent

effects of macro-level and micro-level measures for predicting latent class membership (see also

Table 4-1 presented above that graphically displays the trajectory analyses that will be

estimated).

Multinomial logistic regression

After identifying the latent offender classes and establishing their respective participation

and frequency rates of offending, multinomial logistic regression models will be estimated in



15 Ideally, race-specific models would also be estimated also; however, due to the homogeneity of Racine in the
1940s and the small numbers of non-Whites in the continuous-resident samples, race will remain a covariate in the
multinomial regression models estimated in the following chapter.









selection effect in that individuals who do in fact get married are perhaps qualitatively different

from those who do not get married, and that individuals who are not involved with delinquent

peers and/or engaged in crime may appear to be more suitable marriage prospects.

Maume, Ousey, and Beaver (2005) have since reexamined Warr's (1998) hypotheses and

conclusions using data from the National Youth Survey, but improve upon Warr's earlier study

by specifically investigating the effects of marital attachment on desistance from marijuana use

and control for the possible selectivity bias in the relationship between marital attachment,

delinquent peer associations, and desistance. In line with Sampson and Laub's theory, Maume et

al. found that marriage was important to the desistance, particularly for those marriages that were

characterized by high levels of attachment. Contrary to Warr's (1998) findings, marriage still

maintained significant directs effects for affecting desistance net of the effect of delinquent peer

association (measured as the change in the number of delinquent associates).

Uggen (2000) investigated the impact that employment (i.e., a turning point) has on an

offender's recidivism relying on data from over 3,000 individuals from nine U.S. cities, which

were originally part of the 1977 National Supported Work Demonstration Project that targeted

the underclass and the ghetto poor. The individuals who participated in this project were

randomly assigned to either of two treatment groups (construction work or service jobs) or to a

control group where no job placement was given. The work and arrests histories were collected

for all of these individuals every nine months for three years. The results from the event history

analysis yielded a variety of noteworthy findings. First, although the program failed to reduce

crime overall across the entire sample, its impact was definitely age-graded. Job treatment

significantly reduced recidivism among the older participants (particularly aged 27 or older)

whereas the job treatment had little to no effect on crime among the young adults (late









level factors that may simultaneously affect an individuals' pathways of offending using data

from two of Shannon's (1988, 1991) Racine Birth Cohorts.1

Layout of Dissertation

This dissertation will consist of six chapters. Chapter 2 will begin with a discussion of the

various theoretical frameworks that provide the basis for the study. More specifically, it will

begin with a brief description of how scholars have defined a "community" over time followed

by an outline of the origins and theoretical tenets of social disorganization theory. Next will be

an explanation of the relationship between age and crime and how research on the age-crime

curve complements the emergence of the criminal career paradigm. Similarly, the narrative will

continue with a discussion of how the criminal career paradigm had a direct contribution to the

development of what has become known as developmental/life-course criminology. This chapter

will conclude with a presentation of Sampson and Laub's (1993) age-graded theory of informal

social control (see also Laub & Sampson, 1993) and Moffitt's (1993) taxonomy of offending, the

latter providing the theoretical rationale for hypothesizing why different classes of offenders may

exist and the former describing certain life transitions that may affect an individual's offending

trajectory.

Chapter 3 will continue by offering an in-depth exploration into the prior extant literature

that can be found examining each of the theoretical frameworks mentioned above. Most notably,

this chapter will illustrate the links between social disorganization and individual patterns of

offending, along with addressing the prior research on chronic offending, crime over the life-

course (i.e., Sampson and Laub's age-graded theory of informal social control), and offer a


1 This study relies on data from the 1942 and 1949 Racine Birth Cohorts, as opposed to all three of the cohorts
(1942, 1949, 1955), because there is not any self-reported interview data available for the 1955 cohort precluding its
use for investigating the particular individual-level factors addressed in this research. According to Shannon (1988,
1991), lack of continued federal funding prevented him from interviewing the most recent 1955 cohort.









neuropsychological disorders are more prevalent among males (Rutter, 1983; Singer et al.,

1968); (2) childhood-onset conduct problems are very rare among girls (Werry, 1986); and (3)

female delinquent behavior tends to be more erratic than male delinquent behavior (Moffitt et al.,

1994:283).

While large scale self-report studies versus those based on official data have shown that

virtually every adolescent commits some illegal behavior (see Elliot et al., 1983), according to

Moffitt (1993) there is yet another group of individuals who are characterized as abstainers or

non-offenders. Similarly, Moffitt (1993) provides some theoretical speculations as to why these

individuals fail to become involved in crime and delinquency. Moffitt argues that perhaps these

youth experience late-onset puberty and therefore do not sense the maturity gap or perceive the

social assets associated with crime (1993:689). Prior research has indicated that females that

began menstruating after the age of 15 did not show any involvement in delinquency.

The abstainers may also lack the opportunities to engage in delinquency. Adolescent

delinquency rates have been shown to be higher in urban versus rural areas (Skogan, 1990), and

Sampson and Groves (1989:789) have also commented that one of the strongest correlates of

violence is the "presence of unsupervised groups of teenagers." In another study, Farrington and

West's (1990) research with their London male cohort showed that adolescents who resided in

criminogenic environments but did not engage in delinquency could be characterized as nervous

and withdrawn and also tended to entirely lack friends or only have a few friends.

Most recently, Piquero, Turner, and Brezina (2005) provided one of the first empirical tests

investigating Moffitt's (1993) abstainer hypothesis and found that there did exist a relatively

small group of adolescents (a slightly greater proportion of females) that could be classified as

abstainers. Piquero et al. (2005) indicated that teacher attachment was inversely related to












4-
O




6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

Age

-- G1 pred -i- G2 pred G3 pred G1 avg G2 avg -- G3 avg

Figure 5-6. 1942 Racine Continuous Interviewed Residents in itl Block Data Group-Based Trajectories (N=323)









more than doubled the odds of an individual being assigned to a high-rate chronic offending

group relative to a non-offender group.5

Comparatively, Table 5-23 presents the results from the micro-level-only model. First, the

findings showed a robust effect for sex on the odds of being assigned to either type of offender

group with being male nearly tripling the odds for being assigned to an adolescent-

limited/peaked offender group and increasing the odds by more than six times for an individual

being assigned to a high-rate offender group. Having been socialized in a natural two parent

household yielded a significant protective effect on the odds of being assigned to adolescent-

limited/peaked offender group by reducing the odds by 38%, and it produced a similar reduction

in the odds (41%) of being assigned to a high-rate chronic offender group relative to the non-

offender group. In addition, having moved during the juvenile years more than doubled the odds

of being assigned to a high-rate chronic offending trajectory. Although the majority of the other

individual level risk/protective factors were in the expected direction insofar as increasing or

decreasing the odds for being in a particular group, being Black, living in a low SES household,

stakes in conformity, number of children in the household, perceived delinquent peer influence,

and perceptions of formal social control did not achieve statistical significance for distinguishing

across trajectories.

Table 5-24 displays the multivariate results for the integrated theoretical model for

predicting offender group membership. The findings suggest that once the macro-level and

micro-level factors are estimated simultaneously, social disorganization is left only marginally

5 Percent non-White was originally included in the analyses as a control variable, but it failed to exert any significant
effects for predicting offender group membership. Models were also estimated using backwards deletion and the
results confirmed that percent non-White did not influence trajectory group membership or increase the model fit, so
it was dropped and not included in the analyses presented here. According to Sampson (1997) there has been little
support for this construct in relation to offending when included with neighborhood poverty and residential
instability and given the relatively high degree of homogeneity in these data, it is not surprising that this variable
was non-significant.









turnover mentioned above. This connection is primarily due to the fact that in communities that

are characterized by rapid compositional change, it becomes difficult to promote a unified front

against the "invasion" of potentially threatening groups (Bursik & Grasmick, 1993:33).

Furthermore, heterogeneity can inhibit communication, especially in communities that are

plagued by language barriers. This will more than likely make it increasingly difficult for the

residents to solve or even express their common problems and goals (Kornhauser, 1978:75).

Shaw and McKay (1942) also indicated that high rates of poverty/low socioeconomic

status are characteristic of socially disorganized communities. According to Shaw and McKay,

this situation was attributable to the constant out-migration to the suburbs of residents that had

the ability to move, and the area that the former residents left behind still tended to be plagued by

poverty and increased crime/delinquency rates, regardless of which racial/ethnic group settled

into the area. These individuals who lack the opportunity and/or resources to "escape" this

socially disorganized area (i.e., the inner city) have been referred to by Wilson (1987) as "the

truly disadvantaged" because they have to live in areas of extreme concentrated poverty and with

a high density of subsidized housing. This density of subsidized and predominantly renter-

occupied housing has also been associated with crime rates (see Schuerman & Kobrin, 1986).

Schuerman and Kobrin (1986) have indicated that although little is known about how

neighborhoods evolve into high-crime areas, prior research has evidenced distinct stages of

neighborhood development that typify the reciprocal effects of crime and community decline.

"Emerging" is the first stage where neighborhood deterioration precedes the high-crime rates,

but once communities move through the transitional stage and into the last stage (i.e., enduring

phase), then the high-crime rates precede the further deterioration of the neighborhood. A host

of factors has been attributed to neighborhood deterioration including: movement from single to









Study Strengths

Notwithstanding the limitations noted previously in this chapter, there are several

important strengths of this study that are worth mentioning. First, this study answers the recent

calls among criminologists regarding moving toward theoretical integration of macro-level and

micro-level levels of analysis (Sampson et al., 1997; Messner et al., 1989), specifically focusing

on social disorganization theory (Shaw & McKay, 1942), Sampson and Laub's age-graded

theory of informal social control (1993), and Moffitt's (1993) offender taxonomy. Furthermore,

it approaches this process by employing a relatively new methodological tool that has gained a

substantial recognition in the field as an appropriate modeling technique for identifying distinct

behavioral patterns (for review, see Piquero, 2007).

Second, as an addition to the trajectory literature, this study provides an exploration into

the importance of sex for distinguishing group-based trajectories. Third, prior studies that have

estimated trajectories with the Shannon's (1988, 1991) Racine cohorts have relied on only

felonies and misdemeanors as the measure of offending (see D'Unger et al., 1998); therefore,

this research also includes status offenses in the hopes to better capture adolescent-limited

offending patterns. This study also improves by Land's (2000) study by estimating a series of

trajectory models as a "sensitivity" analysis to discover if similar offending trajectories are

produced by cohort and data source. This is especially of concern once the Racine continuous-

resident subsamples are merged with the interview level data, and subsequently, once the even

fewer of the individuals are merged with the limited amount of available census data

characterizing the community in which they were socialized. Finally, this study is perhaps the

first one (see Land, 2000 for a possible exception) to use trajectory analysis along with multi-

level regression models to predict the likelihood of exhibiting a distinct pattern of offending over

the life-course.









trajectory analyses this group tended to resemble an adolescent-peaked group (D'Unger et al.,

1998; Land, 2000) in that they still exhibited low rates of offending into early adulthood before

desisting in the early 20s. The discussion in the next section now moves toward examining the

results from the ANOVAs in order to further investigate what macro-level and micro-level

factors are significant for distinguishing between offender groups.

Analysis of Variance

The results from the analysis-of-variance tests where the macro-level and micro-level

risk/protective factors are compared across groups and by cohort are presented in Table 5-21.

The group means of offending (number of police contacts) are also displayed in Table 5-21. The

trajectory models displayed in Figure 5-3 (1942 cohort) and Figure 5-6 (1949 cohort) are those

that all of the remaining analyses are based on. The first most obvious findings is that relatively

all of the risk factors are concentrated among the offender groups and the protective factors are

primarily found in the non-offender group (Gl). More specifically, the risk factors tend to be

heightened in the adolescent-limited/peaked groups (G2) and then further elevated in the high-

rate chronic groups (G3) (G3 for both the 1942 and 1949 cohorts will be referred to

interchangeably as high-rate or chronic offenders hereon). For the 1942 and 1949 cohorts, social

disorganization, sex (being male), race (being Black), living a low SES household, living with

natural parents, and perceived formal social control are distinguishing features of group

membership. Additionally, whether an individual had moved during their juvenile years was a

distinguishing feature for 1942 continuous residents, whereas stakes in conformity3 and family


3 In order to further explore the possible influence of marriage and employment on group-based offending
trajectories, the mean age at time of marriage and first full-time job were computed and compared across groups and
by cohort. Regarding the 1942 cohort, 83.0% of the non-offenders (G1), 70.2% of the adolescent-limited/peaked
(G2), and 86.9% of the chronic offenders (G3) reported being married by age 25 and of those that reported being
married, the mean age of marriage was 20.8 (G1), 21.6 (G2), and 20.7 (G3). In contrast, 97.6% of Gl, all of G2,
and 95.6% reported having their first full-time job by age 25 and the mean age of first full-time employment was
18.4 (G1), 18.4 (G2), and 18.0 (G3). When turning to the 1949 cohort, 72.7% of the non-offenders (G1), 63.6% of









yielded varying levels of support for all of the theoretical frameworks mentioned above.

Specifically, all of the trajectory models discovered groups consistent with Moffitt's typology, as

well as uncovering groups not necessarily consistent with her typology. Also consistent with

Moffitt, sex was shown to be a significant covariate for predicting group assignment. Bivariate

support was found for Sampson and Laub's age-graded theory of informal social control. Lastly,

multivariate support was revealed for the impact of living in a socially disorganized area. Yet,

once mlogit models were estimated incorporating the individual and neighborhood effects there

was evidence of mediation. It appeared that individual and family variables at times completely

or at least partially mediated the effect of social disorganization as a distinguishing feature of

group membership. Implications for the current findings and directions for future research are

also discussed.









factor scores, neighborhood context mattered. In addition, they indicated that individual risk-

protective scores had a greater impact on the rate of early-onset offending (before age 13) than

did neighborhood context, but that a direct neighborhood effect was observed for late-onset

offending (after age 12).

Lynam et al. (2000) found that the effects of impulsivity on juvenile offending was

stronger in poor neighborhoods; however, living in a disadvantaged neighborhood did not

heighten nonimpulsive boys' risk for juvenile offending when compared with residing in more

advantaged neighborhoods. Lynam et al. (2000) also noted that the direct neighborhood effects

on individual juvenile delinquency were relatively weak, whereas the neighborhood effects were

stronger when using self-reported measures of the quality of the neighborhood obtained directly

from the inner-city youth.

In a more recent study, Piquero, Moffitt, and Lawton (2005) investigated the possible

interaction effects of individual, familial, and neighborhood-level factors on life-course-

persistent offending. Their data were derived from the National Collaborative Perinatal Project

(Baltimore site) and consisted of information from the original NCPP study that included early

childhood and census-tract measures along with additionally including follow-up interviews with

the children in the 1990s. Piquero et al. (2005) failed to find direct effects for either

neighborhood or individual risk for life-course-persistent offending; however, they did find that

individual, family, and neighborhood-level effects predicted life-course-persistent offending

patterns among African-Americans in disadvantaged neighborhoods. In addition, while Piquero

et al. (2005:230) concluded that their analysis demonstrated similarities across race in life-

course-persistent patterns of offending, there were marked differences in the neighborhood

contexts where the Whites and African-Americans lived.









high-rate chronic offenders (G3) also had significantly lower household SES and less stakes in

conformity than both G1 and G2. Furthermore, G1 was comprised of significantly more

individuals that reported coming from a two natural parent household when compared with G2

and G3. Overall, similar to the 1942 ANOVA results, G3 had significantly more police contacts

than G1 and G2, and G2 had significantly more police contacts than G1. It is now important to

investigate if these same risk/protective factors that were shown to significantly distinguish

group membership in these bivariate analyses are also significant for predicting group

membership when estimated simultaneously, which now becomes the focus of the next section.

Multivariate Analysis

The following set of multinomial regression analyses are estimated in order to compare the

ability of the macro-level and micro-level factors to differentiate among the offender groups.

Each cohort-specific analysis (beginning with the 1942 cohort) will be divided into the following

three separate stages: (1) macro-level level; (2) micro-level level; and (3) the integrated model.

The first two regression analyses will assist in isolating the relative strength of prediction for the

different levels of analysis (macro-level and micro-level) and the latter will allow for an

assessment of the proposed integrated theoretical model for differentiating the non-offenders

from the adolescent-limited/peaked and high-rate chronic offenders.4

According to the results from the first macro-level-only model, using data from the 1942

Racine cohort (see Table 5-22) there was support for the effect of social disorganization in

differentiating the offender groups. Specifically, residing in a socially disorganized area nearly

doubled the odds of an individual being assigned to an adolescent-limited/peaked group and





4 Models were estimated using clustered robust standard errors in Stata 9.2.









With attention to the relative arbitrariness of how prior research has defined offender

groups, more recent research applying Nagin and Land's (1993) group-based modeling technique

has begun to identify latent classes of offenders that exhibit distinct pathways of offending (for

review see Piquero, 2007; see also Nagin, 2005). The consistent finding across these studies has

been the identification of three main types of offending groups: (1) non-offenders; (2)

adolescent-limited offenders; and (3) chronic or high-rate offenders. D'Unger, Land, McCall,

and Nagin (1988) were also able to discover similar offending types across a variety of

longitudinal data sources. It is important to emphasize here that the data relied on in this study is

official data not self-report; therefore it is likely that the non-offender group will be much larger

than it would be if the trajectories were estimated using self-report data instead.

Additionally, D'Unger, Land, and McCall (2002) found evidence of relatively comparable

latent offending classes using data from the Second Philadelphia Birth Cohort (Tracy, Wolfgang,

& Figlio, 1990); however, the adolescent-limited group they uncovered was characterized as

adolescent-peaked because of their elevated levels of offending in adolescence. This adolescent-

peaked grouped has also been noted by other researchers (for review, see Piquero, 2007). Yet,

other groups have also emerged such as low-level chronic or late-onsetters (see D'Unger et al.,

1998, 2002).

Prior applications of this methodology have also found similar offender groups in male and

female trajectories (Cote et al., 2001, 2002; D'Unger et al., 2002; Fergusson & Horwood, 2002;

Land, 2000; Piquero, Brame, & Moffitt, 2005) and other research has identified factors other

than sex such as low socioeconomic status (McDermott & Nagin, 1998) and peer and school

influences (Chung et al., 2001) as being strongly related to the likelihood of being classified as a

chronic offender. Land (2000) has also shown that sex was a significant predictor of chronic









Farrington, D.P. (1986b). Stepping stones to adult criminal careers. In D. Olweus, J. Block, & M.
Radke-Yarrow (eds.), Development of antisocial andprosocial behavior. New York:
Academic Press.

Farrington, D.P. (1995). The development of offending and antisocial behaviour from childhood:
Key findings from the Cambridge Study in Delinquent Development. Journal of
Child Psychology and Psychiatry, 360, 929-964.

Farrington, D.P. (1999). A criminological research agenda for the next millennium. International
Journal of Offender Therapy and Comparative Criminology, 43, 154-167.

Farrington, D.P. (2003). Developmental and life-course criminology: Key theoretical and
empirical issues-The 2002 Sutherland Address Award. Criminology, 41, 221-255.

Farrington, D.P., & West, D.J. (1990). "The Cambridge Study in delinquent development: A
prospective longitudinal survey of 411 males," in H. Kerner & G. Kaiser (eds.),
Criminality: Personality, behavior, and life history. Berlin: Springer-Verlag.

Farrington, D.P., Loeber, R., Elliott, D.S., Hawkins, J.D., Kandel, D.B., Klein, M.W., McCord,
J., Rowe, D.C., & Tremblay, R.E. (1990). Advancing knowledge about the onset of
delinquency and crime. In B.B. Lahey & A.E. Kazdin (eds.), Advances in clinical child
psychology. New York: Plenum Press.

Farrington, D.P., Sampson, R.J., & Wikstrom, P.H. (1994). Integrating individual and ecological
aspects on crime. Stockholm, Sweden: National Council for Crime Prevention.

Felson, R.B., & Haynie, D.L. (2002). Pubertal developmental, social factors, and delinquency
among adolescent boys. Criminology, 40, 967-988.

Fergusson, D.M., Horwood, L.J., Caspi, A., Moffitt, T.E., & Silva, P.E. (1996). The (artefactual)
transmission of reading disability: Psychometric lessons in the study of stability and
change in behavioral development. Developmental Psychology, 32, 132-140.

Fergusson, D.M., Horwood, L.J., & Nagin, D.S. (2000). Offending trajectories in a New Zealand
birth cohort. Criminology, 38, 525-551.

Fergusson, D.M., & Horwood, L.J. (2002). Male and female offending trajectories. Development
and Psychopathology, 14, 159-177.

Gephart, M.A. (1997). Neighborhoods and communities as contexts for development. In J.
Brooks-Gunn, G. Duncan, & J.L. Aber (eds.), Neighborhoodpoverty: Context and
consequences for children. New York: Sage Foundation.

Glueck, S., & Glueck, E. (1950). Unraveling juvenile delinquency. Cambridge, MA: Harvard
University Press.

Glueck, S., & Glueck, E. (1968). Delinquents and nondelinquents in perspective. Cambridge,
MA: Harvard University Press.









Independent Variables

Macro-level independent variables

Social Disorganization. Based on the factors highlighted by Shaw and McKay (1942) that

are believed to contribute to social disorganization and associated with crime (as defined as the

breakdown of the formal and informal social networks for regulating neighborhood resident

behavior), measures of neighborhood socioeconomic status (SES) and residential mobility were

included.3 Neighborhood SES was measured directly by the average housing values of the

residential properties on each block. Prior research has shown the positive effects of living in a

high SES neighborhood has on academic achievement, verbal ability scores, and behavioral

problems (see Klebanov et al., 1997). Residential mobility was indirectly measured by the

percent of housing units on a block that were renter-occupied compared with those that were

owner-occupied.4 Simcha-Fagan and Schwartz (1986:671) have argued that structural

characteristics can affect an individual's ability to form social bonds to the community.

Individuals who rent as opposed to owning their property are less likely to develop the necessary

ties (see Bellair, 1997; Granovetter, 1973) that are important for providing a mechanism of

informal social control. Finally, percent overcrowding was Census-based measure indicating

the percent of rooms within housing units on each block with more than one person per room.

Wilson (1987) has also attributed housing density to concentrated poverty, which has been


3 Because of the sparse amount of U.S. Census data available in 1950 and relative homogeneity of Racine, this study
was unable to include a measure for ethnic heterogeneity in order to provide a complete measure of social
disorganization incorporating all of the processes that lead to crime.
4 The original measure of "social disorganization" used by Land (2000) was a scale created by Shannon (1988,
1991, 1994) where a rank of 1-26 was assigned to an individual to represent their "neighborhood SES" based on the
average Census block characteristics of the area in which they lived. This measure was actually referred to by
Shannon as a "natural area code" with no reference to social disorganization and given that the measures used
(average dollar value of rent, average dollar value of owned property, percent lacking plumbing, percent
overcrowding, and percent renter-occupied) did not load on one construct it is likely that Shannon's (1988, 1991,
1994) "index" and Land's (2000) later use of this index is not an entirely accurate proxy measure of social
disorganization.









method. They also felt that the majority of their criticisms should also be heeded by any future

criminological research incorporating longitudinal data. First, using random samples of varying

sizes, they found that the number of groups only began to stabilize with samples of over 200.

Other research has indicated that the number of identified groups is partly a function of sample

size (see Nagin & Tremblay, 2001) and that the number of groups tends to be consistent with

samples above this range (see D'Unger et al., 1998).

In their final remarks, Sampson et al. (2004:41) indicate that the larger problem with the

use of trajectory analysis and inferences made from longitudinal data with short follow-up

periods concerns the causal theory for using groups. They state that future research should resort

to grounding all methods in strong theory and be wary of models that provide evidence of an

existence of groups just because they are statistically discovered or what is referred to as the

reification of groups (see also Laub & Sampson, 1993; Sampson & Laub, 1993).

As if this above exchange was not enough to settle the debate and prompt future

longitudinal researchers to be mindful to theory and consider the estimation of groups in

trajectory analysis as approximate, this dialogue was continued most recently in a 2005 issue in

the journal of Criminology. Nagin and Tremblay (2005a:873) set forth with an attempt to

debunk three commonly held misconceptions regarding trajectory analysis that originate from

researchers' failure to understand that the method results in an approximation of groups

including: (1) the individuals actually belong to a trajectory group; (2) the number of trajectory

groups is immutable; and (3) the trajectories of group members follow the group-level trajectory

in lock step.

Regarding the first misconception, Nagin and Tremblay (2005a:882) state that trajectory

groups are referred to as is they are real/true entities only for the purpose of "literary











t 1.5

o
0.5

*0
6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

Age

-- G1 pred -i- G2 pred G3 pred G1 avg G2 avg -- G3 avg

Figure 5-3. 1942 Racine Continuous Interviewed Residents n i/h Block Data Group-Based Trajectories (N=235)









average, these chronic offenders represented about 8% of the inmates but self-reported

committing over 60 crimes per year.

Chaiken and Chaiken (1982) conducted the third Rand Inmate study investigating high-rate

offending, wherein they analyzed the criminal histories and survey results from 2,190 prison and

jail inmates in California, Texas, and Michigan. Chaiken and Chaiken identified what they

referred to as a small group of "violent predators" or "omni felons" and these individuals

reported committing robberies, assaults, and drug deals at disproportionately higher rates than

the other inmates. In addition, they tended to commit burglaries, thefts, and other property

crimes at higher rates than even the offenders who self-reported to specialize in those specific

crime types.

Overall, all of the studies described above have shown strong empirical support for

"chronic offenders" and these offenders represent only 2% to 14% of the offending population

(depending on which study is being referred to), yet they are responsible for at least half of all

the offenses committed (Farrington, 1999). The fact that a sizeable fraction of the crime problem

can be attributed to a relatively small number of identifiable chronic offenders provides support

for preventative measures and early identification (see Piquero, Farrington, & Blumstein, 2003).

However, in contrast to the robust finding of chronic offending among various samples and

time periods and the evidence that these individuals tend to be criminally active for long periods

of time, Sampson and Laub (1993) argue that key life transitions such as marriage and/or

employment can have an impact on an individual's pathway of offending, even if they are/were

chronic offenders. The following section addresses the empirical evidence in support of the

effect that these various life events can have on behavioral outcomes, most notably desistance

from crime.









(1989) also identified additional factors including family disruption, urbanization, and rapid

population change that can all impede the ability of a community to create informal networks of

social control for supervising teenage peer groups. Unfortunately, there were not any data

available to measure these important processes, so caution should be taken when interpreting

these results.

Similarly, the stakes in conformity index used to assess Sampson and Laub's (1993) age-

graded theory of informal social control was limited since it relied on a dichotomous index

ranging from 0 to 2. Laub et al. (1998) have argued that the quality or cohesiveness of the adult

social bonds is what is more important than mere participation in these life transitions. Others,

however, have argued that it is likely that being married and/or having a full-time job are

probably indicators that some level of attachment has occurred in that it seems doubtful that

these events would have happened at all if this was not the case (see Piquero et al., 2002).

It is also important to note that this research relied on official data of delinquency/crime as

measured through police contacts. Despite the known strengths and weaknesses of official and

self-report data when investigating criminal behavior (see Junger-Tas & Marshall, 1999), police

contacts may be the better measure of "true" officially-based offending because they are

typically closer in time to the actual criminal event than self-reported offending (see Hindelang

et al., 1981; Thornberry & Krohn, 2003).

Finally, there are some limitations of the trajectory modeling procedure as it specifically

pertains to this study. There was evidence provided from the sensitivity analysis conducted that

data loss did not equally affect all of the trajectory groups. For instance, it appeared that once

the multiple sources of data were merged together that the model only discovered three groups

and that the high-rate offending group (also referred to as chronic) actually showed evidence of









police contacts of this group ranged from a low of 22 to a high of 63 police contacts. This small

group of offenders managed to accumulate 812 police contacts during the ages of 6 to 25!

The mean posterior probabilities for group assignment indicated that, like the trajectory

analyses using the 1942 Racine continuous resident data, this model demonstrated the ability to

determine group membership with a high degree of probability (see Table 5-12). G1 had the

lowest mean probability (M=.84; median=.93), yet this was still way above the standard cutoff of

.70 (Nagin, 2005). The mean posterior probabilities were remarkably high for the low (G2) and

high-rate (G5) chronic offending groups indicating that these individuals' trajectories faced little

risk for mis-classification. According to the coefficients and their corresponding p-values

produced for this sex covariate, there was evidence that for all of the offender groups except the

late-onset offenders (i.e., adolescent-limited/peaked, low and high-rate chronic offenders) being

male significantly increased the likelihood of group membership into an offender group

compared with a non-offender group (see Table 5-13).

Table 5-14 presents the BIC model selection criteria for the 1949 continuous interviewed

residents and similar to the interviewed 1942 sample only three trajectory groups were produced.

This analysis failed to detect a late-onset or a low-rate chronic group as was the case in the prior

1949 continuous resident analysis. The average and predicted curves, as plotted in Figure 5-5,

indicated that the following three trajectory groups could be identified: (1) non-offenders; (2)

adolescent-peaked offenders; and (3) high-rate chronic offenders. G1 (75.1%) represented the

overwhelming majority of the sample and exhibiting practically no offending across the twenty

year time period. Comparatively, the second group (20.5%) did not begin offending until early

adolescence before reaching a peak in mid-adolescence and then desists by their early 20s.

Individuals in this group recorded an average of 4.32 (SD=2.81) police contacts with a range of 2









received the home visits faired significantly better than those that received the minimal amount

of services and had higher levels of employment and fewer cases of child abuse and neglect. In

an outcome evaluation of a similar program, Seitz et al. (1985) also found that pre and post-natal

home visits to disadvantaged mothers reduced the impact of adverse risk factors and the children

whose mothers participated in the program were rated as less aggressive and disruptive than

children who were in the non-experimental group (see Provence and Naylor, 1983 for an

extended discussion of the Yale Child Welfare Research Program).

Programs targeting poor urban children during the preschool years have shown to have a

significant impact on reducing delinquent outcomes. Head Start is one of the most well-known

of these type of initiatives and subsequent evaluations (see Darlington et al., 1980 ; Lee et al.,

1990) have shown that the program was successful in producing short-term intellectual and

social gains during the early preschool years and that Head Start youth were more likely to

graduate high school. Other programs such as the Perry Preschool Project, which focused on

early childhood education (ages 3 to 5), parent meetings, and home visits, have yielded a host of

beneficial outcomes for program participants including better grades, higher standardized test

scores, higher graduation and employment rates, and lower delinquency and arrest rates (see

Berreuta-Clement et al., 1984; Schweinhart & Weikart, 1988). In addition, programs

implemented in the later elementary/early adolescent years specifically directed at improving

problem-solving and social competency among poor youth have been associated with decreased

levels of aggression and problem behaviors (see Weissberg et al., 1981; Hawkins et al., 1991;

Lochman et al., 1984; Sarason & Sarason, 1981; Hammond, 1991).

Family intervention programs facilitating the development of parental support networks

and increasing the link between parents and community services as well as improving parent-









identified in the 1949 continuous resident analysis (N=1,297) where although there were more

males in the group, sex was not found to be a significant covariate for distinguishing this

offender group from the non-offender group. The mlogit results also re-affirmed the importance

of sex for predicting group membership through the consistent and robust effect of being male on

increasing the odds by at least two times for an individual exhibiting either an adolescent-

limited/peaked or a chronic offending trajectory.

The results only revealed limited support for Sampson and Laub's age-graded theory of

informal social control and these findings were only evident at the bivariate level. In both cohort

analyses, the non-offender group had the highest mean score on the stakes in conformity index

when compared with either offender group, but only in the 1949 cohort analysis was the stakes in

conformity index significantly associated with group membership.

In line with the expectation of the influence of social disorganization, there was a minimal

amount of evidence that living in a socially disorganized area increased an individual's odds for

being in an offender group. The bivariate findings suggested that Racine continuous residents,

whose individual offending pattern could be classified into a adolescent-limited/peaked or

chronic offending group-based trajectory, tended to live in more socially disorganized areas than

the non-offenders. The multivariate results confirmed that social disorganization was a

significant predictor of group membership for both the 1942 and 1949 cohort. Although, living

in a socially disorganized area failed to distinguish the adolescent-limited/peaked offenders from

the chronic offenders.

Despite the fact that the bivariate and mlogit (macro-level-only) results indicated that

social disorganization was a distinguishing feature of group membership, the findings from the

mlogits, where both the macro-level and micro-level factors were estimated simultaneously,









The ZIP model allows for the assumption of a dual-state-system. A dual-state-system

signifies two situations: one in a zero-contact state where the assumption is that a group of

individuals exists that can be regarded as police-contact free, and the second state or non-zero

contact state where the expected incidences of police contacts are assumed to follow some

known distribution, such as the Poisson (Long, 1997; Lambert, 1992). Several other studies have

also noted the practicality of the ZIP model when using count data with excess zeros (King,

1989; Land, McCall, & Nagin, 1996; Zorn, 1998).

The ZIP model takes the following form where the first term is the probability of no police

contacts and the second term is the probability of the combined event's being active and the

probability of zero police contacts given Xt > 0 (Nagin, 2005:35).

j y't
S(y,)= (1- a, ) (t e t /y) (t= 1, 2,....)

Once the parametric form of the trajectory model is established it is necessary to determine

the functional form of the groups' trajectories and there are three possibilities including a linear,

quadratic, and cubic form (or age, age-squared, and age-cubed). Depending on the ages and the

length of the observation period, it is unlikely that individual offending will have a linear form.

Similarly, although the quadratic model is more often used in research (see Bushway et al.,

2003:140), it only permits one major directional change in the trajectory, and it is likely that

cohort members will and have undergone periods of intermittency in their involvement with the

police. Intermittency periods refer to those points in time (i.e., years) where no official police

contacts were recorded for any particular cohort members. Therefore, it is likely that the cubic

model will demonstrate the best functional form of the trajectories. The CNORM and the ZIP

equations that allow for all three of these possible functional forms are presented below:

yt* = po' + P0Age,1t + P02Aget,2 + P3Aget3 + t, (CNORM),









LIST OF FIGURES


Figure page

3-1 Integrated Theoretical Model: Social Disorganization, An Age-Graded Theory of
Informal Social Control and Moffitt's Developmental Taxonomy .................................69

5-1 1942 Racine Continuous Residents Group-Based Trajectories.................................138

5-2 1942 Racine Continuous Interviewed Residents Group-Based Trajectories.................139

5-3 1942 Racine Continuous Interviewed Residents i, il/ Block Data Group-Based
T raje cto rie s ........................................................................... 14 0

5-4 1949 Racine Continuous Residents Group-Based Trajectories ................................141

5-5 1949 Racine Continuous Interviewed Residents Group-Based Trajectories.................142

5-6 1942 Racine Continuous Interviewed Residents i, il/ Block Data Group-Based
T raje cto rie s ........................................................................... 14 3









Wikstrom, P.H., & Loeber, R. (2000). Do disadvantaged neighborhoods cause well-adjusted
children to become adolescent delinquents? A study of male juvenile serious offending,
individual risk and protective factors, and neighborhood context. Criminology, 38, 1109-
1142.

Wilson, J.Q., & Kelling, G.L. (March, 1982). Broken windows: The police and neighborhood
safety. Atlantic Monthly, 29-38.

Wilson, J.Q., & Hernstein, R. (1985). Crime and human nature. New York: Simon & Schuster.

Wilson, W.J. (1987). The truly disadvantaged: The inner city, the underclass, and public policy.
Chicago: University of Chicago Press.

Wirth, L. (1938). Urbanism as a way of life. American Journal of Sociology, 44, 1-24.

Wolfgang, M.E., Figlio, R.M., & Sellin, T. (1972). Delinquency in a birth cohort. Chicago:
University of Chicago Press.

Wolfgang, M., Thomberry, T.P., & Figlio, R.M. (1987). From boy to man, from delinquency to
crime. Chicago: University of Chicago Press.

Zorn, C.J. (1998). An analytic and empirical assessment of zero-inflated and hurdle Poisson
specifications. Sociological Methods and Research, 26, 368-400.









administered in adulthood this would provide worthwhile information in order to gauge the level

of cohesiveness or attachment in the adult social bonds to better determine their effect in the

desistence process.

Nevetheless, it is obvious that collecting research in this magnitude, for a lengthy period of

time, involving a numerous amount of actors would require an enormous amount of monetary,

individual, political, and community support, yet there are early indications that research similar

to this magnitude (in a lesser degree) is possible and the preliminary findings leave hope for

future research in integrative mutli-level modeling (see related publications from the Project on

Human Development in Chicago Neighborhoods).

Implications

Taken together, the findings from this research are intended to be preminary and to offer a

starting point to guide and direct future research on theoretically integrated approaches. Given

the minimal significance of social disorganization in the baseline models, it is important that

future studies further investigate other social processes that have been recently discussed by

neighborhood-level researchers as perhaps being more important when estimating neighborhood

effects (for review, see Sampson et al., 2002) before any concerted effort can be made toward

imforming policy. Nevertheless, it is important here to at least highlight some of the possible

implications that research in this area may provide in the future.

Programs such as the Rochester Nurse-Home Visitation Program (see Olds et al., 1986) are

likely to provide the most benefit in targeting the cumulative effects of both neighborhood

disadvantage and factors particularly associated with chronic offending (i.e. low birth weight,

birth trauma). The Rochester program involved random assignment of expectant mothers in a

small semi-rural community to one of three conditions: (1) pre and post natal visits; (2) pre-natal

only visits; and (3) minimal standard services. The results indicated that the mothers who









LIST OF TABLES


Table page

4-1 Resident Samples by Cohort, Size, Data Sources, and Offender Groups........................102

4-2 Observational Periods of Data Sources by Cohort ..................................... ..................102

4-3 Social Disorganization Index Factor Loadings by Cohort................................... 102

5-1 Descriptive Statistics for the Macro-level and Micro-level Measures by Cohort ..........126

5-2 Tabulated Bayesian Information Criteria (BIC) for 1942 Racine Continuous
R e sid en ts ...................................... .................................................... 12 6

5-3 Mean (Median) Posterior Probabilities for Group Assignments: 1942 Racine
C continuous R residents ........................................... ...... ...................... 127

5-4 Coefficient Estimates for Effect of Sex (being Male) on Trajectory Group
M membership: 1942 Racine Continuous Residents..........................................................127

5-5 Tabulated Bayesian Information Criteria (BIC) for 1942 Racine Continuous
Interview ed R residents .................. .............................. ............... ..... 127

5-6 Mean (Median) Posterior Probabilities for Group Assignments: 1942 Racine
Continuous Interviewed Residents ........................................................ ............... 128

5-7 Coefficient Estimates for Effect of Sex (being Male) on Trajectory Group
Membership: 1942 Racine Continuous Interviewed Residents ............ ... ................. 128

5-8 Tabulated Bayesian Information Criteria (BIC) for 1942 Racine Continuous
Interviewed Residents i/ll Block Data.................................. .... ................. 128

5-9 Mean (Median) Posterior Probabilities for Group Assignments: 1942 Racine
Continuous Interviewed Residents i/ll Block Data.................................................. 128

5-10 Coefficient Estimates for Effect of Sex (being Male) on Trajectory Group
Membership: 1942 Racine Continuous Interviewed Residents i, i/ll Block Data ............129

5-11 Tabulated Bayesian Information Criteria (BIC) for 1949 Racine Continuous
R e sid en ts ...................................... .................................................... 12 9

5-12 Mean (Median) Posterior Probabilities for Group Assignments: 1949 Racine
C continuous R residents ........................................... ...... ...................... 129

5-13 Coefficient Estimates for Effect of Sex (being Male) on Trajectory Group
Membership: 1949 Racine Continuous Residents........... .................. ......... ..... 130









individual's pattern of offending by disrupting the ability to form/maintain prosocial bonds with

peers and their schools, which are important factors for inhibiting delinquency (see Hirschi,

1969; Akers, 1985).12

Dependent Variables

Delinquent/Criminal Involvement. The variable used to measure offending was the

number of official police contacts for each individual from age 6 to age 25.13 The police contact

measure was limited to those instances in which an individual came into contact with a Racine

police officer specifically for a crime that he/she was suspected to have committed or did in fact

commit. In addition, the official police contacts were also restricted to contacts that had true

crime codes with a felony/misdemeanor distinction as well as contacts for status offenses to

measure offending (i.e., excluding traffic offenses, and city ordinance violations) because (1)

they are a more accurate official measure of "true" offending when compared with arrests and

convictions; (2) they are closer in time to the actual criminal act than self-reported offending (see

Hindelang et al., 1981; Thornberry & Krohn, 2003); and (3) the inclusion of status offenses

increases the potential for identifying adolescent-limited offenders (see Moffitt, 1993). 14

Offending Group Membership. The results of the semiparametric mixed Poisson-based

regression models will guide the identification of the number of offending groups that exist

within these data and the shapes of their offending trajectories. The identified latent classes of

offenders will operate as the dependent variable for the estimation of the multinomial

12 It is important to note that this measure is limited in that is was impossible to discern from the available data if the
individual moved from a social disorganized area to a more or less disorganized area; therefore, it is important to
interpret the findings with caution and keep in mind that the expectation is that moving is seen as having an adverse
effect on offending patterns within this study.

13 This police contact data refers to the years that are consistently available for the 1942 and 1949 continuous-
residents or up to and including age 25.
14 Prior trajectory analyses with Shannon's (1988, 1991) Racine cohorts relied solely on felonies and misdemeanors
(D'Unger et al., 1998) or everything including traffic offenses and "suspicion" contacts (Land, 2000).









sort of common interest in the preservation of the inhabited area; and (3) a collective identity that

tends to be constant over time (Bursik & Grasmick, 1993:6).

Bursik and Grasmick (1993) further argued that social control may exist within

communities in different forms or what they identify as private, parochial, and/or public

networks. However, they did contend that all communities tend to operate with the same

underlying desire for informal and formal associations for eliciting social control in order to

minimize crime (bearing in mind that particular crimes are likely to be differentially viewed as

unconventional as well as differentially enforced across communities). Although, the

"community/neighborhood" debate may be unresolved, but there appears to be a movement

toward fleshing out the strengths of its definitional attributes and providing a more encompassing

characterization.

This concept of community and the process of identifying particular factors that are

associated or lead to crime were the basis for what has emerged as of one of the classic theories

of crime, or what is commonly known in the discipline as social disorganization theory. This

next section presents the historical roots of the theory and lays out the core assumptions behind

why crime should be higher in socially disorganized communities.

Social Disorganization

Early macro-level research investigating the importance of ecological characteristics for

explaining crime rates has its origins from the Chicago School and the work of Park and Burgess

(1925) and Shaw and McKay (1942). The work of the latter scholars focused on attempting to

offer a conceptual framework to explain the stability of aggregate crime/delinquency rates over

time in the city of Chicago despite the dynamic change in the racial and ethnic composition of its

neighborhoods (Shaw & McKay, 1942). Shaw and McKay argued that the majority of the crime

tended to be concentrated in pockets of the city that Park and Burgess (1925) characterized as the









rates of offending over the entire observation period. There was a tendency for these individuals

to peak at age 16 and then decline, but they still showed non-zero rates of offending through age

25. On average, individuals in this group had 1.98 (SD=1.24) police contacts and no individual

in this group had more than 7 police contacts. G3, representing 2.6% of the sample, is an

atypical group that is not identified in any of the other analyses presented in this study although

this offender group has been the point of discussion in prior research (see D'Unger et al., 1998;

Sampson and Laub, 1993). Individuals in this group are unique in that they had virtually no

officially recorded police contacts until they reached adulthood (age 18) and their rates of

offending increased rather noticeably in young adulthood. No individual in this group had less

than 3 police contacts (M=6.53; SD=3.89) and a handful of these offenders had actually

accumulated more than 10 police contacts despite the fact that their offending did not begin until

young adulthood.

The fourth group, representing 5.7% of the sample, can be best described as an adolescent-

peaked group (see D'Unger et al., 1998; Land, 2000) in that these individuals do not begin

offending until early adolescence but offend at relatively high rates during mid-adolescence

before peaking at age 15 and then virtually have desisted by age 23. No individual in this group

had less than 6 police contacts and more than half of the individuals in this group had

accumulated more than 10 police contacts (the majority of which were during mid-adolescence).

G5, representing the smallest percentage of the sample (1.9%), exhibited a high-rate chronic

pattern of offending that begins early on in life (prior to age 8) and their offending drastically

increases well into adolescence before reaching a peak in late adolescence. These individuals

still display high levels of offending throughout the remainder of the observation period or age

25. Individuals in this group averaged an alarming 32.48 (SD=10.30) police contacts and the









Cohen, L.E., & Felson, M. (1979). Social changes and crime rate trends: A routine activities
approach. American Sociological Review, 44, 588-608.

Cohen, L.E., & Land, K.C. (1987). Age structure and crime: Symmetry versus asymmetry and
the projection of crime rates through the 1990s. American Sociological Review, 52, 170-
183.

Cohen, L.E., & Machalek, R. (1988). A general theory of expropriative crime. American Journal
ofSociology, 94, 465-501.

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.

Cote, S., Zoccolillo, M., Tremblay, R.E., Nagin, D., & Vitaro, F. (2001). Predicting girls'
conduct disorder in adolescence from childhood trajectories of disruptive behaviors.
Journal of the American Academy of Child and Adolescent Psychiatry, 40, 678-684.

Cote, S., Tremblay, R.E., Nagin, D., Zoccolillo, M., Vitaro, F. (2002). The development of
impulsivity, fearfulness, and helpfulness during childhood: Patterns of consistency and
change in the trajectories of boys and girls. Journal of ChildPsychology & Psychiatry &
Allied Disciplines, 43, 609-618.

Cramer, H. (1946). Mathematical models of statistics. Princeton, NJ: Princeton University Press.

Crane, J. (1991). The epidemic theory of ghettos and neighborhood effects on dropping out and
teenage childbearing. American Journal of Sociology, 96, 1126-1159.

Daly, K. (1998). Gender, crime, and criminology. In M. Tonry (ed.), The handbook of crime and
punishment. New York: Oxford University Press.

Darlington, R.B., Royce, J.M., Snipper, A.S., Murray, H.W., & Lazar, I. (1980). Preschool
programs and later school competence of children from low-income families. Science,
208, 202-204.

De Li, S. (1999). Legal sanctions and youths' status achievement: A longitudinal study. Justice
Quarterly, 16, 377-401.

Donnellan, M.B., Ge, X., & Wenk, E. (2000). Cognitive abilities in adolescent-limited and life-
course-persistent criminal offenders. Journal ofAbnormal Psychology, 109, 396-402.

Dumas, J.E. (1989). Treating antisocial behavior in children: Child and family approaches.
Clinical Psychology Review, 9, 197-222.

Duncan, G.J. (1994). Families and neighbors as sources of disadvantage in the schooling
decisions of White and Black adolescents. American Journal ofEducation, 103, 20-53.









Chapter 6 will provide a further discussion of the results from the series of analytical

procedures and elaborate on the possible implications of the research findings. The chapter will

also highlight the limitations of the current study, offer suggestions as to how future research can

improve upon and extend the examination of integrated theoretical models, and close with a few

final thoughts.









Hirschi and Gottfredson (1983:54), in response to what they considered to be the

misapplication of the age-crime distribution by scholars in criticizing social theories of crime and

supporting longitudinal methods over cross-sectional designs, assert that the effect of age is

invariant across time, place, social and cultural conditions, and crime type. They indicated that

rates of delinquency have been comparable over the last 150 years and that the age distribution

of delinquency in England and Wales in the 1940s, Argentina in the 1960s, and the United States

in the 1980s are virtually indistinguishable. They further argued that although non-White

offenders may start their delinquency earlier and offend at a higher rate than Whites and that

while boys offend at higher rates than girls for particular offenses, the age-crime curve is still

invariant across race and sex (1983:556). In addition, they provided evidence against the

differences in the age-distributions of property and person offenses. They illustrated that while

official data may reveal different age-crime curves for person and property offenses, this

difference is not observed in self-report data; therefore, the effects of age are indeed invariant

across crime type.

Hirschi and Gottfredson's (1983) early claims about the invariance of the effects of age

across historical and cultural conditions and crime type were soon met by criticism. Farrington

(1986a) specified that the main limitation of interpreting the types of age-crime curves used by

Hirschi and Gottfredson (1983) is that there is probable ambiguity in differentiating between age

and period effects. Period effects are significant factors that are time-period specific, such as

eras of high unemployment, riots, and/or civil unrest, and these influences are only related to

changes in the crime rates of those cohorts that are aligned with that particular era in history. In

contrast, age effects are universal changes that occur with age (i.e., natural maturation) that lead

to physical deterioration for all cohorts during all periods of observation.









environment of the home. They found that the positive effects of living in a high SES

neighborhood on academic achievement, higher verbal ability scores, and less behavior problems

and the negative effects of living in heterogenous neighborhoods on verbal ability scores were

mediated by the quality of the home learning environment. The effect of home environment was

found across all age groups, measures, and resource-rich and resource-poor neighborhoods.

They also found support for the interaction of the home environment and family characteristics

in each neighborhood type in ways related to criminal outcomes. Similarly, additional research

by Brooks-Gunn et al. (1993) showed that neighborhoods had an effect on childhood IQ and the

likelihood of becoming a teenage parent and/or dropping out of school. Thus, it appears that

neighborhood level effects may be both direct and indirect, but perhaps a substantial proportion

of their effects are indirect in that they operate through individual and family factors.

Simcha-Fagan and Schwartz (1986:671) argue that structural characteristics have an

impact on the organizational networks among members of the community, and these networks

can subsequently affect an individual's ability to form social bonds to the community.

Individuals who rent property as opposed to owning their own property are less likely to have a

stake in the community in which they live, and thus may be less inclined to participate in

neighborhood organizations that provide a mechanism of informal social control. Similarly,

individuals who rent their property may be less likely to care about the maintenance or up-keep

of their dwelling because they do not intend to be there long. This lack of investment could lead

to neighborhoods incivilities (Wilson & Kelling, 1982), which have been linked a breakdown in

informal social control and crime (Sampson, 1995). Dwelling units that are over-crowded such

as public housing units are also likely to lead to a sense of "anonymity" among the residents.

The failure of individuals to get out of their units and develop informal social ties with their









presents both sides of the debate regarding the persistent-heterogeneity versus state-dependent

explanations of criminal behavior.

Developmental Criminology/Life-Course Theory

The life-course perspective can be defined as a series of pathways throughout the

developmental process where there are varying age-dependent expectations and available options

in an individual's decision-making process. These decisions along with the natural course of life

events are seen to shape the trajectory of an individual's life and can be influenced by transitions

and turning points (Elder, 1985). Trajectories are viewed as pathways of development over the

life span such as work life, marriage, schooling, or involvement in crime. They are seen as long-

term behavioral patterns that are marked by a sequence of life events and transitions. These

transitions are then described as distinct life events that occur in shorter intervals that may alter

an individual's behavioral trajectory (such as first job, first marriage, first child, etc.) (Elder,

1985:31-32).

Sampson and Laub (1990) extended this framework by focusing on the strong association

between childhood events and adulthood experiences, and the idea that transitions can redirect an

individual's life-course behavioral pathway. They indicated that life-course analysis focus on

the length, timing, and the order of significant life events (including crime) and their effect on

future social development (Sampson & Laub, 1992). Attention to the impact of life events is

critical because prior research has shown that committing a crime has a considerable behavioral

influence on the likelihood of an individual committing crime in the future (Nagin & Paternoster,

1991).

Loeber (1982) indicated that children who initially exhibited high rates of antisocial

behavior were more likely to continue this behavior than those who initially showed slower rates

of antisocial behavior. Relying on data from 411 South London males, Farrington (1995:956),









Sampson and Laub (1993:317-320) concluded by offering six key issues to guide future

research testing their theory. First, there is some element of luck, randomness, or chance in an

individual's life course (also referred to by Rutter, 1989 as adventitiouss happenings"). Second,

there is individual variation in the effects that life transitions have on behavior. Third, macro-

level opportunity structures for marriage and labor market activity are important (i.e.,

segmentation of the labor market, ethnic enclaves). Fourth, criminal propensity cannot

completely explain adult crime because life transitions can also affect opportunities and

motivations to commit crime (see also Cohen & Machalek, 1988). Fifth, the effect of cumulative

continuity is strongest among those in disadvantaged racial and economic conditions. Sixth,

historical context matters.

With attention to some of these suggestions, Homey et al. (1995) found clear evidence that

meaningful change in life circumstances was associated with desistance from crime based on

their interviews with 658 offenders in a Nebraska Correctional Facility. They also estimated

hierarchical linear models that allowed them to control for criminal propensity and still found

that the offenders' reported reduced involvement in criminal activity when they were living with

their wife. Overall, their results suggested that life transitions can alter behavioral patterns of

offending at particular periods of time, especially in the short-term. Laub et al. (1998) estimated

trajectories with the Glueck's data and found that the timing and quality of marriages were what

was important and that the inhibiting effects of marriage on crime were gradual and grew in

intensity over time. Similar to Homey et al. (1995), these findings were still observed while

controlling for criminal propensity and a host of other factors.

More recently, Piquero et al. (2002) developed a stakes in conformity index where they

combined the life circumstances of employment and marriage to gauge the possible cumulative









group membership to a high-rate chronic offending trajectory with individuals in this group

having significantly lower mean scores on the stakes in conformity index when compared with

those in the adolescent-limited/peaked group and individuals in the non-offender group.

Qualified support was also shown for hypothesis #4 regarding the expectation that social

disorganization would be a significant distinguishing feature for group assignment. Being from a

socially disorganized area significantly increased the odds that an individual would be assigned

to either an adolescent-limited/peaked offender group or a high-rate offender group in relation to

a non-offender group in the 1942 and 1949 cohort mlogit analyses. The effect, however, was

only marginally significant (p=.07) for distinguishing high-rate chronic offenders in the 1949

analysis. The results indicated that social disorganization did not significantly distinguish the

adolescent-limited/peaked offenders from the chronic offenders.

Finally, there was little support provided for hypothesis #5 or the expectation that macro-

level and micro-level effects on the participation and frequency of offending will still be

observed net of the effects of each other. Although the results did indicate that once the multi-

level effects were considered simultaneously, the effect of social disorganization still maintained

a marginally significant effect for distinguishing the adolescent-limited/peaked offenders from

the non-offenders in theoretically integrated mlogit analysis of the 1942 cohort, the majority of

the effect appeared to be indirect. While the direction of the odds were in the expected direction

for both the 1942 and 1949 mlogits (with being from a socially disorganized area increasing the

odds of assignment to an offender group), an examination of the R-squares (McFadden's and

Nagelkerke's) showed that the addition of the social disorganization index contributed relatively

little for improving model fit. Once again, social disorganization was unable to significantly

differentiate the adolescent-limited/peaked group from the chronic offending group.









McDermott and Nagin (1998) found support for the existence of three offender groups

based an analysis of male respondents from the National Youth Survey, which is a well-known

self-report survey. The first group was classified as the non-offenders. The second group was

referred to as adolescent-peaked. They tended to resemble the third or chronic offending group

in adolescence but then began to desist at the age of 18 until there was no observed offending at

age 24.

Additionally, trajectory analyses based on two distinctly different types of offenders

(Gluecks' Boston delinquents vs. California Youth Authority parolees) from different time

periods and contexts found that there seemed to be 4 to 6 trajectory groups that all began to

decline by adulthood (see Laub et al., 1998; Piquero et al., 2002). Laub et al. (1998) and Piquero

et al. (2002) also noted that marriage or stakes-in-conformity (that included marriage) was

related to lower trajectories.

A more comprehensive attempt to identify the consistency of latent offending groups was

performed by D'Unger et al. (1998) using data from the London cohort (West & Farrington,

1973, 1977), the Second Philadelphia Birth Cohort (Wolfgang, Figlio, & Sellin, 1972; Tracy et

al., 1990) and the three Racine Birth Cohorts (Shannon, 1988, 1991). They found support for the

considerable similarity of offending groups across the data sources, with the constant finding of

non-offending, adolescent-limited or peaked offending, and chronic offending groups.

Specifically concerning the Racine Birth Cohorts, D'Unger et al.'s results revealed the similar

four classes of offenders noted above for the 1942, 1949 and 1955 cohorts, but discovered an

additional fifth group for the 1942 and 1955 cohorts with a late-onset pattern of offending.

D'Unger et al. (1998) illustrated that this group showed relatively stable rates of offending

throughout their twenties. Broidy et al. (2003) also found three to four trajectory groups when









summary, Hipwell et al. (2002) concluded that the main difference between boys and girls

behavioral patterns of offending is that only a very small proportion of girls follow the life-

course-persistent trajectory.

Chung et al. (2001) also discovered another trajectory group that they termed desisters that

illustrated an atypical trajectory in that they were characterized by early-onset offending (before

age 13) but instead of showing the steady rise in mid-adolescence into young adulthood, the

group's trajectory declined and appeared to desist by age 21. According to Chung et al. (2001),

further analyses suggested that it was not individual and familial factors that could be attributed

to their desistance but rather the influence of less exposure to peer, neighborhood, and school

factors typical associated with offending. They also indicated that aggressiveness, antisocial

peers, and drug availability were consistent predictors of initial levels of offending measured at

age 13 and that individual factors including aggression, depression, and anxiety significantly

distinguished the late-onset offenders from the non-offending group.

Piquero (2007) provided a comprehensive review on the studies to date applying Nagin's

and Land's (1993) group-based modeling technique and concluded that there existed relative

consistency in the number of trajectory groups and the shape of the curves, but there was

variability across the studies depending on the length of follow-up and the ages that were

represented in the analysis. Additionally, in line with Moffitt these studies have shown evidence

of adolescent-limited and life-course-persistent trajectories across diverse data sets, cohorts,

countries and using various measures for offending including self-reports, teacher ratings, parent

ratings, and police records based on either contacts, arrests, or convictions. However, these

applications have also uncovered other trajectories not consistent with Moffitt's (1993)

predictions such as the low-level chronic and late-onset type.









Family Stability. This measure was used to represent the stability of the family dynamic

in which the cohort member was socialized. Most research has shown an established

relationship between family stability and delinquency/crime (Glueck & Glueck, 1950; McCord

& McCord, 1959; Goetting, 1994) and crime rates (Sampson, 1987). Studies have found that

adolescents living with both biological parents have lower rates of offending when compared

with those who reside in broken homes or reconstituted families. According to Gottfredson and

Hirschi (1990:104), single-parent-headed households (usually by a woman) may be unable to

provide the same amount of supervision because of the time constraints imposed on them and

thus may be more likely to have negative responses to their children's unwanted behaviors.

Furthermore, this variable along with the following measure listed below (family structure) have

been identified as key family-level variables when estimating neighborhood effects (see Brooks-

Gunn & Leventhal, 2000)

Originally, this variable was based on a scale created by Shannon (1988, 1991) which

combined responses to several items regarding family composition ranging from 1 to 20, where

according to Shannon (1988, 1991), 1 represented the most stable family dynamic (two parents

present during the entire childhood/juvenile period) and 20 the least stable family dynamic (lived

only with grandparents, never with parents). For the purposes of this study and due to the little

variability among the categories, the scale was collapsed into a dichotomous variable with 1

indicating the presence of two parents during the entire childhood/juvenile period and 0

indicating any other type of less stable family dynamic.

Family Structure. In order to account for the possible differential opportunities of

parental supervision affecting individuals' offending pathways (Gottfredson & Hirschi, 1990)

along with having a complementary individual measure of household density to the macro-level









and a chronic offender group (with a declining trajectory in young adulthood). However, the

semi-parametric Poisson based models also discovered two groups that were inconsistent with

Moffitt's prediction when relying on police contact data from the largest samples of the 1942

(n=633) and 1949 (n=1,297) Racine continuous residents. A low-level chronic offending group

was identified in both cohorts who were characterized by an early onset of offending and

maintained a fairly low level of offending (yet a non-zero rate) throughout the twenty year

observation period (age 6-25). Comparatively, a late-onset offender group was discovered

among the 1949 cohort who had virtually no invovlement in offending until late in

adolescence/early adulthood but showed relatively high rates of offending throughout young

adulthood. In addition, the adolescent-limited/peaked group found in these analyses slightly

differed from the group as theoretically discussed by Moffitt (1993) in that they were classified

as adolescent-peaked and their offending patterns did not appear to desist until around age 21.

While these groups are atypical insofar as Moffitt's expectations, they have been discovered and

the source of discussion elsewhere (see D'Unger et al., 1998; Land, 2000; Piquero, 2007;

Sampson & Laub, 1993).

Also consistent with Moffitt and others (Rutter, 1983; Singer et al., 1968; Werry, 1986)

that have argued that life-course-persistent offending may be observed almost exclusively among

males and the prior research has shown that females typically show less involvement in crime in

either type of offender group (adolescent-limited offenders and life-course-persistent offenders)

(see Piquero, 2007), all of the trajectory results indicated that being male significantly increased

an individual's likelihood of being classified into any of the offender groups across all data

sources (continuous residents, continuous interviewed residents, and continuous interviewed

residents in i/h block data) and by cohort. The one exception was the late-onset offender group










individual self-reported "peer influence" as the strongest contributing factor toward their

delinquent/criminal behavior in junior high and high school.9

Formal Social Control. This measure was a scale of the individual's perception of the

level of police patrol in their neighborhood during their junior high and high school years.10

This indicator of institutionalized formal social control ranged from 0 representing "not at all" or

"no police patrol," 1 for "lightly patrolled," 2 for "moderately patrolled," to a high of 3 which

indicated that the individual perceived their neighborhood as being "heavily patrolled" during

their early to late adolescent years.1

Moved. In order to provide a contextual measure of residential mobility to complement

the compositional measure included in the social disorganization index, this variable was coded

as 1 if the individual had moved within Racine during their juvenile years. It is likely that

moving, particularly during the adolescent years, is likely to have an adverse effect on an


9 It is worth mentioning here that this is certainly a less than ideal measure of differential association and more direct
measure would have been preferable; however, due to the limitations of data this measure was included as a control
variable.

10 It is important to note that this measure is a weak measure of formal social control and that direct measures such
as prior conviction or incarceration are optimal; however, there was only two items that allowed the individuals to
report having received jail time as a result from their police contact and these questions were only in reference to
their first and second police contact. Because only 1 to 4 individuals across the cohorts actually reported this as a
response I made the determination to use the "police patrol" measure because there was a sufficient amount of
variation in responses to this latter survey item. Additionally, this measure is also limited due to the fact that it was
asking individuals to retrospectively answer this question based on their perceptions of how they felt as a middle and
high-school juvenile (which is at least 10 years prior and for the 1949 cohort and 20 years prior to the date of the
interviews in 1974 for the 1942 cohort) and it is also likely that those individuals who actually came into contact
with the police may have had a bias toward reporting higher levels of neighborhood patrol. Future studies would
benefit from incorporating more direct measures of formal social control such as those discussed by Sampson and
Laub (1997) and De Li (1999), which relied on convictions and on court appearances and convictions, respectively.

1 Shannon (1988, 1991, 1994) originally developed a scale to assess formal social control, but this measure had a
series of flaws, with the most serious being that the scale was constructed by summing self-reported responses of the
cohort member's perceptions of the (1) level of police patrol when they were in junior high and high school
(l=heavily, 2=moderately, 3=lightly; 4=not at all); (2) their and a couple of their friends' views toward the police in
junior and senior high, positivei, 2=negative, 3=indifferent); and (3) to report whether or not they got into trouble
with the police with a couple of their friends (yes/no). Even more oddly, this scale was constructed by coding an
individual's response as "1" if they had either negative or indifferent views toward the police, adding "2 points" if
the individual reported that they and their friends got in trouble with the police, and adding "4 points" if the
individual also responded that they perceived police patrol was either heavy or moderate during their juvenile years.









Sampson, R.J. (1997). The embeddedness of child and adolescent development: A
neighborhood-level perspective on urban violence. In J. McCord (ed.), Violence and
childhood in the inner city. Cambridge, UK: University of Cambridge Press.

Sampson, R.J. (2002). Transcending tradition: New directions in community research, Chicago
style. Criminology, 40, 213-230.

Sampson, R.J., & Groves, W.B. (1989). Community structure and crime: Testing social-
disorganization theory. American Journal of Sociology, 94, 774-802.

Sampson, R.J., Raudenbush, S.W., & Earls, F. (1997). Neighborhoods and violent crime: A
multilevel study of collective efficacy. Science, 277, 918-924.

Sampson, R.J., 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., & Laub, J.H. (1992). Crime and deviance in the life course. Annual Journal of
Sociology, 18, 63-84.

Sampson, R.J., & Laub, J.H. (1993). Crime in the making: Puathii uy and turning points through
life. Cambridge, MA: Harvard University Press.

Sampson, R.J., & Laub, J.H. (1995). Understanding variability in lives through time:
Contributions of life-course criminology. Studies on Crime and Crime Prevention, 4,
143-158.

Sampson, R.J., & Laub, J.H. (1997). A life-course theory of cumulative disadvantage and the
stability of delinquency. In T.P. Thornberry (ed.), Developmental dile, ile, of crime and
delinquency (pp. 131-161). New Brunswick, NJ: Transaction.

Sampson, R.J., Morenoff, J.D., & Gannon-Rowley, T. (2002). Assessing "neighborhood effects":
Social processes and new directions in research. Annual Review of Sociology, 28, 443-
478.

Sampson, R.J., & Laub, J.H. (2003). Life-course desisters? Trajectories of crime among
delinquent boys followed to age 70. Criminology, 41, 555-592.

Sampson, R.J., & Raudenbush, S.W. (2004). Seeing disorder: Neighborhood stigma and the
social construction of "broken windows." Social Psychological Quarterly, 67, 319-342.

Sampson, R.J., Laub, J.H., & Eggleston, E.P. (2004). On the robustness and validity of groups:
Response to Daniel Nagin. Journal of Quantitative Criminology, 20, 37-42.

Sampson, R.J., & Laub, J.H. (2005a). Seductions of method: Rejoinder to Nagin and Tremblay's
"Developmental Trajectory Groups: Fact or Fiction." Criminology, 43, 905-914.









Piquero, A.R. (2007). Taking stock of developmental trajectories of criminal activity over the
life course. In A. Liberman (ed.), Longitudinal research on crime and delinquency. New
York: Springer.

Piquero, A.R., & Brezina, T. (2001). Testing Moffitt's account of adolescence-limited
delinquency. Criminology, 39, 353-370.

Piquero, A.R., Blumstein, A., Brame, R., Haapanen, R., Mulvey, E.P., & Nagin, D.S. (2001).
Assessing the impact of exposure time and incapacitation on longitudinal trajectories of
criminal offending. Journal ofAdolescent Research, 16, 54-74.

Piquero, A.R., & Buka, S. (2002). Linking juvenile and adult patterns of criminal activity in the
Providence Cohort of the National Collaborative Perinatal Project. Journal of Criminal
Justice, 30, 259-272.

Piquero, A.R., Brame, R., Mazzerole, P., & Haapanen, R. (2002). Crime in emerging adulthood.
Criminology, 40, 137-169.

Piquero, A.R., Farrington, D.P., & Blumstein, A. (2003). The criminal career paradigm. Crime
and Justice, 30, 359-506.

Piquero, A.R., & White, N. (2003). On the relationship between cognitive abilities and life-
course-persistent offending among a sample of African Americans: A longitudinal test of
Moffitt's hypothesis. Journal of Criminal Justice, 31, 399-409.

Piquero, A.R., & Moffitt, T.E. (2004). Life-course-persistent offending. In J.R. Adler (ed.),
Forensic Psychology: Concepts, debates, and practices. Portland, OR: Willan Publishing.

Piquero, A.R., Brame, R., & Lynam, D. (2004). Studying career length through early adulthood
among serious offenders. Crime & Delinquency, 50, 412-435.

Piquero, A.R., Brame, R., & Moffitt, T.E. (2005). Extending the study of continuity and change:
Gender differences in the linkage between adolescent and adult offending. Journal of
Quantitative Criminology, 21, 219-243.

Piquero, A.R., Brezina, T., & Turner, M.G. (2005). Testing Moffitt's account of delinquency
abstention. Journal ofResearch in Crime andDelinquency, 42, 27-54.

Piquero, A.R., Moffitt, T.E., & Lawton, B. (2005). Race and Crime: The contribution of
individual, familial, and neighborhood level risk factors to life-course-persistent
offending. In D. Hawkins & K. Kempf-Leonard (eds.), Our children, your children:
Confronting race and ethnicity in the American juvenile justice system. Chicago, IL:
University of Chicago Press.

Plomin, R. (1990). The role of inheritance in behavior. Science, 248, 183-188.

Plomin, R., Chipuer, H.M., & Loehlin, J.C. (1990). Behavioral genetics and personality. In L.A.
Pervin (ed.), Handbook ofpersonality theory and research. New York: Guilford Press.









conditions because of their parent's vulnerability to a host of problems including their inability to

deal with difficult children because they lack sufficient coping resources (see also Synder &

Patterson, 1987). Furthermore, Moffitt (1997) has found that these life-course-persistent

offenders were always more likely to be involved in crime regardless of the status of the

neighborhood in which they resided.

Moffitt continues by asserting that age will exert a direct influence on the strength of the

association between an individual's underlying criminal propensity and their offending patterns.

According to Moffitt, it is expected that criminal propensity should be strongly associated with

antisocial behavior in early childhood, yet the strength of the association should be greatly

reduced in adolescence because the adolescent-limited offenders exhibit the highest prevalence

rates during this time period and these offenders lack a criminal propensity. The strong

correlation between criminal propensity and crime should again resurface in adulthood because

the life-course-persistent offenders become the predominant group once again in this time period

(see also Piquero & Brezina, 2001).1

Guided by prior research, Moffitt and colleagues have offered some theoretical refinements

to the two offender typologies to explain differential rates of offending across sex. The

adolescent-limited type exemplifies female offending patterns (Moffitt, 1994) and life-course-

persistent offending pattern is almost exclusively a male phenomenon because (1)



1 It is important to note here that there is a strong degree of overlap between Moffitt's life-course-persistent offender
and Gottfredson and Hirschi's (1990) characterization of an offender with low self-control. The main similarity is
found in their definitions of propensity, which is believed to be influenced by impulsivity, hyperactivity, and low
verbal ability (see also Bartusch et al., 1997). Similarly, in line with Moffitt et al.'s (1994) sex expectations for life-
course-persistent offending, Gottfredson and Hirschi (1990:148-153) argue that male and female differences in
opportunity or supervision can explain the disproportionate rates of offending across sex. Blackwell and Piquero
(2005) found support for the relationship between low self-control and offending among males and females, but
indicated that there were sex differences in their responses toward their parents' control. Therefore, it is likely that
Gottfredson and Hirschi would argue that sex differences in self-control can explain the concentration of life-course-
persistent offending patterns among males.









proposed taxonomy of offenders and highlights the features that distinguish an adolescent-

limited offender from a life-course-persistent offender. Also discussed is Moffitt's (1994; see

also Moffitt et al., 1994) more recent claims as to why life-course-persistent offending may be

limited almost exclusively to males.

Moffitt's Taxonomy

Moffitt (1993) identifies the adolescent-limited offenders as normative and believes they

include individuals who are generally well socialized and adjusted, do not begin their offending

early in life, and are simply acting out a passing phase in the developmental process of

adolescence. These individuals only offend because of the process that she identifies as "social

mimicry," whereby the youth achieve status and power in their adolescent social world by

mirroring the antisocial behaviors of others (most notably the behavior of their life-course-

persistent counterparts). Therefore, through direct observation and reinforcement, the

adolescent-limited offenders begin to engage in similar behaviors in order to assert their

individuality and emerging "adulthood" status. However, although their rates of offending may

converge with that of their life-course-persistent counterparts only briefly in mid-adolescence

(ages 14 to 17), the adolescent-limited offenders (by definition) desist from offending soon after

this developmental time period. These adolescent-limited offenders are also typified by

inconsistent, erratic, and brief offending careers. Moffitt (1993) further argues that they may

shoplift and smoke some pot, but at the same time follow school rules.

In contrast, Moffitt (1993) identifies the life-course-persistent group as offenders who

begin their criminality/behavior problems early in childhood and exhibit an ongoing pattern of

behavioral problems throughout their life. These offenders begin biting and hitting at age 4,

shoplifting and truancy at age 10, selling drugs and stealing cars at age 16, robbery and rape at

age 22, and fraud and child abuse at age 30 (Moffitt, 1993:679). Moffitt argues that there is no









of 10% of the original Philadelphia birth cohort, Wolfgang, Thornberry, and Figlio (1987) found

that these chronic, originally defined as having had five or more police contacts as juveniles,

had increased the seriousness of their offending into adulthood. Tracy, Wolfgang, and Figlio

(1990) found even higher rates of chronic offending among members of the Second Philadelphia

Birth Cohort Study. They found that the chronic represented 7% of the cohort and 23% of the

offenders, yet were responsible for committing 61% of all the offenses including 60% of

homicides, 75% of forcible rapes, 73% of robberies, and 65% of the aggravated assaults.

In a similar study, Snyder (1988) examined a cohort of Phoenix and Utah youths and found

that the chronic, defined as having had five or more referrals to juvenile court, represented only

5% of the cohort. However, like the original Philadelphia Birth Cohort, they were responsible

for 51% of all the offenses committed including 61% of murders, 64% of rapes, 67% of

robberies, 61% of aggravated assaults, and 66% of burglaries. Shannon (1988) studied three

youth cohorts in Racine, Wisconsin and found a slightly higher percentage of chronic (8% to

14%) than the previously mentioned studies, but still found that they were involved in an

overwhelming majority (75%) of all of the felony arrests.

Hamparian, Schuster, Davis, and White (1985) in a follow-up study of the original

Columbus Cohort (Hamparian, Schuster, Dinitz, & Conrad, 1978) found that 60% of the

previously identified chronic (representing only 2% of the cohort) had been arrested at least

once by their early-twenties for an adult felony offense. These offenders were also more likely

to have had an earlier onset of delinquency, more frequent and serious offending, and displayed

continuity in offending from their youth through their mid-twenties.

The high frequency offending of chronic that has been illustrated when examining official

arrest data is also evident in self-report surveys. Results from the National Youth Survey (NYS)









a sharp declining trajectory in early adulthood and even perhaps an indication of desistance by

age 25. Unfortunately, given the data limitations and the reliance on only official measures of

offending to model the trajectories, this research is unable to answer the question of whether or

not this is true evidence of desistance. Furthermore, there is also the possibility that these

individuals in the high-rate offending group which accumulated a rather large number of police

contacts over time had become incarcerated for some period of time by their young adulthood

years, thus not providing them an opportunity to offend. This is also an empirical question that

cannot be explored with this data. Lastly, once the individual and neighborhood level factors

were estimated in multivariate models, the results indicated that while these risk/protective

factors significantly distinguished the adolescent-limited/peaked offenders and the chronic-

offenders from the non-offenders, these same variables (with the exception of sex) were not able

to significantly distinguish the adolescent-limited/peaked offenders from the chronic-offenders.

Once again, this is likely the result of the data loss that appeared to differentially affect offender

groups.

Directions for Future Research

The findings from this research provided mixed results for the various theories including

Moffitt's developmental typology and the saliency of sex when distinguishing offender groups,

Sampson and Laub's age-graded theory of informal social control, and social disorganization

theory. This research did generate some preliminary findings that can direct future research on

integrated multi-level theoretical models.

Sex was found to be rather robust for significantly affecting group assignment when

estimated as a covariate in the trajectory analyses. It also exerted a consistent independent effect

for significantly increasing the odds for group assignment to either an adolescent-limited/peaked

offender group or a chronic offender group. Future research would benefit from estimating sex-









probability (see Table 5-6). Only in G2 did the mean posterior probabilities fall slightly below

.90 but still remained well above the .70 cut-off (Nagin, 2005). Consistent with the previous

model, the results in Table 5-7 indicated that being male significantly increased an individual's

likelihood of being assigned to either offender group (adolescent-limited/peaked or chronic)

when compared with the non-offender group.

Table 5-8 presents the BIC model selection criteria for the trajectory analysis using the

data from the 1942 Racine continuous interviewed residents n i/h block data that provides the

trajectories and groups that will be of interest in the 1942 ANOVA and mlogit analysis that

follows. A three-group model once again provided the best fit for these data and the average and

predicted offending curves are plotted below in Figure 5-3.

Gl, representing 70.2% of the sample, are virtually a non-offending group with little to no

police contacts during the ages of 6 to 25. The second group, representing 20% of the sample is

similar to the adolescent-limited offender type proposed by Moffitt (1993) with the individuals in

this group beginning to offend in adolescence before peaking around the age of 16 before

declining and showing evidence of having desisted by 21. No individual in this group had

accumulated more than 5 total police contacts during the twenty years of follow-up. Also

consistent with the earlier analyses, the third group exhibited an early onset of offending (prior to

age 10), relatively high levels of offending throughout adolescence before reaching a peak at age

16; however, these individuals continued to maintain comparatively high rates of offending into

adulthood and through the age of 25, although the trajectory was declining. No individual in this

group had accumulated less than 4 police contacts and a handful of these individuals had as

many as more than ten police contacts prior to reaching age 26 (M=8.09; SD=3.30). This small

number of Racine residents amassed 186 police contacts between the ages of 6 and 25.









Gottfredson and Hirschi refer to as criminal propensity, does not necessarily mean a crime will

occur. The theorists indicate that this propensity will only cause an individual to engage in a

crime or an analogous behavior once it converges with an opportunity to commit a criminal act.

However, this temporal convergence of propensity and opportunity is not meant to argue the fact

that it occurs without the individual's knowledge of some social, legal, or natural sanctions

associated with their behavior. It is meant to indicate that individual's with low self-control will

tend to behave criminally given the opportunity because they typically perceive fewer negative

consequences and have less to lose when compared with individuals with high self-control

(Gottfredson & Hirschi, 1990).

In contrast to Gottfredson and Hirschi's (1990) viewpoint, a state-dependence

interpretation (or dynamic approach) has also been presented to account for the positive

association between past and future criminal behavior. This perspective asserts that prior

participation in criminal activity reduces internal inhibitions and external constraints to future

criminal activity and intensifies the prior offender's motivation to commit future crime. Or in

other words, the prior commission of a crime causes a subsequent increase in the probability of

an individual engaging in a future crime (Nagin & Paternoster, 1991; Nagin & Farrington, 1992).

The state-dependence hypothesis also recognizes the impact of life events on criminal behavior

(Piquero, Brame, Mazerolle, & Haapanen, 2002). In support of this interpretation, Sampson and

Laub contend that:

Lives are often unpredictable and dynamic; exogenous changes are ever present. Many
changes in life result from chance or random events in individual lives...while other
changes in life directions stem from macro-level 'exogenous shocks' (1995:50).

Despite the existing points of contention between the two rival explanations, the results

from Piquero et al. (2002) based on data from 524 males released from California Youth

Authority (CYA) institutions indicated that criminal offending appears to involve a combination









structure (number of children in family) were also distinguishing features for group membership

among the 1949 continuous residents.

While the preliminary results from the f-test in the ANOVAs revealed the concentration of

risk factors among the offender groups, it was necessary to conduct Tukey's B post hoc tests in

order to determine which (if any) of the groups significantly differed from one another.

Concerning the 1942 cohort, individuals who were in either offender group lived in significantly

more socially disorganized communities and both the adolescent-limited/peaked groups (G2) and

the high-rate chronic groups (G3) had significantly more males and G3 was characterized by a

significantly greater number of Blacks. G1 also had a significantly greater number of individuals

reporting having grown up in a two natural parent household when compared with both G2 and

G3. In addition, all of the groups' mean number of police contacts significantly differed, or in

other words G3 had significantly more police contacts than G1 and G2, and G2 had significantly

more police contacts than G1.

When turning to the Tukey's B post hoc tests results for the 1949 cohort, while individuals

in G3 and G2 lived in more socially disorganized areas than G1 the groups means did not

significantly differ. Comparatively, there was a host of individual level factors that significantly

varied particularly when comparing G3 and G2 to G1. The offender groups had significantly

more males and Blacks in addition to having a significantly greater number of children in the

household and had significantly higher perceptions of formal social control (police presence)

when compared with G1 (although G2 and G3 did not significantly differ from each other). The

the adolescent-limited/peaked (G2), and only 33.3% of the chronic offenders (G3) experienced marriage by age 25
and of those that reported being married, the mean age of marriage was 21.0 (G1), 21.1 (G2), and 21.3 (G3).
Comparatively, 96.7% of G1, 95.5% of G2, and 91.6% reported having their first full-time job by age 25 and the
mean age of first full-time employment was 19.0 (G1), 18.8 (G2), 17.3 (G3). Thus it appeared that typically fewer
adolescent-limited/peaked offenders and fewer chronic offenders reported having experienced marriage and/or
employment by age 25, yet there was not consistent evidence that the mean age of having experienced these life
transitions varied greatly across groups and by cohort.









(polytomous) regression models (Models 1 and 2), with the non-offender group omitted in order

to serve as the reference category. The following sections describe the group-based modeling

technique designed by Nagin and Land (1993) that will be utilized in this study for identifying

the offender groups along with a discussion of the existing debate that has been ongoing related

to the usefulness of the methodology.

Study Limitations

There are a series of limitations of the data source itself and the measures used in this study

that are worth noting. First, the timeliness of the data is an issue since the cohorts themselves

were born in the 1940s, socialized as children and/or adolescents in the 1950s and 1960s.

Therefore, it is likely that these samples may not be generalizable to other cohorts from more

recent time periods because they were born during WWII (e.g., 1942 cohort) or are from the

baby-boomer generation (e.g., 1949 cohort) who grew up in the post-WWII era. Second, Racine,

WI was a Midwestern industrial city with an estimated population of 89,000 residents, and it is

not likely that the results from this study would apply to the larger metropolitan urban cities of

today (i.e., cities with populations greater than 100,000).

Third, particularly during the time of data collection, Racine was a relatively

homogenous city with limited numbers of minorities, and according to Shannon (1988), although

the overall population did rise to 95,000 persons by 1970, this was followed by a decline to

86,000 persons by 1980. He attributed this population decline to a growing number of young

people in need or work and increased female participation in the work force, but this was met

with a decreased number of individuals leaving for retirement. Taken together, the results

discovered here may not necessarily apply to areas with larger minority representation or

dissimilar population shifts.









Farrington (1986a) further argued that there is also the issue of disentangling cohort effects

when interpreting the age-crime relationship. Cohort effects are influences that are due to an

offender's membership in a specific cohort, such as belonging to the baby-boomer generation.

Taken together, all of these matters present problems because an individual's actual age is

always measured in number of years and typically equals the current year (period) subtracted by

birth year (cohort) (Farrington, 1986a).

Furthermore, according to Farrington (1986a) the inherent characteristic of the age-crime

curve that receives the least amount of attention is that it reflects variations in the prevalence

(how many individuals are involved in crime) rather than the incidence (how many crimes are

committed) of offending (see also Blumstein et al., 1988). While age-crime curves for

individuals are typically unimodal (with only one peak in the teenage years), Farrington (1986a)

argues that these curves do not mirror the aggregate curve that was referred to by Hirschi and

Gottfredson (1983). Additionally, individual crime rates have been shown to be constant during

a criminal career, thus providing evidence that arrest rates do not always decrease with age for

all offenders (Blumstein & Cohen, 1979).

Farrington (1986a) also showed that aggregate age-crime curves can be comparable, but it

is their measures of distribution and central tendency that vary across time, place, sex, and

offense type. Nevertheless, despite the conflicting arguments discussed above particularly

between Hirschi and Gottfredson (1983) and Farrington (1986a), there is virtually no definitive

evidence refuting the strong relationship between age and crime. Given the robustness of this

relationship, recent attention has drawn researchers to focus on how criminal behavior changes

throughout the various stages of life (Sampson & Laub, 1990, 1992) and this approach has









This definitional disagreement over the "appropriate" unit of analysis has been at the heart of

macro-level research for some time.

What is a Community?

Community theory gained increased attention within the realms of sociology in response to

the ecological work of Park and his colleagues at the University of Chicago in the 1920s and

1930s. Their study of the dynamics of urban life in Chicago supported the idea that variables

such as population size, density, and heterogeneity interact to produce communities that can be

characterized as "urban." Following a brief decline in community-based sociological research in

the 1950s and 1960s, the 1970s and 1980s hosted a general resurgence in attention devoted to the

"community/neighborhood" debate. This movement was fueled by research that relied on using

local community areas (Heitgerd & Bursik, 1987; Bursik, 1984, 1986), census tracts or some

aggregation thereof (Schuerman & Kobrin, 1986; Shannon, 1988, 1991; Taylor & Covington,

1988), police districts (Shannon, 1988, 1991), and even electoral wards (Sampson & Groves,

1989) as the units of analysis. The consistency among the findings that were produced based on

these varying aggregations and definitions of community/neighborhood was that there existed

significant differences in local politics, economics, and lifestyles between communities and that

the world is no longer comprised of "small isolated groups of human beings scattered over a vast

territory" (Wirth, 1938:2).

An extension to the "community/neighborhood" definitional debate and its historical

origins was more recently provided in Bursik and Grasmick's (1993) classic work,

Neighborhoods and Crime: The Dimensions of Effective Community Control. The authors began

with the contention that although consensus had yet to be obtained regarding the definition of

neighborhoods, there were several characteristics that the majority of sociologists would indicate

that delineated a neighborhood including: (1) a small physical area within a larger area; (2) some









Table 5-17. Tabulated Bayesian Information Criteria (BIC) for 1949 Racine Continuous
Interviewed Residents i/ i/i Block Data (N=323)
Optimal # of Groups BIC
Cubic ZIP Model
2 -1816.65
3 -1747.74
4 False Convergence


Table 5-18. Mean (Median) Posterior Probabilities for Group Assignments: 1949 Racine
Continuous Interviewed Residents i/i/ Block Data (N=323)
Group Prob Prob Prob 25th 75th
(GI) (G2) (G3) Percentile Percentile
1 (n=245) .98(.99) .02(.00) .00(.00) .98 .99
2 (n=66) .11(.01) .89(.98) .01(.00) .82 .99
3 (n=12) .00(.00) .00(.00) .99(.99) .99 .99
Note: Minimum and Maximum values presented for the group/prob(group) combination.



Table 5-19. Coefficient Estimates for Effect of Sex (being Male) on Trajectory Group
Membership: 1949 Racine Continuous Interviewed Residents ii ilh Block Data
(N=323)
Groups coefficient SE p-value
Cubic ZIP Model
2 1.493 0.364 <0.001***
3 1.887 0.790 0.016*
*p<.05 **p<.01 ***p<.001
a Group 1 (the lowest rate group) is the reference category of comparison.




Full Text

PAGE 1

1 TRAJECTORIES OF TWO RACINE BI RTH COHORTS: A THEORETICALLY INTEGRATED MODEL FOR EXPLAINING OFFENDING By WESLEY G. JENNINGS 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

2 2007 Wesley G. Jennings

PAGE 3

3 ACKNOWLEDGMENTS I would like to provide a brief acknowledgement to all of those who have helped me along the way to achieve this ultimate goal. First and foremost, I would like to thank my family for all of their continued love and support throughout my many, many years of schooling. They have been a source of strength and financial support. I would also like to sin cerely thank my wife for her willingness to understand how long this proc ess may take and working hard to support us financially throughout my graduate education. Lastly, I would lik e to offer an extreme amount of gratitude to all of the f aculty at the Department of Cr iminology, Law, and Society, and specifically to my committee me mbers who were steadfast in their support throughout my entire graduate student career. Alex Piquero, Lonn Lanza-Kaduce, Brian Stul ts, Karen Parker, and Chuck Peek played an extremely integral role in my success in completing this dissertation and have provided me with excellent advice, criticisms, and thoughts th at have certainly helped me mature as an academic professional. For this I will always be grateful and I hope to continue collaborating with all of these individuals for a long, long time.

PAGE 4

4 TABLE OF CONTENTS page ACKNOWLEDGMENTS...............................................................................................................3 LIST OF TABLES................................................................................................................. ..........6 LIST OF FIGURES................................................................................................................ .........8 ABSTRACT....................................................................................................................... ..............9 CHAPTER 1 INTRODUCTION..................................................................................................................11 Introduction................................................................................................................... ..........11 Background..................................................................................................................... ........11 Present Study.................................................................................................................. ........15 Layout of Dissertation......................................................................................................... ...16 2 THEORETICAL PERSPECTIVES........................................................................................18 Introduction................................................................................................................... ..........18 What is a Community?...........................................................................................................20 Social Disorganization.....................................................................................................21 Age-Crime Relationship a nd Criminal Careers...............................................................26 Developmental Criminol ogy/Life-Course Theory..........................................................33 Sampson and Laubs Age-Graded Theory of Informal Social Control...........................36 Moffitts Taxonomy........................................................................................................40 3 REVIEW OF THE LITERATURE........................................................................................45 Introduction................................................................................................................... ..........45 Impact of Social Disorganiza tion at the Individual Level......................................................45 Prevalence and Incidence of Chronic Offending.............................................................50 The Effect of Life Transitions on Offending...................................................................54 The Saliency of Moffitts Taxonomy across Sex............................................................58 Prior Applications of GroupBased Models of Offending..............................................61 Integrated Theoretical Model for Explaining Offending.................................................65 Hypotheses and Expectations..........................................................................................68 4 DATA AND METHODS.......................................................................................................70 Introduction................................................................................................................... ..........70 Data Sources................................................................................................................... ........71 Independent Variables.....................................................................................................73 Macro-level independent variables..........................................................................73

PAGE 5

5 Micro-level indepe ndent variables...........................................................................74 Dependent Variables.......................................................................................................81 Study Limitations.............................................................................................................. ......82 Methodology.................................................................................................................... .......85 Group-Based Modeling...................................................................................................85 A Seductive or Useful Statistical Method?.....................................................................90 Analytic Procedure..........................................................................................................96 Trajectory estimation................................................................................................96 Multinomial logistic regression................................................................................97 Study Strengths..............................................................................................................101 5 RESULTS........................................................................................................................ .....103 Introduction................................................................................................................... ........103 Descriptive Statistics......................................................................................................... ...103 Independent Variables...................................................................................................103 Dependent Variable.......................................................................................................105 Trajectory Analysis............................................................................................................ ...105 Analysis of Variance........................................................................................................... ..116 Multivariate Analysis.......................................................................................................... ..118 Conclusions.................................................................................................................... .......122 6 DISCUSSION AND CONCLUSION..................................................................................144 Discussion..................................................................................................................... ........144 Limitations.................................................................................................................... ........147 Directions for Future Research.............................................................................................150 Implications................................................................................................................... .......153 Final Thoughts................................................................................................................. .....156 APPENDIX: ZERO-ORDER CORRELATIONS BY COHORT...............................................158 REFERENCES..................................................................................................................... .......159 BIOGRAPHICAL SKETCH.......................................................................................................181

PAGE 6

6 LIST OF TABLES Table page 4-1 Resident Samples by Cohort, Size, Data Sources, and Offender Groups........................102 4-2 Observational Periods of Data Sources by Cohort..........................................................102 4-3 Social Disorganization Inde x Factor Loadings by Cohort...............................................102 5-1 Descriptive Statistics for the Macrolevel and Micro-level Measures by Cohort...........126 5-2 Tabulated Bayesian Information Crit eria (BIC) for 1942 Racine Continuous Residents...................................................................................................................... ....126 5-3 Mean (Median) Posterior Probabili ties for Group Assignments: 1942 Racine Continuous Residents.......................................................................................................127 5-4 Coefficient Estimates for Effect of Sex (being Male) on Trajectory Group Membership: 1942 Racine Continuous Residents...........................................................127 5-5 Tabulated Bayesian Information Crit eria (BIC) for 1942 Racine Continuous Interviewed Residents......................................................................................................127 5-6 Mean (Median) Posterior Probabili ties for Group Assignments: 1942 Racine Continuous Interviewed Residents...................................................................................128 5-7 Coefficient Estimates for Effect of Sex (being Male) on Trajectory Group Membership: 1942 Racine Continuous Interviewed Residents.......................................128 5-8 Tabulated Bayesian Information Crit eria (BIC) for 1942 Racine Continuous Interviewed Residents with Block Data ...........................................................................128 5-9 Mean (Median) Posterior Probabili ties for Group Assignments: 1942 Racine Continuous Interviewed Residents with Block Data ........................................................128 5-10 Coefficient Estimates for Effect of Sex (being Male) on Trajectory Group Membership: 1942 Racine Continuous Interviewed Residents with Block Data ............129 5-11 Tabulated Bayesian Information Crit eria (BIC) for 1949 Racine Continuous Residents...................................................................................................................... ....129 5-12 Mean (Median) Posterior Probabili ties for Group Assignments: 1949 Racine Continuous Residents.......................................................................................................129 5-13 Coefficient Estimates for Effect of Sex (being Male) on Trajectory Group Membership: 1949 Racine Continuous Residents...........................................................130

PAGE 7

7 5-14 Tabulated Bayesian Information Crit eria (BIC) for 1949 Racine Continuous Interviewed Residents......................................................................................................130 5-15 Mean (Median) Posterior Probabili ties for Group Assignments: 1949 Racine Continuous Interviewed Residents...................................................................................130 5-16 Coefficient Estimates for Effect of Sex (being Male) on Trajectory Group Membership: 1949 Racine Continuous Interviewed Residents.......................................130 5-17 Tabulated Bayesian Information Crit eria (BIC) for 1949 Racine Continuous Interviewed Residents with Block Data ...........................................................................131 5-18 Mean (Median) Posterior Probabili ties for Group Assignments: 1949 Racine Continuous Interviewed Residents with Block Data ........................................................131 5-19 Coefficient Estimates for Effect of Sex (being Male) on Trajectory Group Membership: 1949 Racine Continuous Interviewed Residents with Block Data ............131 5-20 Trajectory Groups by Sample and Cohort.......................................................................132 5-21 Mean Covariate Levels by Traj ectory Group Membership and Cohort..........................133 5-22 1942 Racine Continuous Interviewed Residents with Block Data : Odds Ratios and Probabilities for Macro-Level Predic tors of Offending Group Membership..................134 5-23 1942 Racine Continuous Interviewed Residents with Block Data : Odds Ratios and Probabilities for Micro-Level Predic tors of Offending Group Membership...................134 5-24 1942 Racine Continuous Interviewed Residents with Block Data : Odds Ratios and Probabilities for Macro-level and Micr o-level Predictors of Offending Group Membership.....................................................................................................................135 5-25 1949 Racine Continuous Interviewed Residents with Block Data : Odds Ratios and Probabilities for Macro-Level Predic tors of Offending Group Membership..................136 5-26 1949 Racine Continuous Interviewed Residents with Block Data : Odds Ratios and Probabilities for Micro-Level Predic tors of Offending Group Membership...................136 5-27 1949 Racine Continuous Interviewed Residents with Block Data : Odds Ratios and Probabilities for Macro-level and Micr o-level Predictors of Offending Group Membership.....................................................................................................................137 A-1 1942 Cohort................................................................................................................ .....158 A-2 1949 Cohort................................................................................................................ .....158

PAGE 8

8 LIST OF FIGURES Figure page 3-1 Integrated Theoretical Model: Social Di sorganization, An Age-Graded Theory of Informal Social Control and Mo ffitts Developmental Taxonomy...................................69 5-1 1942 Racine Continuous Reside nts Group-Based Trajectories.......................................138 5-2 1942 Racine Continuous Interviewed Residents Group-Based Trajectories...................139 5-3 1942 Racine Continuous Interviewed Residents with Block Data Group-Based Trajectories................................................................................................................... ...140 5-4 1949 Racine Continuous Reside nts Group-Based Trajectories.......................................141 5-5 1949 Racine Continuous Interviewed Residents Group-Based Trajectories...................142 5-6 1942 Racine Continuous Interviewed Residents with Block Data Group-Based Trajectories................................................................................................................... ...143

PAGE 9

9 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 TRAJECTORIES OF TWO RACINE BI RTH COHORTS: A THEORETICALLY INTEGRATED MODEL FOR EXPLAINING OFFENDING By Wesley G. Jennings December 2007 Chair: Lonn Lanza-Kaduce Cochair: Alex Piquero Major: Criminology, Law, and Society Macro-level research evidencing the relationship between communities and crime dates back to the early ecological work of Park and Burgess and later further developed by Shaw and McKay into what is commonly referred to as soci al disorganization theor y. This research has since shown that poverty, ethnic heterogeneity, and residential mobility lead to the inability for members of the community to develop mechanisms of informal social control, which in turn leads to higher ra tes of crime. In contrast, micro-level theories such as Sa mpson and Laubs theory argues that individual pathways of offending can be altered by life tr ansitions such as marri age and employment. Comparatively, Moffitts taxonomy asserts that ther e are two theoretically distinct groups of offenders, where one group displays an adolesce nt-limited type of offending and the other group are considered to be life-course-persistent or chronic offenders. Moffitt has also argued that the life-course-persistent type is predominantly a male phenomenon. This study will provide one of th e first attempts to explore th e possibility of developing a theoretically integrated model for explaining gr oup offending using trajectory analysis and relying on data from two Racine Birth Cohorts. Re sults from this study relying on official data

PAGE 10

10 yielded varying levels of suppor t for all of the theoretical frameworks mentioned above. Specifically, all of the trajectory models discove red groups consistent with Moffitts typology, as well as uncovering groups not necessarily consis tent with her typology. Also consistent with Moffitt, sex was shown to be a significant cova riate for predicting group assignment. Bivariate support was found for Sampson and Laubs age-graded th eory of informal social control. Lastly, multivariate support was revealed for the impact of living in a socially disorganized area. Yet, once mlogit models were estimated incorporating the individual and neighborhood effects there was evidence of mediation. It appeared that in dividual and family variables at times completely or at least partially mediated th e effect of social disorganizatio n as a distinguishing feature of group membership. Implications for the current fi ndings and directions for future research are also discussed.

PAGE 11

11 CHAPTER 1 INTRODUCTION Introduction There exists a large body of research that has identified the association between the variation in the rate of offenders and community structural dynamics, most notably in the social disorganization literature (Bursik & Grasmick, 1993). Similarly, there has been a long line of research that has linked indivi dual-level factors to involvement in offending (Farrington, 1995). Yet, few efforts combine these levels of analysis On this score, Samp son (1997:32) provided an accurate summary of the current state of research in integrated approaches to studying offending in noting that few studies have successfully de monstrated a unified approach to individual and neighborhood-level dimensions of crime. The primary focus of this dissert ation is to present empirical da ta that will shed light on the question of the extent to which macro-level, micr o-level, or some combination of factors best relates to offending over the life course. In orde r to answers these questio ns, this study relies on extensive longitudinal data from two birth c ohorts of Racine, Wisconsin residents (Shannon, 1988, 1991) and utilizes a novel methodology for identif ying latent classes of offenders. After classifying the Racine residents into offender gr oups, a series of multinom ial regression models will allow for an assessment of the relative strength macro-level and micro-level factors for increasing and/or decreasing the likelihood of belonging to one tr ajectory group as opposed to another. Background While a handful of studies have compared ra tes of offending with individual/family SES and area (or school) SES and have found that th e highest rates of of fending generally were observed in the lowest SES areas among the indi viduals with the lowest SES (Braithwaite, 1979;

PAGE 12

12 Reiss & Rhodes, 1961; Wikstrom, 1 991a), only a few studies have i nvestigated the influence of other individual and community covariates on o ffending. Lindstrom (1996) found that the level of family interaction was significantly associated with juvenile offendi ng and this relationship was the most pronounced in more socially di sadvantaged neighborhoods. Kupersmidt et al. (1995) showed that juveniles living in singleparent families in low SES neighborhoods were more likely to be aggressive than their c ounterparts in high SES neighborhoods. Relying on a comparison of two different groups of P ittsburgh neighborhoods (i.e., underclass and nonunderclass), Peeples and Loeber ( 1994) indicated that the relati onship between race and juvenile offending was only apparent in the underclass neighborhoods. Even less research has targeted the integr ation of individual and neighborhood effects on offending under the auspices of the criminal career paradigm (Blumstein et al., 1986) and recent developmental frameworks. Wikstrom (1991b) conduc ted one of the first st udies to address this deficiency in the existing research and found that individuals that resided in the most socially disadvantaged areas had the highe st levels of both participati on and frequency of offending. Loeber and Wikstrom (1993) also showed that no t only did juveniles offend at higher rates (i.e., greater participation and frequency) in low SE S neighborhoods when compared with those from neighborhoods with higher SES a si gnificantly larger proportion of the offenders in the low SES neighborhoods became serious offenders. In a more recent study, Wikstrom and Loeber (2000) examined whether the interaction of risk and/or protective factors and community cont ext had an influence on early and late-onset offending among three samples of boys in 90 Pittsburgh neighborhoods. They found that neighborhood socioeconomic context did not have a direct effect on the prevalence of offending for boys who had a high number of risk factors, but for those who had balanced risk-protective

PAGE 13

13 factor scores, neighborhood contex t mattered. In addition, they i ndicated that individual riskprotective scores had a greater im pact on the rate of early-onset offending (before age 13) than did neighborhood context, but th at a direct neighborhood effect was observed for late-onset offending (after age 12). Lynam et al. (2000) found that the effects of impulsivity on juvenile offending was stronger in poor neighborhoods; however, livi ng in a disadvantaged neighborhood did not heighten nonimpulsive boys risk for juvenile of fending when compared with residing in more advantaged neighborhoods. Lynam et al. (2000) also noted that the direct neighborhood effects on individual juvenile delinquency were relativ ely weak, whereas the neighborhood effects were stronger when using self-reported measures of the quality of the neighborhood obtained directly from the inner-city youth. In a more recent study, Piquero, Moffitt, a nd Lawton (2005) investigated the possible interaction effects of individual, familial, and neighborhood-level factors on life-coursepersistent offending. Their data were derived from the National Collabo rative Perinatal Project (Baltimore site) and consisted of information fr om the original NCPP study that included early childhood and census-tract measures along with addi tionally including follow-up interviews with the children in the 1990s. Piquero et al. ( 2005) failed to find direct effects for either neighborhood or individual risk for life-course-persistent offendi ng; however, they did find that individual, family, and neighbor hood-level effects predicted li fe-course-persistent offending patterns among African-Americans in disadvantage d neighborhoods. In a ddition, while Piquero et al. (2005:230) concluded that their analysis demonstrated si milarities across race in lifecourse-persistent patterns of offending, ther e were marked differences in the neighborhood contexts where the Whites and African-Americans lived.

PAGE 14

14 With attention to the relative arbitrariness of how prior research has defined offender groups, more recent research applying Nagin a nd Lands (1993) group-based modeling technique has begun to identify latent classes of offenders th at exhibit distinct pathways of offending (for review see Piquero, 2007; see also Nagin, 2005). The consistent finding across these studies has been the identification of three main type s of offending groups: (1) non-offenders; (2) adolescent-limited offenders; and (3) chronic or high-rate offe nders. DUnger, Land, McCall, and Nagin (1988) were also ab le to discover similar offendi ng types across a variety of longitudinal data sources. It is important to emphasize here that the data relied on in this study is official data not self-report; therefore it is likel y that the non-offender group will be much larger than it would be if the trajectories were estimated using self-report data instead. Additionally, DUnger, Land, a nd McCall (2002) found evidence of relatively comparable latent offending classes using data from the Second Philadelphia Birth Cohort (Tracy, Wolfgang, & Figlio, 1990); however, the ad olescent-limited group they unc overed was characterized as adolescent-peaked because of their elevated levels of of fending in adolescence. This adolescentpeaked grouped has also been noted by other res earchers (for review, se e Piquero, 2007). Yet, other groups have also emerged su ch as low-level chronics or la te-onsetters (see DUnger et al., 1998, 2002). Prior applications of this methodology have al so found similar offender groups in male and female trajectories (Cote et al., 2001, 2002; DUnger et al., 2002; Fergusson & Horwood, 2002; Land, 2000; Piquero, Brame, & Moffitt, 2005) and ot her research has identified factors other than sex such as low socioeconomic status (McDermott & Nagin, 1998) and peer and school influences (Chung et al., 2001) as being strongly related to the likelihood of be ing classified as a chronic offender. Land (2000) has also shown th at sex was a significant predictor of chronic

PAGE 15

15 offending patterns, although its effect was comple tely mediated by the inclusion of neighborhood context. Present Study This current study contributes to the literature in following ways First, by drawing from the present advancements in the criminological lite rature regarding the application of trajectory analysis for classifying offenders (Nagin & Land, 1993; Nagin, 2005), this study seeks to provide another replication of th e latent class groups that have been previously identified by prior researchers (for review, see Piquero, 2007) such as the non-offenders, adolescentlimited/peaked offenders and chronic offenders across two Racine Birth Cohorts. Second, this study seeks to investigate whether sex is a disti nguishing feature of groupbased trajectories as has been discussed in prior applications of this group-based modeling techni que (see Cote et al., 2001, 2002; DUnger et al., 2002; Fergusson & Horwood, 2002; Land, 2000; Piquero et al., 2005). Third, this research will include an explora tion into the impact that social disorganization has on individual offending patterns. Four th, guided by two prominent developmental theoretical frameworks, specifica lly Sampson and Laubs (1993) ag e-graded theory of informal social control and Moffitts (1993) taxonomy of adolescent-limited and life-course-persistent offenders, this study seeks to explore the impact of key life transitions (marriage, employment) on desistance from offending in adulthood and assess the viability of Moffitts taxonomy. Fourth, the saliency of Moffitts (1994; see also Moffitt et al., 1994) claims that males are more likely to be life-course-persisten t offenders than females will al so be explored. Finally, this study seeks to address the rather limited amount of research integrati ng macro-level and micro-

PAGE 16

16 level factors that may simultaneously affect an individuals pathways of offending using data from two of Shannons (1988, 1991) Racine Birth Cohorts.1 Layout of Dissertation This dissertation will consist of six chapters. Chapter 2 will begin with a discussion of the various theoretical frameworks th at provide the basis for the st udy. More specifically, it will begin with a brief description of how scholar s have defined a community over time followed by an outline of the origins and theoretical tenets of social disorganization theory. Next will be an explanation of the relationship between ag e and crime and how research on the age-crime curve complements the emergence of the criminal career paradigm. Similarly, the narrative will continue with a discussion of how the criminal career paradigm had a dire ct contribution to the development of what has become known as deve lopmental/life-course cr iminology. This chapter will conclude with a presentation of Sampson and Laubs (1993) age-graded theory of informal social control (see also Laub & Sampson, 1993) and Moffitts (1993) taxonomy of offending, the latter providing the theoretical rationale for hypothesizing why diffe rent classes of offenders may exist and the former describing certain life transi tions that may affect an individuals offending trajectory. Chapter 3 will continue by offering an in-depth exploration into the prior extant literature that can be found examining each of the theoreti cal frameworks mentioned above. Most notably, this chapter will illustrate the links between so cial disorganization and individual patterns of offending, along with addressing the prior rese arch on chronic offending, crime over the lifecourse (i.e., Sampson and Laubs age-graded theo ry of informal social control), and offer a 1 This study relies on data from the 1942 and 1949 Racine Birth Cohorts, as opposed to all three of the cohorts (1942, 1949, 1955), because there is not any self-reported interview data available for the 1955 cohort precluding its use for investigating the particular individual-level factors addressed in this research. According to Shannon (1988, 1991), lack of continued federal funding prevented him from interviewing the most recent 1955 cohort.

PAGE 17

17 description of studies that have supported Moffitts (1993 ) proposed taxonomy and also demonstrated evidence that life-course-persi stent offending may be disproportionately found among males. This chapter concludes with a pr esentation of the findings from several recent applications of Nagin and Lands (1993) met hodological technique for identifying group-based offending trajectories. Chapter 4 will provide a systematic explanat ion of the source of the data and their limitations for the current rese arch along with an in-depth description of the 1942 and 1949 Racine Birth Cohorts used in the analysis. Th e sample and data information will be followed by a comprehensive identification of all of the meas ures incorporated in this study including the macro-level and micro-level fact ors and the two dependent measur es (number of police contacts and offending group membership). The final section of this chapter will expl ain the stages of the analysis as far as the ordering of the analytic procedures and the model specifications and will conclude with a brief discussion of the strengths of the study. Chapter 5 will present the findings from all of the statistical analyses including the results from (1) the group-based trajectory estimations a nd model selection criter ia; (2) a discussion of the importance of sex in predicting group assign ment; (3) description of factors that are associated with group assignment; and (4) an ex amination of the impact of macro-level and micro-level variables (both se parately and simultaneously) fo r significantly distinguishing between trajectory groups. The fi nal chapter (Chapter 6) will be gin with a further discussion on the findings as well as limitations of this resear ch. This section will be followed by an offering of some suggestions and guidance for future research on mutli-level modeling and theoretical integration and will conclude with a few final thoughts.

PAGE 18

18 CHAPTER 2 THEORETICAL PERSPECTIVES Introduction The origins of macro-level res earch, specifically social disorg anization, can be traced back to the early work in the first half of the twentieth century by Park and Burgess (1925) and the later expansions by Shaw and McKay (1942). These researchers and others that followed in their footsteps have formed what is commonly refe rred to today as the Chicago School. These researchers, most notably Shaw and McKay, we re influenced by the d eep-rooted sociological tradition of studying the eco logy of crime and focused on explaining why crime rates not individual criminality was higher in certain areas of the co mmunity. This tradition of studying and theorizing about crime in the aggregate persisted for some time until there became an increased interest in studying i ndividual-level factors that a ffect offending, which was strongly influenced by the increased use of survey me thodology in sociology and criminology. This early individual-level research sparked some of the most influential criminological theories that exist today including social learning theory (Burgess & Akers, 1966; Akers et al., 1979) and social control/bonding theory (Hirschi, 1969). By the late 1970s and early 1980s individual-leve l research had evolved in the sociological and criminological literature and led to what is still considered one of the most robust relationships in the social scienc es, i.e., the age-crime relationshi p, or the notion that crime rises to a peak in middle to late adolescence follo wed by a precipitous decl ine in early adulthood. Hirschi and Gottfredson (1983) argued that there is no need for any theory to account for the relationship between age and crime because it is invariant across culture, time, and space. Although, there has been a relatively strong amount of debate over this early a ssertion (which will be returned to later), th is along with evidence from Wolfgang et al.s (1972) research on

PAGE 19

19 chronic offending fueled the development of the criminal career paradigm. This perspective involves a series of dimensions that will be addressed below, but most importantly, this framework highlighted the similarities and di fferences among offenders and their criminal careers (see Blumstein et al., 1986). Coinciding with this new research interest was the birth of developmental criminology or life-course criminology. This theoretical pers pective focuses on how involvement in crime changes over the life course. It hi ghlights factors such as life tran sitions or turning points that can alter an individuals o ffending trajectory (Elder, 1985). In addition, Sampson and Laub (1993) formulated an age-graded th eory of informal social contro l that recognizes that different types of individuals have different offending ca reers and that there ar e age-dependent or agegraded processes that influence, shape, or al ter ones pathway of offending. In comparison, one of the most recognized developmental theories of crime (other than Sampson and Laubs agegraded theory of informal social control) is that of Moffitts (1993) taxonomy of adolescentlimited and life-course-persistent offenders. Moffitt argues that there are three theoreti cally and qualitatively distinct groups of individuals including: (1 ) the non-offenders; (2) the adolesce nt-limited offenders; and (3) the life-course-persistent offenders. Furthermore, she also argues that the patterns of offending for the two latter groups (i.e., the groups who are actually involved in o ffending) are driven by different causal processes, have different ages of onset, and are particular ly distinct in their frequency of offending and career lengths. All of the theoretical issues touched upon above will be the focus for the remainder of this chapter, but it is first important to briefly visit the debate among scholars (sociologists and criminologists alike) on what is a community or neighborhood.

PAGE 20

20 This definitional disagreement over the appropriat e unit of analysis has been at the heart of macro-level research for some time. What is a Community? Community theory gained increased attention w ithin the realms of sociology in response to the ecological work of Park and his colleagues at the University of Chicago in the 1920s and 1930s. Their study of the dynamics of urban life in Chicago supported th e idea that variables such as population size, density, and heterogene ity interact to produce communities that can be characterized as urban. Following a brief decl ine in community-based sociological research in the 1950s and 1960s, the 1970s and 1980s hosted a gene ral resurgence in atte ntion devoted to the community/neighborhood debate. This movement was fueled by research that relied on using local community areas (Heitgerd & Bursik, 198 7; Bursik, 1984, 1986), census tracts or some aggregation thereof (Schuerman & Kobr in, 1986; Shannon, 1988, 1991; Taylor & Covington, 1988), police districts (Shannon, 1988, 1991), and ev en electoral wards (Sampson & Groves, 1989) as the units of analysis. The consistenc y among the findings that were produced based on these varying aggregations and definitions of community/neighborhood wa s that there existed significant differences in local politics, economic s, and lifestyles between communities and that the world is no longer comprised of small isolated groups of huma n beings scattered over a vast territory (Wirth, 1938:2). An extension to the community/neighborhood definitional debate and its historical origins was more recently provided in Burs ik and Grasmicks (1993) classic work, Neighborhoods and Crime: The Dimensions of Effective Community Control The authors began with the contention that although consensus had ye t to be obtained regarding the definition of neighborhoods, there were several ch aracteristics that the majority of sociologists would indicate that delineated a neighborhood includ ing: (1) a small physical area wi thin a larger area; (2) some

PAGE 21

21 sort of common interest in the preservation of the inhabited area; and (3) a collective identity that tends to be constant over tim e (Bursik & Grasmick, 1993:6). Bursik and Grasmick (1993) further argued that social control may exist within communities in different forms or what they id entify as private, parochial, and/or public networks. However, they did contend that all communities tend to operate with the same underlying desire for informal and formal associa tions for eliciting social control in order to minimize crime (bearing in mind that particular crim es are likely to be differentially viewed as unconventional as well as differentially enforced across communities). Although, the community/neighborhood debate may be unresol ved, but there appears to be a movement toward fleshing out the strengths of its definiti onal attributes and providing a more encompassing characterization. This concept of community and the process of identifying particular factors that are associated or lead to crime were the basis for wh at has emerged as of one of the classic theories of crime, or what is commonly known in the disc ipline as social disorgan ization theory. This next section presents the historic al roots of the theory and lays out the core assumptions behind why crime should be higher in socially disorganized communities. Social Disorganization Early macro-level research inve stigating the importance of ecological characteristics for explaining crime rates has its origins from the Ch icago School and the work of Park and Burgess (1925) and Shaw and McKay (1942). The work of the latter scholars focused on attempting to offer a conceptual framework to explain the st ability of aggregate cr ime/delinquency rates over time in the city of Chicago despite the dynamic ch ange in the racial and ethnic composition of its neighborhoods (Shaw & McKay, 1942). Shaw and Mc Kay argued that the majority of the crime tended to be concentrated in pockets of the city that Park and Burgess (19 25) characterized as the

PAGE 22

22 zones of transition that surround th e central business district. Th e negative inverse relationship that Shaw and McKay observed between the ne ighborhoods spatial distance from the central business district and their crime rates provide d the framework for their theory of social disorganization. More specifically, their theory emphasizes the concentration of high levels of population turnover, racial and/ or ethnic heterogeneity, and poverty in the neighborhoods that exhibit the highest rate s of crime/delinquency. At the fundamental level, social disorganizati on is assumed to be the causal factor between neighborhood characteristics and o ffending patterns at the aggreg ate level because it emphasizes the importance of the ability of a communitys re sidents to realize shared goals and values and regulate the conduct of its citizen s (i.e., informal social contro l) (Berry & Kasarda, 1977; Kornhauser, 1978; Sampson & Groves, 1989; Sampson, Raudenbush, & Earls, 1997). High population turnover (i.e., residentia l instability) as identified by Shaw and McKay (1942), is seen as a structural antecedent that may impede the ab ility of communities to organize effectively and to self-regulate residents behavior because of th e constant out-migration that characterizes highcrime communities. This high rate of reside ntial mobility is believe d to produce a social atmosphere where many residents are uninterested in communities they hope to leave at the first opportunity (Kornhauser, 1978:78). This situation leads to th e potential failure of the communitys residents to forge interpersonal ties or even weak ties which some scholars have argued are the most important (see Bellair, 1997 ; Granovetter, 1973), because of the prevailing opinion that either they or their neighbors will soon be leaving the area. Racial and/or ethnic heter ogeneity is the second structur al characteristic recognized by Shaw and McKay (1942) that may perhaps hinder the ability of communities to develop mechanisms of informal social control. It is also causally linked to the high rates of population

PAGE 23

23 turnover mentioned above. This connection is prim arily due to the fact that in communities that are characterized by rapid compositional change, it becomes difficult to promote a unified front against the invasion of potentially thr eatening groups (Bursik & Grasmick, 1993:33). Furthermore, heterogeneity can inhibit comm unication, especially in communities that are plagued by language barriers. This will more th an likely make it increasingly difficult for the residents to solve or even express their co mmon problems and goals (Kornhauser, 1978:75). Shaw and McKay (1942) also indicated that high rates of poverty/low socioeconomic status are characteristic of socially disorgan ized communities. According to Shaw and McKay, this situation was attrib utable to the constant out-migration to the suburbs of residents that had the ability to move, and the area that the former residents left behind stil l tended to be plagued by poverty and increased crime/delinque ncy rates, regardless of which racial/ethnic group settled into the area. These individuals who lack the opportunity and/ or resources to escape this socially disorganized area (i.e ., the inner city) have been referred to by Wilson (1987) as the truly disadvantaged because they have to live in areas of extreme concentrated poverty and with a high density of subsidized housing. This dens ity of subsidized and predominantly renteroccupied housing has also been associated with cr ime rates (see Schuerman & Kobrin, 1986). Schuerman and Kobrin (1986) have indicated that although little is known about how neighborhoods evolve into high-cr ime areas, prior research has evidenced distinct stages of neighborhood development that typify the recipro cal effects of crime a nd community decline. Emerging is the first stage where neighborhood deterioration precedes the high-crime rates, but once communities move through the transitional stage and into the last stage (i.e., enduring phase), then the high-crime ra tes precede the further deterioration of the neighborhood. A host of factors has been at tributed to neighborhood de terioration including: movement from single to

PAGE 24

24 multiple family dwellings, residential mobilit y, broken homes, rise in young adult population, minority group relations, female participation in the work force, and even advanced education. In addition, according to Taylor (1995), prior scholars have contended that community decline is primarily the result of residentia l mobility or at least the reside nts desire to move from their current area. This resident turnover has been connected to many adverse neighborhood outcomes including weaker attachments among residents, lower sa tisfaction with the neighborhood, less local involvement, a nd decreased housing values. Although not explicitly referenced by Sh aw and McKay in the formulation of their theory, the relative inability of researchers to perhaps appropriately interpret Shaw and McKays underlying assumptions and/or dise ntangle the reciprocal effect s of social disorganization meaning that the factors identified above may both influence crime rates and foster social disorganizationled to a shift in the discipline in favor of indi vidual-level theories of crime causation (Land, 2000). However, several researcher s have made recent attempts to revitalize the work of Shaw and McKay (1942) and better define the causal pathwa ys that exist between neighborhood characteristics and offending (see Bursik & Grasmick, 1993; Sampson & Groves, 1989). Bursik and Grasmick (1993) put forth an atte mpt to define their systemic theory of neighborhood organization for explaining crime. At the fundamental level, systemic theory is predicated upon the ability of neighborhoods to control themselves and their surrounding environment in order to minimize their likeli hood of victimization a nd reduce crime. The authors argue that formal and informal mechanisms of social control are equally important if the neighborhood is to be successful in reducing crime.

PAGE 25

25 The private, parochial, and public spheres are regarded as the three specific elements that are important in eliciting th e necessary controls on the behaviors of the neighborhoods members. The private level of control is expressed through the strong intimate informal networks that exist within the communities th at are able to reduce crime by threatening or withdrawing social support from those members who engage in inappropriate or non-normative behavior. The parochial level of control is directly tied to the ability of so cial institutions such as the school and church to influence behavior th rough the shared sentimen ts of the neighbors for reducing victimization risk. The public level of control is perhaps the most difficult to obtain and emphasizes the ability of the neighborhood to mobilize government resources at the local level and acquiring crime controllin g agents (i.e., police) as an ex ternal resource to reduce the crime in their neighborhood. These three mechan isms are believed to manifest themselves through neighborhood-level in teractions over time; however, the greatest impediment for the formation of the ties is residential instabil ity, which presents a d ilemma being that those communities who need these three mechanisms the most will be the least capable of developing them (Bursik & Grasmick, 1993). Sampson and Groves (1989) also emphasized the importance of family disruption when examining the effect of social disorganization on crim e. They argued that marital and family disruption (i.e., divorce and singl e-parent households) may decrea se the ability for community residents to develop informal soci al controls. This assumption is based on the belief that twoparent households provide stabili ty and increased supervision fo r their children. Two parents also allow for more supervision over their person al property. Sampson (1987) has shown rather large effects for family disruption on rates of juvenile crime. Although, Sampson and Groves (1989) demonstrated that social disorganizatio n (as measured by sparse friendship networks,

PAGE 26

26 unsupervised teenage peer groups, and low organi zational participation) mediated a rather significant portion of the direct effect of family disruption on crime/delinquency rates as well as mediating the effects of other structural characte ristics such as classical social disorganization measures (low SES, residential mobility, and ethnic heterogeneity). In sum, the social disorganization fram ework provides a backdrop for explaining why crime rates may be higher in communities that are characterized by hi gh population turnover, ethnic/racial heterogeneity, and re sidential instability, but just how social disorganization can also be expected to have an affect on indivi dual outcomes will become clearer in the next Chapter (Ch. 3). The next section (along with th e remaining sections in Chapter 2) illustrates a complete shift in theoretical framework by not focusing on explanations of crime in the aggregate but by presenting a series of criminol ogical paradigms and micro-level developmental theories that have been offered to e xplain individual variation in crime. Age-Crime Relationship and Criminal Careers Perhaps the most commonly accepted fact about crime is the association between age and crime (Cohen & Land, 1987; Hirschi & Gottf redson, 1983; Hirschi & Gottfredson, 1985; Farrington, 1986a; Greenberg, 1985; Obrien, St ockard, & Isaacson, 1999; Sampson & Laub, 1990, 1992; Shavit & Rattner, 1988; Steffensme ier, Allan, Harer, & Streifel, 1989; Steffensmeier, Streifel, & Shihadeh, 1992). The progr ession of what has been referred to in the literature as the age-crime curve is that the crime rate, as measured by arrest, typically increases from the minimum age of criminal re sponsibility and peaks in the teenage years followed by a decline in early adulthood (Farri ngton, 1986a). Although th is general statement accurately characterizes the age-cr ime curve, in order to fully understand the complexity the agecrime relationship it is necessary to revisit Hi rschi and Gottfredsons ( 1983) original article discussing this phenomenon entitled Age and the Explanation of Crime

PAGE 27

27 Hirschi and Gottfredson (1983:54), in response to what they considered to be the misapplication of the age-crime distribution by schol ars in criticizing social theories of crime and supporting longitudinal methods over cross-sectional designs, assert that the effect of age is invariant across time, place, social and cultural conditions, and crime type. They indicated that rates of delinquency have been comparable over the last 150 years and that the age distribution of delinquency in England and Wales in the 1940s, Argentina in the 1960s, and the United States in the 1980s are virtually indi stinguishable. They further argued that although non-White offenders may start their delinque ncy earlier and offend at a highe r rate than Whites and that while boys offend at higher rates th an girls for particular offenses the age-crime curve is still invariant across race and sex (1983:556). In addition, they provided evidence against the differences in the age-distributions of property and person offenses. They illustrated that while official data may reveal diffe rent age-crime curves for pers on and property offenses, this difference is not observed in self -report data; therefore, the effects of age are indeed invariant across crime type. Hirschi and Gottfredsons (1983) early claims about the invariance of the effects of age across historical and cultural cond itions and crime type were s oon met by criticism. Farrington (1986a) specified that the main limitation of inte rpreting the types of age-crime curves used by Hirschi and Gottfredson (1983) is th at there is probable ambiguity in differentiating between age and period effects. Period eff ects are significant factors that ar e time-period specific, such as eras of high unemployment, riots, and/or civil unrest, and these influenc es are only related to changes in the crime rates of those cohorts that are aligned with that particul ar era in history. In contrast, age effects are universal changes that occur with age (i.e., natural maturation) that lead to physical deterioration for all cohort s during all periods of observation.

PAGE 28

28 Farrington (1986a) further argued that there is al so the issue of disent angling cohort effects when interpreting the age-crime re lationship. Cohort effects are in fluences that are due to an offenders membership in a specific cohort, su ch as belonging to th e baby-boomer generation. Taken together, all of these matters present prob lems because an individuals actual age is always measured in number of years and typically equals the current year (period) subtracted by birth year (cohort) (Farrington, 1986a). Furthermore, according to Farrington (1986a) the inherent characteristic of the age-crime curve that receives the least am ount of attention is that it reflects variations in the prevalence (how many individuals are involved in crime) rather than the incidence (how many crimes are committed) of offending (see also Blumstein et al., 1988). While age-crime curves for individuals are typically unimoda l (with only one peak in the teenage years), Farrington (1986a) argues that these curves do not mirror the aggreg ate curve that was referred to by Hirschi and Gottfredson (1983). Additionally, in dividual crime rates have been shown to be constant during a criminal career, thus providing evidence that ar rest rates do not always decrease with age for all offenders (Blumstein & Cohen, 1979). Farrington (1986a) also showed that aggregate age-crime curv es can be comparable, but it is their measures of distribution and central te ndency that vary across time, place, sex, and offense type. Nevertheless, despite the conflic ting arguments discussed above particularly between Hirschi and Gottfredson (1983) and Farr ington (1986a), there is virtually no definitive evidence refuting the str ong relationship between age and crime. Given the robustness of this relationship, recent attention has drawn research ers to focus on how criminal behavior changes throughout the various stages of life (Sam pson & Laub, 1990, 1992) and this approach has

PAGE 29

29 become characterized as the criminal career paradigm (Blu mstein, Cohen, Roth, & Visher, 1986). Blumstein et al. (1986) pointed out that the bulk of prior research in criminology was devoted to investigating crime using aggregate crim e rates and/or crimes pe r capita in the general population. This aggregate crime rate is the product of the particip ation rate of offending and the frequency of offending among the general populati on involved in crime, which according to Blumstein et al. makes studying individual criminalit y difficult. Hence, in an effort to better enable researchers to study diffe rential offending patterns, the crim inal career paradigm is unique because it divides the aggregate crime rate into two modules and has its own set of terminology and dimensions. The first module of the aggregate crime rate is participation in offe nding, or prevalence. Criminal career research distinct ly categorizes indivi duals into those who are criminally active and those who are not (Blumstein & Cohen, 1987). Participation m easurement is an expression of both the offenders involvement in crime and the range of the criminal acts committed during the observation period (i.e. serious or serious and non-serious offenses ). Regardless of the range of specified offenses or the particular observa tion period, active offenders always include new offenders (i.e., first offense occurs during obser vational period) and offenders who began their offending prior to the observati on period but remain criminally active during the period of study (Blumstein et al., 1986:18). Therefore, according to Piquero, Farrington, and Blumstein (2003:246), the longer the aver age duration of offending, the greater the c ontribution of persisters to measured participation in successive observation periods. According to Blumstein et al (1986), national estimates of participation in offending can range anywhere from a low of 25% to almost 45% depending on the operationalization of what

PAGE 30

30 constitutes an offense. Visher and Roth (1986) conducted one of the most comprehensive attempts to capture a true estimate of the participation rate of offending across the United States and Great Britain. They found that across both countries, th e average particip ation rate for non-traffic related offenses was approximately 30%. However, the rate fluctuated based on which type of criteria was employed, with the participation rate being the lowest when a stringent official measure was used such as conviction (West & Farrington, 1973, 1977) and the participation rate tended to be the highest when a more comprehensive official definition of offending was utilized such as police contact (Shannon, 1988, 1991). The second partition of the aggregate crime rate is frequency of offending. This refers to the average annual crime rate contributed by a pa rticular group of offe nders. Participation illustrates the pervasiveness of offending in th e general population, whereas the frequency refers to the intensity of th e individual offenders cr iminal activity (Blumste in et al., 1988). The frequency of offending is referred to as lambda ( ), in order to discriminate it from the aggregate rate used in the Uniform Crime Reports (U CR) (Blumstein & Cohen, 1987), and regarding lambda, most studies have found relative stabil ity over age, (Blumstein et al., 1986; Chaiken & Chaiken, 1982), sex (Blumstein et al., 1986), and crime type (P eterson & Braiker, 1980). The length of an individuals criminal caree r (duration) and the seriousness of offending are also important concepts in the criminal car eer paradigm (Blumstein et al., 1986:13). The duration of a criminal career indicates the total length of time of a criminals offending (Blumstein & Cohen, 1988). It represents the time that elapsed from the offenders initial offense to the time of his/her last offense (Blu mstein et al., 1986:13), or in other words the interval between an initiation a nd a termination in offending (Pique ro et al., 2003). Research has shown that there is a significant amount of variat ion in criminal careers, where many offenders

PAGE 31

31 criminal careers are brief and end in their teenage years and others continue to commit crimes up to and beyond middle age. Piquero, Brame, and Lynam (2004) have argued that the relative inattention of research regarding the duration of criminal careers is due to the difficulty in determining the true end of a criminal career. This difficulty has been addr essed by Blumstein et al (1986:18) where they stated the importance of distingu ishing total career length from residual career length, wherein the later refers to the estimated time remaining in an offenders criminal career given the specific observational period. Attention to this element is what Piquero et al. ( 2003) referred to as subsequently calculating estimates of career termin ation probabilities after an offender's arrest. Numerous studies (for review, see Piquero et al., 2003) have attempted to measure offender career lengths in actual number of years produci ng varying results ranging from a low of five years (Blumstein, Cohen, & Hsieh, 1982) to a relative high of 17.27 years (Piquero, Brame, & Lynam, 2004). Most recently, Kazemian and Farrington (2006) found evidence of a general decline in residual career length and offending frequency over time based on data from the Cambridge Study in Delinquent Development. The decline in residual career lengt h was observed in the original South London male sample and among th eir sons, although offending frequency did not decline as a result of conviction for the fathers. In addition, age of onset was associated with both residual career length and offending frequency, but the rela tionship was noticeably weaker among the fathers. The seriousness dimension of criminal career re search reflects both the number of offenses committed, as well as the particular patterns of switching (i.e., vers atility) among offenses (Blumstein et al., 1986:13). There are numerous t ypes of offenses in which criminals can engage

PAGE 32

32 in across the course of their career, as is there are wide variation among offenders in their levels of specialization and seriousness of offending. Offenders have been classified as either specialists (concentrating in only one offens e type or closely correlated offenses) or generalists (active in numerous types of re lated and non-related offe nses) and the types of offenses can also fluctuate, either escalating or de-escalating in seriousness (Blumstein et al., 1986:18). In a summary provided by Piquero et al. (2003:2 50) the authors note that most research has provided support for the generality of offending th roughout the duration of a criminal career and that specialization might occur more frequently among adult offenders versus juvenile offenders. In addition, there is some evidence in favor of offense versatility, although this switching tends to occur across clusters of logi cally related crime types (i.e., violence, property, and other). Also, minimal research has addressed the issue of co-offending patterns an d found that juveniles tend to commit crimes in groups, especially rega rding burglary and robbery, and males tended to co-offend with other males. Overall, criminal career research, comprised of all its relative dimensions, attempts to address questions of if how and why different factors influence the beginning of criminal activity, the continued participati on and prevalence of offending, and factors related to desistance or termination of criminal activity. This crimin al career paradigm also directly coincided with the movement toward developmental and life-cour se perspectives for explaining crime; however, the life-course approach differs from the traditio nal approach to criminal career research because it recognizes that there may be di fferent types of offenders that may have different types of careers (Farrington, 1999). The next section is de voted toward describing life-course theory and

PAGE 33

33 presents both sides of the debate regarding the persistent-heterogeneity versus state-dependent explanations of criminal behavior. Developmental Criminology/Life-Course Theory The life-course perspective can be defined as a series of pathways throughout the developmental process where there are varying ag e-dependent expectations and available options in an individuals decision-making process. Th ese decisions along with th e natural course of life events are seen to shape the trajectory of an i ndividuals life and can be influenced by transitions and turning points (Elder, 1985). Trajectories are viewed as pa thways of development over the life span such as work life, marriage, schooling, or involvement in crime. They are seen as longterm behavioral patterns that are marked by a sequence of life events and transitions. These transitions are then described as distinct life even ts that occur in shorter intervals that may alter an individuals behavioral trajec tory (such as first job, first ma rriage, first child, etc.) (Elder, 1985:31-32). Sampson and Laub (1990) extended this fram ework by focusing on the strong association between childhood events and adulthood experiences, a nd the idea that transiti ons can redirect an individuals life-course behavioral pathway. They indicated that life-course analysis focus on the length, timing, and the order of significant lif e events (including crime) and their effect on future social development (Sampson & Laub, 1992). Attention to the imp act of life events is critical because prior research has shown that committing a crime has a considerable behavioral influence on the likelihood of an individual comm itting crime in the future (Nagin & Paternoster, 1991). Loeber (1982) indicated that children who initially exhibi ted high rates of antisocial behavior were more likely to continue this behavi or than those who initia lly showed slower rates of antisocial behavior. Rely ing on data from 411 South L ondon males, Farrington (1995:956),

PAGE 34

34 concluded that offending may increase or decrease over time, but the worst offenders at one age still tend to be the worst at another age. This statement is essentially the basis of Gottfredson and Hirschis persistent-heterogeneity interpre tation (or static appro ach) of the relationship between past and future offending where they state that: The best predictor of crime is prior crim inal behavior [and] research shows that differences between people in the likelihood that they will commit criminal acts persist over timeand these differences appear early and remain stable over much of the life course (1990:107-108). Gottfredson and Hirschi (1990) expand upon this viewpoint in their di scussions regarding their theory of low self-control. According to Gottfredson and Hirsch i (1990: 97), the main underlying cause of low self-control is ineffective child-re aring. They argue th at it is the primary caregiver's (i.e., the parents) role to teach the child self-control through monitoring their behavior, recognizing deviant beha vior when it occurs, and eff ectively punishing the observed deviant behavior. These practices are believed to foster the child's development of self-control and an ability to delay gratificati on. The formulation of this core trait will also better enable the child to become more sensitive to the interests and desires of others, become more independent and willing to restrain their beha viors, and have a lesser propensity for using force or violence as a mechanism for resolving disputes or achieving their ends. Gottfredson and Hirschi (1990:90-91) further ar gue that the six primary elements that characterize individuals who exhibit low self-contr ol are impulsivity, insensitivity, physically as opposed to mentally oriented, short-sightedness, risk-takers, and displaying a low tolerance for frustration. Although these six co mponents are known to be distinct personality and behavioral traits, these various dimensions are all believed to be estab lished early on in childhood and coalesce within individuals to manifest a single construct that remains persistent throughout the life course. The fact that these elements fuse together to generate a unitary construct, what

PAGE 35

35 Gottfredson and Hirschi refer to as criminal pr opensity, does not necessarily mean a crime will occur. The theorists indicate that this propens ity will only cause an individual to engage in a crime or an analogous behavior once it converges w ith an opportunity to co mmit a criminal act. However, this temporal convergence of propensity and opportunity is not meant to argue the fact that it occurs without the indi vidual's knowledge of some social legal, or natural sanctions associated with their behavior. It is meant to indicate that individual's with low self-control will tend to behave criminally given the opportunity b ecause they typically perceive fewer negative consequences and have less to lose when comp ared with individuals with high self-control (Gottfredson & Hirs chi, 1990). In contrast to Gottfredson and Hirsch is (1990) viewpoint, a state-dependence interpretation (or dynamic approach) has also been presented to account for the positive association between past and future criminal be havior. This perspec tive asserts that prior participation in criminal activity reduces internal inhibitions and external constraints to future criminal activity and intensifies the prior offende rs motivation to commit future crime. Or in other words, the prior commission of a crime cause s a subsequent increase in the probability of an individual engaging in a future crime (Nagin & Paternoster, 1991; Nagin & Farrington, 1992). The state-dependence hypothesis al so recognizes the impact of life events on criminal behavior (Piquero, Brame, Mazerolle, & Haapanen, 2002). In support of this interpretation, Sampson and Laub contend that: Lives are often unpredictable and dynamic; e xogenous changes are ever present. Many changes in life result from chance or random events in individual liveswhile other changes in life directions stem from macro-level exogenous shocks (1995:50). Despite the existing points of contention betw een the two rival explanations, the results from Piquero et al. (2002) ba sed on data from 524 males re leased from California Youth Authority (CYA) institutions indi cated that criminal offending a ppears to involve a combination

PAGE 36

36 of the persistent-heterogeneity and state-dependence explanations. This finding led Piquero et al. to conclude that theoretical models must account for the relative contribution of both explanations in attempting to describe the ove rall progression of criminal behavior. Other research has also argued in favor of taking th e theoretical middle gr ound when studying crime (Paternoster & Brame, 1997; Paternoster et al., 1997). Additionally, previous resear ch has found evidence for the relationship between the stability of offending and antisocial behavior, but indicate that the trajectory of offending will differ across ages and vary over the life course (Nagin, Farrington, & Moffitt, 1995). Taking this into account, Sampson and Laub (1993) proposed a theoretical model that addresses the impact that life events and transitions can have on an individuals behavioral pathway and the following section provides a larger discus sion of their theory as contained within a developmental lifecourse perspective. Sampson and Laubs Age-Graded Theo ry of Informal Social Control Sampson and Laubs (1993) age-graded theory in formal social control can be viewed as a merging of the ideas of state dependence (Nagin & Paternoster, 1991) and cumulative continuity, wherein cumulative continuity assumes that adve rse life circumstances accumulate and exert a more significant effect as their frequency incr eases over time (Caspi & Moffitt, 1993; Moffitt, 1993) in order to create a model for explaining crim e over the life course. Th ey argue that social capital (investment or stakes in conformity) and turning points (employment, marriage) are important factors that may challenge traditional dete rministic theories of crime that assume that early childhood experiences lay the groundwork for later development (Bandura, 1982). According to Sampson and Laub (1993:301), tw o books rocked the criminological field and renewed researchers interest in childhood: A General Theory of Crime and Crime and Human Nature In Gottfredson and Hirschis (1990) work, A General Theory of Crime they

PAGE 37

37 argue that ineffective child re aring (including the pa rents failure to effectively monitor, recognize and punish their childs un wanted behaviors) in the early years of a childs life (prior to age 8) fosters the development of low self-cont rol. This latent trait is then viewed as a relatively stable phenomenon that increases th e likelihood of involvement in criminal and analogous behavior throughout an individuals life. Comparativ ely, in Wilson and Hernsteins (1985) book, Crime and Human Nature these scholars state that cr ime could be best explained by individual differences in impulsivity and temp erament, along with an interaction of these qualities with familial factors (s ee also Grasmick et al., 1993; Nagin & Paternoster, 1991). Fueled by this renewed interest in childhood and drawing from social control theory (Hirschi, 1969) and Elders life-course a pproach (1975, 1985), Sampson and Laub (1993:303) began to focus on age-graded life transitions and argue that institutions of formal and informal social control matter and vary over the life course. They contend that age-graded informal social control is essential for promoti ng interpersonal bonds that link i ndividuals to the larger social institutions in which they live (i.e., work, fa mily, school). They also acknowledge that both continuity and within-individual changes do occur over time and that specific life transitions and other factors related to development may interv ene in ones pathway of offending (i.e., also referred to as the stepping-stone approach by Farrington, 1986b). Sampson and Laub (1993) further developed thei r theoretical framewor k with attention to what they characterize as a series of building blocks. The first building block focuses on the intervening role of informal family and school social bonds. Their second building block focuses on the continuity of antisocial behavior th at begins early on in a childs development and extends throughout adulthood. They also recognize th at particular life even ts and social ties in adulthood can alter an individuals behavioral trajectory and this viewpoint is central to their

PAGE 38

38 third building block, i.e., the im portance of adult social bonds. They propose that social bonds developed in adulthood (employment, marriage) can explain involvement/lack of involvement in crime regardless of any underlyi ng individual criminal propensit y. In addition, they emphasize that it is not only the mere expos ure of individuals to these bonds the quality and intensity of these bonds is critical. Drawing on the work of Elder (1975, 1985), Sampson and Laub also hypothesize that different adaptations to key life events can affect an individuals trajectory, and crucial turning points such as divorce or death of a love one can redirect life trajectories. Two additional important concepts in Sampson and Laubs age-graded theory of informal social control include the proces ses of self-selection a nd cumulative continuit y. Self-selection is the classic birds of a feathe r flock together argument, and is based on the belief that individuals with low self-contro l resulting from ineffective parenting systematically sort themselves into criminogenic environments. For example, delinquent youth will tend to gravitate toward delinquent peers in adolescenc e, and similarly, will find delinquent spouses in adulthood and continue their offending beha vior (Gottfredson & Hirschi, 1990). In comparison, cumulative continuity (Moffitt, 1993) integrated with the concept of statedependence (Nagin & Paternoster, 1991) refers to the idea that involvement in delinquency has an incremental effect on the inte rpersonal social bonds formed in adulthood (i.e., labor force attachment, marital cohesion). For instance, an arrest and subseque nt incarceration in adolescence may lead the youth toward dropping ou t of school. This practice would logically affect their future job prospects and thus result in the failure of the individual to develop strong adulthood bonds to the labor force, which inev itably increasing their lik elihood of involvement in crime (Tittle, 1988).

PAGE 39

39 Furthermore, Sampson and Laub (1993) view change as occurring along a continuum where change can be shaped by a shift in societal roles and environments and this individual or environmental change can lead to the devel opment of social capital (Coleman, 1988, 1990; Nagin & Paternoster, 1992). Sampson and Laub ( 1993) further argue that the lack of social capital is one of the central antecedents of w eak social bonds. Similarl y, individuals who lack social capital are expected to be closed off from the informal networks among individuals in the social system, which also has a negative influenc e on their future life prospects (Coleman, 1990). However, Sampson and Laub (1993) also caution that just getting married or employed is insufficient for manufacturing social capital. Th ese social ties are only seen as important in cases where they impose obligations and restraints on individuals th at lead them to refrain from criminal involvement. According to Sampson and Laub (1993), this process offers an explanation for not only career de sistance but late-onset criminal ity because regardless of an individuals early developmental hi story, if they are able to form social capital or investment in societal institutions, then they will always be less delinquent/cri minal. Additionally, they argue that social capital can be reciprocal in nature and that there are racial ethnic, and structural variations in social capital (s ee Anderson, 1990; Short, 1990). In short, according to Sampson and Laub, their life-cour se theory seeks to: Return development to where it probably s hould have been all along, conceived as the constant interaction between individuals and their environment, coupled with purposeful human agency and random developmental noise (2005:40). Sampson and Laubs (1993) age-graded theory of informal social control soon sparked researchers interests; however, in the same year developmental psychologist Terrie Moffitt (1993) unveiled a theoretical typology that claims that there are differe nt types of offending trajectories that exist in the social world and her taxonomy classifies offenders as either adolescent-limited or life-course-persistent. Th e next section provides a discussion of Moffitts

PAGE 40

40 proposed taxonomy of offenders and highlights th e features that dis tinguish an adolescentlimited offender from a life-course-persistent offe nder. Also discussed is Moffitts (1994; see also Moffitt et al., 1994) more recent claims as to why life-course-persistent offending may be limited almost exclusively to males. Moffitts Taxonomy Moffitt (1993) identifies the adolescent-limited offenders as normative and believes they include individuals who are generally well socialized and adjusted, do not begin their offending early in life, and are simply acting out a pa ssing phase in the developmental process of adolescence. These individuals only offend because of the process that she identifies as social mimicry, whereby the youth achieve status a nd power in their adoles cent social world by mirroring the antisocial behaviors of others (m ost notably the behavior of their life-coursepersistent counterparts). Th erefore, through direct observation and reinforcement, the adolescent-limited offenders begi n to engage in similar behavi ors in order to assert their individuality and emerging adulth ood status. However, alt hough their rates of offending may converge with that of their life-course-persist ent counterparts only brie fly in mid-adolescence (ages 14 to 17), the adolescent-limited offenders (by definition) desist from offending soon after this developmental time period. These adolescen t-limited offenders are also typified by inconsistent, erratic, and brief offending careers. Moffitt (1993) further argues that they may shoplift and smoke some pot, but at th e same time follow school rules. In contrast, Moffitt (1993) identifies the li fe-course-persistent group as offenders who begin their criminality/behavior problems early in childhood and exhibi t an ongoing pattern of behavioral problems throughout their life. Thes e offenders begin biting and hitting at age 4, shoplifting and truancy at age 10, selling drugs and stealing cars at age 16, robbery and rape at age 22, and fraud and child abuse at age 30 (Mo ffitt, 1993:679). Moffitt argues that there is no

PAGE 41

41 reason to believe that these offenders ever extr aordinarily abandon their an tisocial tendencies and assume a prosocial identity later in life. A lthough prior research has shown that while offending may decline in middle to late adulthood among crim inal psychopaths, their antisocial personality traits (Cleckley, 1976) have b een shown to persist until ag e 69 (Harpur & Hare, 1994). Farrington et al. (1990) ha ve also indicated that early-onset offending is st rongly correlated with the seriousness of offending and the length of an individuals offending career. Furthermore, compared with their adolescen t-limited counterparts, life-course-persistent offenders problematic behavior is not the result of social mimicr y but it is believed to result from neurodevelopmental impairments that lead to a lack of self-control. This inability to form self-control (see also Gottfredson & Hirschi, 1990) affects their socialization and development and thus makes these individuals vulnerable to committing crime. Moffitt (1993) has also attributed this groups persistent offending to an interaction of liv ing in a poor social environment and neurological problems duri ng early fetal development that inhibit tempermental, cognitive, and behavioral developmen t. A series of studies have shown that one of the most robust predictors of antisocial beha vior is neurological deficiencies (see Moffitt, 1990; Moffitt & Henry, 1991; Hirschi & Hindelang, 1977). Moffitt (1993) continues to argue that a subs tantial portion of the antisocial behavior exhibited by these life-course-persistent offe nders can be attributed to heritability (or intergenerational transmission of antisocial beha vior) (Heusmann et al., 1984). Prior research has shown that there is a strong consistency between parents and thei r children in their temperament and personality traits (Plomin, Ch ipuer, & Loehlin, 1990) and in their cognitive abilities and intelligence scores (Plomin, 1990; Loehlin, 1989). Loeber et al. (2000) indicated that at-risk children are often exposed to adverse home environments and neighborhood

PAGE 42

42 conditions because of their parents vulnerability to a host of problems incl uding their inability to deal with difficult children because they lack sufficient coping resources (see also Synder & Patterson, 1987). Furthermore, Moffitt (1997) has found that these lif e-course-persistent offenders were always more likel y to be involved in crime regardless of the status of the neighborhood in which they resided. Moffitt continues by asserting that age will exert a direct influence on the strength of the association between an individual s underlying criminal propensity and their offending patterns. According to Moffitt, it is expected that crimin al propensity should be strongly associated with antisocial behavior in early ch ildhood, yet the strength of the association should be greatly reduced in adolescence because the adolescent-l imited offenders exhibit the highest prevalence rates during this time period and these offende rs lack a criminal propensity. The strong correlation between criminal propensity and crim e should again resurface in adulthood because the life-course-persistent offenders become the predominant group once again in this time period (see also Piquero & Brezina, 2001).1 Guided by prior research, Moffitt and colleagues have offered some theoretical refinements to the two offender typologies to explain di fferential rates of offending across sex. The adolescent-limited type exemplifies female offe nding patterns (Moffitt, 1994) and life-coursepersistent offending pattern is almost exclusively a male phenomenon because (1) 1 It is important to note here that there is a strong degr ee of overlap between Moffitts life-course-persistent offender and Gottfredson and Hirschis (1990) charact erization of an offender with low se lf-control. The main similarity is found in their definitions of propensity, which is believed to be influenced by impulsivity, hyperactivity, and low verbal ability (see also Bartusch et al., 1997). Similarly, in line with Moffitt et al.s (1994) sex expectations for lifecourse-persistent offending, Gottfredson and Hirschi (1990:148-153) argue that male and female differences in opportunity or supervision can explain the disproportiona te rates of offending across sex. Blackwell and Piquero (2005) found support for the relationship between low self-control and offending among males and females, but indicated that there were sex di fferences in their responses toward their pare nts' control. Therefore, it is likely that Gottfredson and Hirschi would argue that sex differences in self-control can explain the concentration of life-coursepersistent offending patterns among males.

PAGE 43

43 neuropsychological disorders are more preval ent among males (Rutter, 1983; Singer et al., 1968); (2) childhood-onset conduct problems are ve ry rare among girls (Werry, 1986); and (3) female delinquent behavior tends to be more erratic than male delinquent behavior (Moffitt et al., 1994:283). While large scale self-report studies versus th ose based on official data have shown that virtually every adolescent commits some illegal behavior (see Elliot et al., 1983), according to Moffitt (1993) there is yet another group of indivi duals who are characterized as abstainers or non-offenders. Similarly, Moffitt (1993) provides some theoretical speculations as to why these individuals fail to become involved in crime a nd delinquency. Moffitt argues that perhaps these youth experience late-onset puberty and therefore do not sense the maturity gap or perceive the social assets associated with crime (1993:689). Prior research has indicated that females that began menstruating after the age of 15 did not show any involvement in delinquency. The abstainers may also lack the opportuni ties to engage in delinquency. Adolescent delinquency rates have been shown to be higher in urban versus rural areas (Skogan, 1990), and Sampson and Groves (1989:789) have also commented that one of the str ongest correlates of violence is the presence of unsupervised groups of teenagers. In another study, Farrington and Wests (1990) research with their London male cohort showed that adolescents who resided in criminogenic environments but did not engage in delinquency could be characterized as nervous and withdrawn and also tended to entirely lack friends or only have a few friends. Most recently, Piquero, Turner, a nd Brezina (2005) provided one of the first empirical tests investigating Moffitts (1993) ab stainer hypothesis and found that there did exist a relatively small group of adolescents (a sligh tly greater proportion of females) that could be classified as abstainers. Piquero et al. (2005) indicated that teacher attach ment was inversely related to

PAGE 44

44 association with delinquent peers for the females and that less associati on with delinquent peers was related to being an abstainer among the males. However, Piquero et al. also noted evidence that was contrary to Moffitts expectations. Mo re specifically, they did not find that that the abstainers were more likely to be sad and depr essed or completely excluded from all peer networks. Piquero et al. (2005) showed that th e abstainers were still relatively active in the dating scene and were not typically loners, but ar gued that they were just adolescents who were involved in a more prosocial network of peers. Overall, Moffitts taxonomy provides a solid theoretical framework to explain not only differential offending patterns of individuals, but also offers sp eculations as to why some youth never become involved in delinquency. Furtherm ore, her offender typology can also account for sex differences in offending. This concludes the chapter on theoretical fr ameworks. It attended to the assumptions underlying social disorganization theory, the age-crime curve and the criminal career paradigm, the recent theoretical advancements of an age-graded theory of in formal social control (Sampson & Laub, 1993), and the existence of adolescent-limited and lifecourse-persistent offenders (Moffitt, 1993) including the re cent theoretical extensions to account for sex differences in offending. The next chapter cont ributes a thorough discussion of what the prior literature has discovered regarding these perspectives.

PAGE 45

45 CHAPTER 3 REVIEW OF THE LITERATURE Introduction The theoretical perspectives pr esented above have been the fo cus of a considerable amount of empirical testing. While, soci al disorganization theory has r eceived a substa ntial amount of attention for explaining crime rates at the macrolevel (for review, see Bursik & Grasmick, 1993; Sampson & Groves, 1989). The theory (particu larly surrounding neighborhood SES) has also been applied to accounting for individual variatio n in offending. This line of research will be discussed in more depth below. It will be followed by an examinati on of the studies (most notably that of Wolfgang et al ., 1972) that have identified the existence of a relatively small proportion of chronic high-rate offenders (bet ween 2% and 14% depending on which study is being referred to) who are responsi ble for at least half of the ove rall crime and even a larger percentage of the violent offenses. The next se ctions will present the research evidence on the impact that life transitions such as marriag e and employment can ha ve on an individuals pathway of offending and the literature that has provided support for Moffitts (1993) taxonomy as an explanation of sex differences in offending. Finally, this chapter will conclude with an illustration of the prior applica tions of Nagin and Lands (19 93) methodological technique for identifying groups of individuals with similar o ffending trajectories as well as providing several hypotheses that will be test ed in this current study. Impact of Social Disorganizat ion at the Individual Level In general, community-level scholars (e.g. Bu rsik & Grasmick, 1993) have associated the geographic concentration of soci oeconomic disadvantage to the so cial disorganization of the neighborhoods, and these disorgan ized neighborhoods have been ar gued to have adverse effects on the individuals and families that reside within those areas. In contrast, other scholars contend

PAGE 46

46 that the concentrated effects of poverty cannot be fully explained w ithout including control measures for both familial and individual-level characteristics (see Gephart, 1997:14). Although, while social disorganization theo ry (as originally formulated) sought to explain crime at the aggregate level, there are empirically-based reas ons to expect that social disorganization may also have an effect on individua l behavior. Simcha-Fagan and Sc hwartz took a step toward this possibility: An aggregate-level theory cannot be expected to provide a comprehe nsive explanation of individual delinquency or crim inality, it should clarify thos e contextual effects which, independently of or in interaction with th e adolescents predispositi on, affect individual behavior (1986:688). Jencks and Mayer (1990) provided a review of the current extant knowledge on the effects that social compositional measures (e.g., poverty) have on a series of developmental outcomes including: educational attainment, cognitive sk ills, criminal activity, and economic success. They indicated that the research is generally mi xed at best regarding the influence of average SES (in the aggregate) on the likelihood of an i ndividuals chances for planning to go to college, attending college, and/or gradua ting from college, especially for Whites. Research has shown that youth from high SES neighborhoods are more likely to have higher educational attainment compared with their counterparts who come fr om lower SES neighborhoods. In addition, Black mens chances of competing in the job market and finding well-paid jobs were greatly reduced when they resided in a heavily concentrated Black or welfare depende nt neighborhood (see also Aber et al., 1997). Klebanov et al. (1997) provided an extension of the resear ch on the influences of neighborhood effects on crime by specifically testi ng whether or not these effects are mediated and/or moderated (or interact with) by family processes such as maternal warmth, cognitive stimulation in the home, maternal coping style, av ailability of social s upport, and the physical

PAGE 47

47 environment of the home. They found that the positive effects of living in a high SES neighborhood on academic achievement, higher verbal ability scores, and less behavior problems and the negative effects of living in heterogenou s neighborhoods on verbal ability scores were mediated by the quality of the home learning envir onment. The effect of home environment was found across all age groups, measures, and res ource-rich and resource -poor neighborhoods. They also found support for the interaction of th e home environment and family characteristics in each neighborhood type in ways related to criminal outcomes. Similarly, additional research by Brooks-Gunn et al. (1993) showed that neighbor hoods had an effect on childhood IQ and the likelihood of becoming a teenage parent and/or dropping out of school. Thus, it appears that neighborhood level effects may be both direct an d indirect, but perhaps a substantial proportion of their effects are indirect in that they operate through i ndividual and family factors. Simcha-Fagan and Schwartz (1986:671) argue that structural char acteristics have an impact on the organizational networks among me mbers of the community, and these networks can subsequently affect an individuals abil ity to form social bonds to the community. Individuals who rent property as opposed to ow ning their own property are less likely to have a stake in the community in which they live, and thus may be less inclined to participate in neighborhood organizations that provide a mechanis m of informal social control. Similarly, individuals who rent their propert y may be less likely to care a bout the maintenance or up-keep of their dwelling because they do not intend to be there long. This lack of investment could lead to neighborhoods incivilities (W ilson & Kelling, 1982), which have been linked a breakdown in informal social control and crime (Sampson, 1995) Dwelling units that are over-crowded such as public housing units are also likely to lead to a sense of anonymity among the residents. The failure of individuals to ge t out of their units and develop informal social ties with their

PAGE 48

48 neighbors may be due to their exposure to viol ence and fear of crime (Roncek, 1981; Sampson, 1995). Sampson and Groves (1989) ha ve also shown that high rate s of residentia l instability were associated with juvenile delinquency and crime. Gephart (1997) reviewed th e developmental impact of community effects on adverse individual outcomes. In early childhood, children who reside in areas of extreme disadvantage face a slightly higher risk for having a low bi rth weight and suffering from child abuse. Comparatively, additional analyses have shown that the presence of affluent neighbors was associated with higher IQs and the presence of low-income neighbors was related to a greater likelihood of behavioral probl ems in early childhood. Furthermore, research has also found that the influence of a neighborhoods socioeconomic status on individual outcomes in early chil dhood and middle childhood were different across races. At age 3, the benefit of affluent nei ghbors (versus middle SES neighbors) on childrens IQ scores was more salient among White child ren than African-American children (BrooksGunn et al., 1993). Similar findi ngs were also found among children at ages 5 and 6 for IQ scores, verbal ability, and read ing achievement scores (Chase-L ansdale et al., 1997; Duncan et al., 1994). In addition, all of thes e studies revealed that the pr esence of low SES neighbors was positively associated with early childhood behavior problems. Gephart (1997) also indicated that research investigati ng neighborhood effects during late childhood and early adolescence found that commun ity impoverishment, the burden of childcare, and residential instability were significant predictors of va rious forms of delinquency such as violent crime, drug trafficking, juvenile de linquency, and teen pregna ncy (see Gephart, 1997). Two additional studies have also shown that neighborhood SES had a stronger effect for male adolescents than female adolescents (Entwisl e et al., 1994; HalpernFelsher et al., 1997).

PAGE 49

49 Research has also shown that living in a high SES neighborhood increased older adolescents chances of gra duating high school and attending college (Brooks-Gunn et al., 1993; Duncan, 1994; Ensminger et al., 1996; Halpern-Fe lsher et al., 1997). All of these studies suggested that the protective e ffect of living in a higher SES neighborhood was more salient for males. Crane (1991) has also provided evidence that males had the worst outcomes when living among low SES neighbors. In addition, a series of studies reviewed earlier have all consistently shown that living in a low SES neighborhood was positively associated with higher rates of criminal and delinquent behavi or (Braithwaite, 1979; Lindstrom, 1996; Loeber & Wikstrom, 1993; Lynam et al., 2000; Peeples & Loeber, 199 4; Piquero et al., 2005; Reiss & Rhodes, 1961; Sampson & Groves, 1989; Wikstrom, 1991a, 1991b; Wikstrom & Loeber, 2000). The U.S. Department of Housing and Urban Development (HUD) initiated what is known as the Moving to Opportunity for Fair Housi ng Demonstration (or MTO) in 1994 motivated by the success of the Gautreaux Program that has been in operation in Chicago for almost three decades. Rosenbaum (1995; see also Rosenbaum & Popkin, 1991) presented quasi-experimental evidence that Gautreaux youth w ho moved to the suburbs greatly increased their educational attainment and job opportunities when compared with those that remained in the poor urban areas. Intrigued by these findings and the idea that residential mobility may benefit individuals from disadvantaged communities, the MTO program became operational in five selected sites (Baltimore, Boston, Chicago, Los Angeles, and New York City). The MTO program involved randomly providing families vouchers allowing some families with children in high poverty areas to move to less poor areas with the h ope that this relocation would increase their educational and employment opportunities. Eligibility for the program was determined by the families that lived in Sect ion 8 public housing units with at least a 40%

PAGE 50

50 poverty rate according to 1990 census tract data. Prior MTO evaluations have indicated that residential relocation significan tly contributed to a reduction in violent arrests for males and better overall health among the children that moved from the public housing units (Katz et al., 2001; Ludwig et al., 2001). Katz et al. (2001) also noted signif icant reductions in criminal victimization among individua ls that moved out of the Section 8 housing. Leventhal and Brooks-Gunn (2003) noted that the MTO program had profound effects on mental health outcomes for the individuals and fa milies that relocated. They reported that the parents who moved reported less distress than did those that remained a nd the boys that moved (especially those between ages 8 and 13) repor ted significantly less depr ession, anxiety, and/or dependency problems when compared with their counterparts who were still in public housing. Most recently, Kling et al. (2007) noted the substa ntial mental health and education benefits for the adults and females, but commented on the ad verse effects associated with the males who received vouchers (for an exte nded review of MTO impact evalua tions, see also Orr et al., 2003). Overall, there is reason to believe that stru ctural factors includi ng those highlighted in social disorganization theory can have an eff ect on individuals outcomes and perhaps influence their pathway of offending. It is likely that living in areas character ized by social disorganization will increase an individuals likelihood of bei ng a chronic offender. This concept of chronic offending is the topic of the discussion presented in the following section. Prevalence and Incidence of Chronic Offending Relying on data from the well-known Philade lphia Birth Cohort Study, Wolfgang, Figlio, and Sellin (1972) provided some of the earliest ev idence of the existence of chronic offenders. Wolfgang et al. found that there were a group of delinquents that repres ented only 6% of the cohort and 18% of the cohort's delinquents but we re responsible for comm itting roughly half of all the offenses and about two-thir ds of all of the violent offenses In a ten-year follow-up study

PAGE 51

51 of 10% of the original Philade lphia birth cohort, Wolfgang, T hornberry, and Figlio (1987) found that these chronics, originally de fined as having had five or more police contacts as juveniles, had increased the seriousness of their offending into adulthood. Tracy, Wolfgang, and Figlio (1990) found even higher rates of chronic offe nding among members of the Second Philadelphia Birth Cohort Study. They found that the chronics represented 7% of the cohort and 23% of the offenders, yet were responsible for committing 61% of all the offenses including 60% of homicides, 75% of forcible rapes, 73% of robbe ries, and 65% of the aggravated assaults. In a similar study, Snyder (1988) examined a cohort of Phoenix and Utah youths and found that the chronics, defined as havi ng had five or more referrals to juvenile court, represented only 5% of the cohort. However, li ke the original Philadelphia Birt h Cohort, they were responsible for 51% of all the offenses committed includi ng 61% of murders, 64% of rapes, 67% of robberies, 61% of aggravated assaults, and 66% of burglaries. Sha nnon (1988) studied three youth cohorts in Racine, Wisconsin and found a sli ghtly higher percentage of chronics (8% to 14%) than the previously mentioned studies, bu t still found that they were involved in an overwhelming majority (75%) of all of the felony arrests. Hamparian, Schuster, Davis, and White ( 1985) in a follow-up study of the original Columbus Cohort (Hamparian, Schuster, Din itz, & Conrad, 1978) found that 60% of the previously identified chronics (r epresenting only 2% of the cohor t) had been arrested at least once by their early-twenties for an adult felony offe nse. These offenders were also more likely to have had an earlier onset of delinquency, more frequent an d serious offending, and displayed continuity in offending from thei r youth through their mid-twenties. The high frequency offending of chronics that ha s been illustrated when examining official arrest data is also evident in self-report survey s. Results from the National Youth Survey (NYS)

PAGE 52

52 have shown that about 5% of juveniles at each age level from twelve to seventeen years old can be classified as serious, chronic offenders. Thes e offenders are characterized as having engaged in three or more aggravated as saults, sexual assaults, gang fight s, and/or strong arm robberies within the previous year. In addition, on average these chronics committed 132 delinquent offenses annually and 8% of these offenses are considered serious and violent. Perhaps more alarming is the fact that 84% of the offenders iden tified as chronics in the survey had no official record of delinquency and this disparity between official measures and self-reported measures of offending cannot be underestimated (Elliot, 1994; Elliot, Huizinga, & Morse, 1986). Further support for the existen ce of chronic offenders was also provided in a series of RAND Inmate studies conducted during the late 1970s and earl y 1980s. Petersilia, Greenwood, and Lavin (1977) conducted the firs t of these studies and focused on the criminal careers of 49 inmates in a California medium-security prison. All of these offenders were currently serving time for armed robbery and were characterized as re peat offenders because they all had served at least one prior prison term. Petersilia et al. (1977) gathered information on the frequency of offending from structured interviews with the o ffenders and their official criminal records. Overall, the 49 inmates self-reported committi ng 10,500 serious crimes even though the majority of the offenders had spent the better part of their criminal careers behind bars. Peterson and Braiker (1980) carried out the second Rand Inmate study where they administered self-report questionn aires to 624 incarcerated male fe lons in five of Californias state prisons asking the inmates to report what ty pes of offenses they committed during the three years prior to their current imprisonment. Pete rson and Braiker found that the majority of the inmates did not specialize in offending and that most of the felons who committed one crime type at a high frequency also re ported significant criminal activity in other types of crime. On

PAGE 53

53 average, these chronic offenders represented about 8% of the inmates but self-reported committing over 60 crimes per year. Chaiken and Chaiken (1982) conducted the thir d Rand Inmate study investigating high-rate offending, wherein they analyzed the criminal hi stories and survey results from 2,190 prison and jail inmates in California, Texas, and Michig an. Chaiken and Chaiken identified what they referred to as a small group of violent pred ators or omni felons and these individuals reported committing robberies, assaults, and drug deals at disproportionately higher rates than the other inmates. In addition, they tended to commit burglaries, thef ts, and other property crimes at higher rates than even the offenders who self-reported to spec ialize in those specific crime types. Overall, all of the studies described above have shown strong empirical support for chronic offenders and these offenders represen t only 2% to 14% of the offending population (depending on which study is being referred to), yet they are respons ible for at least half of all the offenses committed (Farrington, 1999). The fact that a sizeable fraction of the crime problem can be attributed to a relatively small number of identifiable chronic offenders provides support for preventative measures and early identifica tion (see Piquero, Farringto n, & Blumstein, 2003). However, in contrast to the robust finding of chronic offending among various samples and time periods and the evidence that these individual s tend to be criminally active for long periods of time, Sampson and Laub (1993) argue that ke y life transitions such as marriage and/or employment can have an impact on an individual s pathway of offending, even if they are/were chronic offenders. The following section addres ses the empirical evidence in support of the effect that these various life events can have on behavioral outcomes, most notably desistance from crime.

PAGE 54

54 The Effect of Life Tran sitions on Offending Sampson and Laubs (1993) age-graded theory of informal social control takes the position that life key life transitio ns (including delinquency) can affect adult outcomes. Similarly, Caspi and Moffitt (1993) have referred to delinquency (particularly serious) as a knifing off of future opportunities, or in other words delinquency in ad olescence is expected to impede the formation of conventional social bonds that will c ontinue to have an affect in adulthood Relying on the Gluecks data (1950, 1968) Sampson and Laub (1993) presented preliminary support for their theory, whereby in dividuals who exhibited the strongest ties to work and family had the lowest rates of delinquenc y. In contrast, incarcer ation in the adolescent years was inversely related to job stability and positively associated with continued criminal involvement. Nagin and Waldfogel (1998) have also noted the advers e effects that an adolescent conviction has on future job opportunities amon g West and Farrington s (1977) London male cohort. Additional research has shown that invo lvement in delinquency results in the worst adult outcomes for lower class boys, particularly amon g lower class boys with a high frequency of police contacts (Hagan, 1991). Jess or et al. (1991) have also i ndicated that delinquency is not a major impediment for the formation of prosocia l adult bonds for middle class adolescents. Based on the results of Sampson and Laubs (1993) original test of their age-graded theory of informal social control, support was found th at delinquency was associated with crime in adulthood as well as being significan tly related to a series of adverse life events including being AWOL from the military, economic dependence, and having had martial problems. Contrary to what Gottfredson and Hirschi (1990) would pred ict, they also found significant independent effects for marital attachment and job stability on involvement in adult crime net of the effects of criminal propensity.

PAGE 55

55 Sampson and Laub (1993:317-320) c oncluded by offering six key issues to guide future research testing their theory. First, there is some element of luck, randomness, or chance in an individuals life course (also referred to by Rutter, 1989 as adve ntitious happenings). Second, there is individual variation in the effects that life transitions have on be havior. Third, macrolevel opportunity structures for marriage and labor market activity are important (i.e., segmentation of the labor market, ethnic encl aves). Fourth, criminal propensity cannot completely explain adult crime because life transitions can also affect opportunities and motivations to commit crime (see also Cohen & M achalek, 1988). Fifth, the effect of cumulative continuity is strongest among those in disadv antaged racial and econom ic conditions. Sixth, historical context matters. With attention to some of these suggestions, Horney et al. (1995) f ound clear evidence that meaningful change in life circumstances was associated with desistance from crime based on their interviews with 658 offenders in a Nebraska Correctional Facility. They also estimated hierarchical linear models that allowed them to control for criminal propensity and still found that the offenders reported reduced involvement in criminal activity when they were living with their wife. Overall, their results suggested that life transitions can alte r behavioral patterns of offending at particular periods of time, especially in the short-term. Laub et al. (1998) estimated trajectories with the Gluecks da ta and found that the timing and qua lity of marriages were what was important and that the inhibiting effects of marriage on crime were gradual and grew in intensity over time. Similar to Horney et al (1995), these findings were still observed while controlling for criminal propensity and a host of other factors. More recently, Piquero et al. (2002) developed a stakes in conformity index where they combined the life circumstances of employment and marriage to gauge the possible cumulative

PAGE 56

56 effects of local life circumstances on pathwa ys of offending among serious offenders from the California Youth Authority. Based on their results, Piquero et al. concluded that it is likely the case the these mechanisms of informal social control (or life transiti ons) can have a stronger effect on instrumental as opposed to interperso nal crimes because commission of the former tends to involve a relative degree of planning. Laub and Sampson (2003) estimated models that took into acc ount within-individual change and found that when controlling for age, individuals who were currently married had lower criminal propensity scores when compared w ith their scores when they were not married. They also found these differences in propensity existed in times where the individuals were employed or in active military service compared with periods when they were not. In another study testing Sampson and Laubs th eory, Warr (1998) provided an analysis of two waves of longitudinal data from the Nationa l Youth Survey of 1,725 individuals. Counter to Sampson and Laubs (1993) position, Warr hypothe sized that the link between life transitions and crime might be better explained by the effect that these transitions have on altering the relationships with delinquent peers (Akers et al., 1979, 1985; Sutherland, 1947). Warrs results did reveal that marriage was strongl y associated with desistance from crime, but he indicated that for many of these individuals, getting married was a key life transition that signaled a severing of heavy involvement with their delinquent peers a nd instead re-focused th eir attention to their spouse and the desire to have children. These findings led Warr to conclude that marriage not only reduces individuals motivation for continuing criminal behavior by re -directing their goals, it also greatly minimizes their opportunities for offending by limiting the amount of time that they have to spend with their former delinquent peers. In addition, Warr caution ed that there still exis ted the possibility of a

PAGE 57

57 selection effect in that individua ls who do in fact get married ar e perhaps qualitatively different from those who do not get married, and that i ndividuals who are not i nvolved with delinquent peers and/or engaged in crime may appear to be more suitable marriage prospects. Maume, Ousey, and Beaver (2005) have si nce reexamined Warr s (1998) hypotheses and conclusions using data from the National Y outh Survey, but improve upon Warrs earlier study by specifically investigating the e ffects of marital attachment on desistance from marijuana use and control for the possible selectivity bias in the relationship between marital attachment, delinquent peer associations, and desistance. In line with Samp son and Laubs theory, Maume et al. found that marriage was important to the desistan ce, particularly for those marriages that were characterized by high levels of attachment. C ontrary to Warrs (1998) findings, marriage still maintained significant directs effects for affecting desistance net of the effect of delinquent peer association (measured as the change in the number of delinquent associates). Uggen (2000) investigated the impact that employment (i.e., a turning point) has on an offenders recidivism relying on data from over 3,000 individuals from nine U.S. cities, which were originally part of the 1977 National Supported Work Dem onstration Project that targeted the underclass and the ghetto poor. The individu als who participated in this project were randomly assigned to either of two treatment groups (construction work or service jobs) or to a control group where no job placement was given. Th e work and arrests histories were collected for all of these individuals every nine months for three years. The results from the event history analysis yielded a variety of noteworthy findings First, although the program failed to reduce crime overall across the entire sample, its imp act was definitely age-graded. Job treatment significantly reduced recidivism among the older participants (p articularly aged 27 or older) whereas the job treatment had little to no effect on crim e among the young adults (late

PAGE 58

58 teens/early twenties). Therefore, it appeared that this life transition (employment) affected desistance from crime only for a certain age group of individuals (older adults) providing support for Sampson and Laubs (1993) theory. In a recent review of the state of their theory and in response to some noted criticisms (see Moffitt, 2005), Sampson and Laub (2005c:18) refined their views toward the influence of life transitions on adult offending by arguing that involvement in institutions such as marriage and work provide both short-term and l ong-benefits. In the short-term, these life transitions can alter an individuals motivation for crime by re-f ocusing their energy and attention toward conventional outlets (spending time and caring for their spouse/family or their job) and can produce the long-term benefit of fo stering their development of pros ocial adult bonds or stakes in conformity. Sampson and Laub (2005c:36) continued by argu ing that marriage (or employment) should not be viewed as a one particular point in ti me, but viewed as a dynamic and cumulative causal factor that affects desistance from crime over tim e (see also Laub et al., 1998). Given the above presented evidence in support for Sampson and La ubs age-graded theory of informal social control, it is important also to visit the re search providing support for Moffitts taxonomy of offending for explaining sex variat ions in offending and this becomes the focus of the following section. The Saliency of Moffitts Taxonomy across Sex According to Moffitt et al. (2001), females typi cally have fewer risk factors than males; therefore, they are less likely to be early-onset and/or life-c ourse-persistent offenders. In contrast, the sex similarities in adolescent-limite d offending are due to th e process of social mimicry where there is an equal-opportunity for offending among males and females in the adolescent years (Moffitt et al., 2001). Moffitt et al. (2001) have also suggested that while

PAGE 59

59 neurodevelopmental disorders affect the life-cou rse pathways of offending similarly for males and females, males experience early cognitive deficits far more frequently than females. Their results showed that significant sex differences existed in almost all cognitive areas including hyperactivity, impulsivity, and temperament. Tibbetts and Piquero (1999) conducted one of the earliest tests of whether Moffitts taxonomy applies to males and females and report ed that they failed to find enough females who exhibited the life-course-persisten t pattern of offending in order to provide sufficient statistical power in a Philadelphia cohort. Subsequently, Fer gusson et al. (2000) were able to show that a single model could explain antiso cial behavior across sex, but the ratio of involvement in earlyonset offending and late-onset offending was 4:1 and 2:1 for males and females respectively. In contrast, Aguilar et al. (2000) found that the prevalence of early-onset offending was similar across sex. In another study, Moffitt and Caspi (2001) f ound that the sex ratio for life-coursepersistent offending was 10:1, whereas the sex ra tio was virtually negligible for adolescentlimited offending (1.5:1). Similarly, Kratzer an d Hodgins (1999) found a rather large sex ratio for life-course-persistent offending (15:1) and adolescent-limited offending (4:1) among a Swedish cohort of 13,000 individuals. Mazerolle et al. (2000) reported that early-onset offending was associated with persistent offendi ng and a lack of specialization among members of a Philadelphia cohort (n= 3,655). In addition, in a longitudinal study of 820 girls, Cote et al. (2001) did observe an antisocial group charac terized by persistent and frequent offending, although this group only made up 1.4% of the entire sample of females. Comparatively, Silverthorn and Frick (1999) have argued anti social girls tend to share many of the correlates of chil dhood-onset offending and adulthood adjustment problems that the

PAGE 60

60 childhood-onset boys express. The difference is that antisocial girls typically demonstrate a delayed-onset pathway (do not begin their o ffending until adolescence). They propose that this third developmental pathway is unique to females and while the cognitive and neurological deficits and the dysfunctional family environment is present in these gi rls childhood, they do not produce overt antisocial behavi or until adolescence. Silverthorn and Frick (2001) provided a prelimin ary test of their delayed-onset offender typology among 72 incarcerated youth. They found that there were similar percentages of early and late-onset offenders among the males, but that the females were more likely to show the delayed-onset (late-onset) patterns of offending and these girls disp layed similar risk factors as the childhood-onset boys. Subsequently, two addi tional tests found that early-onset and delayedonset offending were similar among boys and girls a nd that the early-onset girls shared the same risk factors as the early-onset boys (see Fergusson & Horwood, 2001; Moffitt & Caspi, 2001). More recently, White and Piquero (2 004) provided another test of sex differences/similarities in early versus delaye d-onset offending, except us ing a large sample of African-American participants from the Philadelphia portion of the National Collaborative Perinatal Project. White and Piquero (2004) discovere d that females were just as likely as males to display early-onset patterns of offending, but comparatively, the late-onset females shared many risk factors with the early-onset males. Noting all of the research presented above, according to recent review, Piquero and Moff itt (2005) argued that females do show less involvement in criminal activities at both leve ls (adolescent-limited of fenders and life-coursepersistent offenders) when compared with males. Taken together, all of the prev iously reviewed research has indicated that (1) macro-level factors can affect individual be havior, (2) there exists a sma ll group of chronic offenders who

PAGE 61

61 commit disproportionate amounts of crime; (3) li fe transitions can al ter individual offending pathways; and (4) Moffitts taxonomy can be used to explain sex differences in offending, particularly the findings that males are more likely to be found among the life-course-persistent offender group than the females. Noting this ev idence, the following section directs attention toward examining the stability of offender groups across recent prior applications of Nagin and Lands group-based modeling techni que, along with a discussion of the factors that have been found to be associated with offending group membership. Prior Applications of GroupBased Models of Offending Similar to the theoretical developments in criminal career and developmental research, modern innovations in met hodological approaches to studying group-based modeling, specifically the technique devel oped by Nagin and Land (1993) (i.e ., what has been referred to by many names including trajectory analysis, finite mixture models, or semiparametric mixed Poisson-based regression models). The tec hnique permits disaggregating aggregate offending patterns into group-based o ffending trajectories of i ndividuals that exhibit distinct pathways over time (Nagin & Land, 1993; Land, McCall, & Na gin, 1996; Land & Nagin, 1996; Nagin, 2005). This method for modeling multiple pathways of offending provides a mechanism to identify and distinguish between certain types of behavioral trajectories. It also enables the use of regression models such as multinomial regression in order to examine which covariates are best able to predict which offending group that an individual is assigned. Nagin and Land (1993) provided the first application of th is methodology based on data from the well known South London cohort of 411 males (West & Farrington, 1973, 1977). Relying on convictions as their measure of offe nding, they found evidence of four groups of offenders including a non-offender group, an adolescent-limited group, and two groups of chronic offenders with different degrees of high-rate offending.

PAGE 62

62 McDermott and Nagin (1998) found support for the existence of three offender groups based an analysis of male respondents from th e National Youth Survey, which is a well-known self-report survey. The first group was classi fied as the non-offenders. The second group was referred to as adolescent-peaked. They tended to resemble the third or chronic offending group in adolescence but then began to desist at th e age of 18 until there was no observed offending at age 24. Additionally, trajectory analyses based on two distinctly di fferent types of offenders (Gluecks Boston delinquents vs. California Yo uth Authority parolees ) from different time periods and contexts found that there seemed to be 4 to 6 trajectory groups that all began to decline by adulthood (see Laub et al., 1998; Piquer o et al., 2002). Laub et al. (1998) and Piquero et al. (2002) also noted that marriage or stakes-in-conformity (that included marriage) was related to lower trajectories. A more comprehensive attempt to identify th e consistency of latent offending groups was performed by DUnger et al. (1998) using da ta from the London cohort (West & Farrington, 1973, 1977), the Second Philadelphia Birth Cohort (W olfgang, Figlio, & Se llin, 1972; Tracy et al., 1990) and the three Racine Birth Cohorts (Shannon, 1988, 1991). They found support for the considerable similarity of offending groups across the data sources, with the constant finding of non-offending, adolescent-limited or peaked offending, and chronic offending groups. Specifically concerning the Racine Birth Cohorts, DUnger et al. s results revealed the similar four classes of offenders noted above for the 1942, 1949 and 1955 cohorts, but discovered an additional fifth group for the 1942 and 1955 cohorts with a late-onset pattern of offending. DUnger et al. (1998) illustrate d that this group show ed relatively stable rates of offending throughout their twenties. Broidy et al. (2003) also found three to four trajectory groups when

PAGE 63

63 examining childhood physical aggression and later involvement in delinquency/crime using data with various demographic groups (males/femal es) and six different study sites represented including: Child Development Project; Christ church Health and Development Study, Dunedin Multidisciplinary Health and Human Deve lopment Study, Montreal Longitudinal Study, Pittsburgh Youth Study and the Quebec Provincial Study. More recent research has also observed th ese similar offending gr oups disaggregated by sex (Cote et al., 2001, 2002; DUnger et al ., 2002; Fergusson & Horwood, 2002; Land, 2000; Piquero et al., 2005), although some studies were unable to identify a chronic offending group among females (DUnger et al., 2002; Piquero et al., 2005). DUnger et al.s (2002) analysis provided five offender groups for the males and only three for the females. Furthermore, they indicated that while the age patt erns of offending tended to be similar across sex, the frequency of offending was higher among the males. In contrast, Cote et al. (2001) discovered a life-course-per sistent group of offenders among the females, but noted that this group only comprised 1.4% of the sample of nearly 900 females. In a more recent study, Cote et al. (2002) found that the broad patterns of development (as far as groups) were similar across sex, but males were more likely to exhibit higher impulsivity trajectories when compared with the females. Relying on data from 896 members of the New Zealand Dunedin Birth Cohort, Fergusson and Horwood (2002) indicated that while identical offending trajectories were found for the males and females, the females were more likely to be of low risk and found predominantly among th e adolescent-limited gr oup, whereas the males were more likely to be in the chronic offendi ng groups. Most recently, Piquero et al. (2005) discovered three classes of of fending (low, medium, and high-rate) among males and only two for the female models (low and medium-rate) us ing data from the Dunedin birth cohort. In

PAGE 64

64 summary, Hipwell et al. (2002) concluded that the main difference between boys and girls behavioral patterns of offending is that only a very small proportion of girls follow the lifecourse-persistent trajectory. Chung et al. (2001) also discovere d another trajectory group that they termed desisters that illustrated an atypical trajectory in that they were characterized by early-onset offending (before age 13) but instead of showing the steady rise in mid-adolescence in to young adulthood, the groups trajectory declined and appeared to de sist by age 21. According to Chung et al. (2001), further analyses suggested that it was not individual and familial f actors that could be attributed to their desistance but rather the influence of less exposure to peer neighborhood, and school factors typical associated with offending. They also indicated that aggressiveness, antisocial peers, and drug availability were consistent predic tors of initial levels of offending measured at age 13 and that individual factors including ag gression, depression, and anxiety significantly distinguished the late-onset offenders from the non-offending group. Piquero (2007) provided a comprehensive revi ew on the studies to date applying Nagins and Lands (1993) group-based modeling technique and concluded that th ere existed relative consistency in the number of trajectory groups and the shape of the curves, but there was variability across the studies depending on the length of follo w-up and the ages that were represented in the analysis. A dditionally, in line with Moffitt these studies have shown evidence of adolescent-limited and life-course-persistent trajectories across dive rse data sets, cohorts, countries and using various measures for offending including self-reports, teacher ratings, parent ratings, and police records based on either contacts, arrests, or convictions. However, these applications have also uncovered other traj ectories not consistent with Moffitts (1993) predictions such as the low-level chronics and late-onset type.

PAGE 65

65 The next section provides an elaboration of the theoretically integrated model using macrolevel and micro-level factors for explaining group-based offending tr ajectories. Specifically, the discussion focuses on how the social disorganiz ation framework can be incorporated with Samspon and Laubs age-graded theory of inform al social control and Moffitts developmental taxonomy. Integrated Theoretical Mode l for Explaining Offending Typically, criminological and sociological research focuses on either examining community level or aggregate rates of delinquency/crime or investigating the differences between delinquents/criminals and nondelinquents/cr iminals. Although, in recent years scholars have begun to recognize the importance of multi-leve l research and integrated theoretical models (see Reiss, 1986; Moffitt, 1997a; Sampson, 1997; Fa rrington, Sampson, & Lauritsen, 1994). In an attempt to answer these calls for research in this area, the proposed model in this study emphasizes the importance of three particular th eoretical perspectives (social disorganization, age-graded theory of informal social cont rol, and Moffitts taxonomy) for explaining groupbased offending trajectories. The proposed mode l is further elaborated on below and the conceptual figure (along with addi tional variables included in the an alyses) is displayed in Figure 3-1. In the 1920s Shaw and McKay reported that the areas that were characterized by the highest rates of crime and delinque ncy also tended to display comp aratively high rates of infant mortality, low birth weight, child abuse, along with other adverse factors for early childhood development. According to Sampson (1997:43) the consistent finding of moderate-strong continuity of early onset offending behavior and later criminal i nvolvement and the longstanding covariation of delinquency rates and social prob lems clearly emphasizes that there is a connection between healthful de velopment of young children and community structure. The

PAGE 66

66 quality of health care services has also been shown to vary across communities with the most socially disorganized areas ha ving the poorest services (Lash, Si gal, & Dudzinski, 1980). Similarly, Moffitt (1997a) and others (see Bronfenbrenner, Moen, & Garbarino, 1984) have argued nueropsychologically vulnerable ch ildren may be disproportionately found in poor areas given the association between brain de ficiencies and povert y. Moffitt (1997a) also highlights the concentration of low birth wei ght, poor infant nutrition, and the incidence of exposure to toxic agents as be ing greater in poor neighborhoods. Furthermore, children with early developmental/neurological deficits are not typically born into protective and supportive environments with in-tact families, but rather into families with pare nts who socialize their children in a criminogenic environment where both pa rent and child are vuln erable to the adverse neighborhood conditions (Moffitt, 1997a). These at-r isk or difficult children are then likely to receive negative reactions from their parents in efforts to cont rol their problem behavior in early development and this negative reaction is likely to be con tinued throughout adolescence among peer groups at school and indivi duals in the community (see Go ldsmith, Bradshaw, and ReiserDanner, 1986) whereby further exacerbating the de velopment/continuity of antisocial behavior. Caspi et al. (1987:308) summarizes this intera ction as one in which the child acts; the environment reacts; and the child reacts back in mutually interl ocking evocative interaction. Thus, in the proposed theoretical model, living in a socially disorgan ized area should impact group-based offending trajectories by increasing the likelihood of being assigned to a chronic offender group. Moffitt (1994) has also suggested that chroni c offending can perhaps be regarded as an exclusively male phenomenon because of the dispr oportionate concentration of risk factors such as neurodevelopmental disorders and early be havior problems among males. Additionally,

PAGE 67

67 Moffitt and others (see Moffitt et al., 1994) have argued that female delinquency tends to be more erratic when compared with male delinqu ency. More recently, Piquero and Moffitt (2005) have contended that males do tend to show more involvement in crime/delinquency regardless of offender group, i.e., adolescent-limited or life-cou rse-persistent. Consequently, it is expected that a greater proportion of males will exhibit both types of offense trajectories and more specifically, being male will be a significantly distinguishing feature of group-based patterns of offending. In contrast, Sampson and Laubs age-graded theo ry of informal social control argues that social bonds or stakes in conformity in adulthood affect desistance from crime. It is expected that individuals who exhibit a virtual non-offe nding trajectory will also display the greatest amount of stakes in conformity since they are likely to be al ready bonded as youth and adolescence since they show no involvement in delinquency. Similarly, this assumed early attachment to institutions such as family and school is likely to continue into adulthood in that they will also be more likely to be married a nd have full-time jobs which will further restrict their involvement in crime. Second, while soci al disorganization is li kely to affect group assignment to an offender group, it is believed that the experience of life transitions such as marriage and employment will increase the likelihood that individuals will express an adolescent-limited/peaked offens e trajectory compared with a chronic offending pattern since these social bonds are theorized to have an eff ect on criminal career desistance. Or in other words, while these groups may have relatively similar patterns of offe nding during adolescence, these life transitions will be related to desist ance among the adolescent-limited/peaked offender group.

PAGE 68

68 In short, recent theoretical and methodologica l advancements in cr iminological research permits me to examine the possible generalizabi lity of formerly discussed offender groups (nonoffenders, adolescent-limited/peaked offenders, a nd chronic offenders) and exploring a potential integration of macro-level and micro-level theories for explai ning offending. These issues along with a description of the methodology and analytic procedures that wi ll be utilized in this current study will be the focal point of the following chapte r (Chapter 4) and the re sults chapter (Chapter 5) will involve a systematic exploration into the several key hypotheses mentioned below. Hypotheses and Expectations Hypothesis #1: According to Moffitts (1993) taxonomy a nd prior applications of trajectory analysis (for review, se e Piquero, 2007), it is e xpected that the semiparametric Poisson-based regression analyses will produce th ree distinct classes of offend ers including: 1) non-offenders; 2) adolescent-limited; an d 3) chronic offenders. Hypothesis #2: In accordance with the theoretical ex tensions provided by Moffitt (1994) and Moffitt et al. (1994), it is expected that sex wi th be a significant covariate for distinguishing group-based trajectories or in other words, males will be more likely to be in offending groups than females. Hypothesis #3: In line with Sampson and Laubs age-graded theory of informal social control, it is hypothesized that individuals who self-report havi ng experienced these key life events (or stakes in conformity) will be less likely to be chronic offenders. Hypothesis #4: It is expected that indi viduals whose neighborhood of origin is characterized by poverty/low neighborhood socioeconomic stat us and residential instability (i.e., social disorganization ) will be more likely to be chronic offenders than adolescent-limited offenders or non-offenders. Hypothesis #5: Macro-level and micro-level effects on the participation and frequency of offending will still be observed net of the effects of each otheror in other words, both macrolevel and micro-level effects will be significantly related to latent class membership. However, the possibility of mediation will also be explored.

PAGE 69

69 Figure 3-1. Integrated Theoretical Model: Social Disorganization, An Age-Graded Theory of Informal Social Control and Moffitts Developmental Taxonomy Social Disorganization Poverty Residential Instability Overcrowding Non-offenders AdolescentLimited/Peaked Offenders High Stakes in Conformity Male Low SES Household Black Moved # of Children Formal Social Control Natural Parents Chronic Offenders Differential Association

PAGE 70

70 CHAPTER 4 DATA AND METHODS Introduction The former two chapters (Chapters 2 and 3) provided the theoretical framework for explaining individual variations in offending and the latter chap ter expressed how macro-level and micro-level factors could affect an individu als pathway of offending and why males can be expected to be disproportiona tely found in chronic offending groups. This information along with the closing section of Chapter 3 that di scussed prior applications of Nagin and Lands (1993) group-based methodology all bear equal importa nce in framing the outline of this current chapter. This chapter begins with a description of the data and sample for which all the analyses of this current study are based. This is followe d by a section highlighti ng the macro-level and micro-level independent variables incorporated in this study to empirically test Sampson and Laubs (1993) age-graded theory of informal social control and Moffitts (1993) taxonomy of adolescent-limited and life-course -persistent offending. The macrolevel measures used in this study were developed with attention to those factors emphasized as having relatively strong effect sizes according to a recent meta analysis by Pratt and Cullen (2005) and the micro-level measures were theoretically-deriv ed and based on prior empirica l research. After the section describing the measures, a discussion of the sp ecific group-based modeling technique (Nagin & Land, 1993) used in this current st udy to identify latent classes of offenders is presented. Next, there is also a discussion devoted to ward fleshing out the recent debate on the usefulness of the technique that has occurred be tween the originators of the methodological tool and some of its most vocal critics. This chapte r will conclude with a description of the software and statistical procedures that will be used in or der to estimate the group trajectories and then to

PAGE 71

71 assess the relative strength of macro-level and micro-levels variables for predicting the likelihood of group membership. Data Sources The data used for these analyses were drawn from the first two birth cohorts collected by Shannon (1988, 1991) in Racine, Wisconsin. Racine was a Midwestern indus trial city with an estimated population of 89,000 residents according to the 1960 U.S. Census. Shannon initially gathered detailed information on three births co horts, the first being 1942, followed by a second cohort in 1949, and the third from the year of 1955. Overall, official police contact data was collected on all of th e individuals from each of the three cohorts from birth up until the end of the follow-up period (i.e., 1974) which yielded sample sizes of 1,352 for the 1942 cohort, 2,099 for the 1949 cohort and 2,676 for the 1955 cohort. In order to create a uniform baseline age of offending across cohorts, this current study only included information on police contacts afte r the age of six. Mo re specifically, Shannon was particularly interested in a dolescents that were defined as c ontinuous residents of Racine through their eighteenth birthdays (regardless of their mobility within the citys limits). This continuous-residence data (which provi des the data for the focus of this study) generated sub-samples of 633 for the 1942 cohor t, 1,297 for the 1949 cohort, and 2,159 from the 1955 cohort. Of these sub-samples, Shannon (1988, 1991) selected random samples for the purposes of collecting in-person interview data for the first tw o cohorts that began on June 1, 1976 (when the 1942 cohort was age 34 and the 1949 cohort was 27 years of age).1 Group-based offending trajectories will be estimated sepa rately for the 1942 and 1949 continuous-resident subsamples presented below. A later section in this chapter entitled Trajectory Estimation 1 According to Shannon (1988, 1991), the lack of continued federal funding restricted him from collecting interview data for the 1955 cohort.

PAGE 72

72 provides a more thorough discussion of the vari ous models and a tabular display can also be found below in Table 4-1. In addition, Table 42 illustrates the different years of observation that can be attributed to each of the data sour ces along with the correspon ding ages of the Racine continuous-residents by cohort. Thus, the final continuous-residence based sa mples previously alluded to involves taking the police contact and interview-level data separately for the 1942 and 1949 residents and merging this individual-level data with the limited available neighborhood-level data based on a unique Census block identifier for each individua l. Once the individu al and neighborhood-level data were merged and sample frequencies we re conducted, it was observed that a number of individuals were missing information on all of the neighborhood-level vari ables of interest. Therefore, in order to create an accurate base line and minimize missing data, we chose only to include the cases that had existi ng 1950 Census block information.2 This selection criterion left a total sample of 235 and 323 individuals fo r the 1942 and 1949 continuous-residents samples respectively. These samples were predominantly White residents (92% and 85%), followed by Black (6% and 11%), and the approximately remaining 3% of each sample were Chicanos. There was an almost even split as far as the sex composition between the males (47% and 53%) and the females (53% and 48%). 2 In order to incorporate change into the models (i.e., such as computing change scores for the Census measures by using 1960 data as well) would have resulted in the loss of more than 60 additional individuals which would have affected the stability of the trajectory models (see Nagin & Tremblay, 2005; Sampson & Laub, 2004). Preliminary comparisons indicated that the 1950 and 1960 block data were relatively similar; therefore, the 1950 Census measures were used with the intention to preserve any additional sample loss and to represent macro-level-level characteristics that existed closer to the early developmen tal years of the Racine residents that Moffitt (1993) has indicated as a crucial time period (see also Gottfredson & Hirschi, 1990). Prior research has shown the adverse effect of living in a low SES neighborhood on factors asso ciated with life-course-persi stent patterns of offending, especially in early childhood (see Brooks-Gunn et al., 1993; Chase-Lansdale et al., 1997; Duncan et al., 1994). The original principal investigator Shannon (1994:7) also indicat ed that if a researcher was interested in the ecological characteristics of which the cohort memb ers spent their early childhood/juvenile years, then the 1950 Census data should be used.

PAGE 73

73 Independent Variables Macro-level independent variables Social Disorganization. Based on the factors highlighted by Shaw and McKay (1942) that are believed to contribute to soci al disorganization and associated with crime (as defined as the breakdown of the formal and informal social networks for regulating neighborhood resident behavior), measures of neighborhood socioeconomi c status (SES) and residential mobility were included.3 Neighborhood SES was measured directly by the average housing values of the residential properties on each block. Prior research has shown the positive effects of living in a high SES neighborhood has on academic achievement, verbal ability scores, and behavioral problems (see Klebanov et al., 1997). Resident ial mobility was indirectly measured by the percent of housing units on a bloc k that were renter-occupied co mpared with those that were owner-occupied.4 Simcha-Fagan and Schwartz (1986: 671) have argued that structural characteristics can affect an individuals ab ility to form social bonds to the community. Individuals who rent as opposed to owning their pr operty are less likely to develop the necessary ties (see Bellair, 1997; Granovetter, 1973) that are important for providing a mechanism of informal social control. Finally, percent overcrowding was Census-based measure indicating the percent of rooms within hous ing units on each block with more than one person per room. Wilson (1987) has also attributed housing dens ity to concentrated poverty, which has been 3 Because of the sparse amount of U.S. Census data availa ble in 1950 and relative homogeneity of Racine, this study was unable to include a measure for ethnic heterogeneity in order to provide a complete measure of social disorganization incorporating all of the processes that lead to crime. 4 The original measure of social disorganization used by Land (2000) was a scal e created by Shannon (1988, 1991, 1994) where a rank of 1-26 was assigned to an individual to represent their neighborhood SES based on the average Census block characteristics of the area in which th ey lived. This measure was actually referred to by Shannon as a natural area code with no reference to social disorganization and given that the measures used (average dollar value of rent, average dollar value of owned property, percent lacking plumbing, percent overcrowding, and percent renter-occupied) did not load on one construct it is likely that Shannons (1988, 1991, 1994) index and Lands (2000) later use of this index is not an entirely accurate proxy measure of social disorganization.

PAGE 74

74 posited as a precursor of social disorganization. Dwelling units that are over-crowded also may lead to sense of anonymity between the residents because of their failure to interact with their neighbors (see Roncek, 1981; Sampson, 1995). The results from principal components analysis confirmed that these three variables were measuring a similar construct (i.e., social disorg anization) across the two cohorts as presented below in Table 4-3; therefore, the corresponding component loadings were utilized in order to construct the appropriate indexes.5 This procedure allows for pres erving the integrity of the data by removing the potential of a partialing fallacy (Gordon, 1968) and reduci ng the potential for multicollinearity which is common in macro-le vel data (Land, McCall, & Cohen, 1990). Thus, positive z-scores were indicative of more socially disorganized areas Micro-level independent variables Stakes in Conformity. Sampson and Laub (2005c) argue that involvement in institutions such as marriage and work provide both short-te rm and long-benefits for individuals. These life transitions can change an individuals motivatio n for crime by re-focusing their attention toward spending time with their spouse and/or investing tim e in their job. These events are believed to foster an individuals development of prosocia l adult bonds. Prior research has shown support for the impact that these transitions can ha ve on desistance from crime (Sampson & Laub, 1993; Warr, 1998) and other research has shown that the e ffect of these life transitions are age-graded (see Uggen, 2000). Therefore, in accordance with Sampson and Laubs (1993) theory and prior research (see Piquero et al., 2002) a stakes in conformity inde x was created for the two local life circumstances 5 This procedure involves multiplying the raw variable scores by the standardized weights obtained as a result of the principal components analyses and then summing these values together to produce a single measure (see Kim & Mueller, 1985).

PAGE 75

75 of marriage and employment. Each of these tran sitions are dummy-coded with 1 indicating if the individual self-reported havi ng experienced the particular life transition by age 25 and 0 indicating if the individual faile d to report having experienced th e specific life transition by age 25.6 Therefore, the index ranged from 0 to 2, w ith 2 indicating that th e individual reported having experienced both of these life circumstances.7 Sex. Sex has long been recognized as one of the strongest predictors of involvement in delinquency and violent crime where crime rates have consistently shown higher prevalence rates of offending for males when compared with females (Rowe et al., 1995). Daly (1998) has also noted that self-repor t studies have repeatedly indicated th at males report a higher prevalence (involvement) and incidence (number of acts) of offending. Wilson and Hernstein (1985) have commented on the strength of the relationship be tween sex and crime over time and cultures, and according to Gottfredson and Hirschi (1990), an ywhere and everywhere males tend to commit more crime than females (with the exception, check fraud, shoplifting, and prostitution). Additionally, Moffitt and some of her colleagues have suggested that life-course-persistent offending may be observed almost exclusiv ely among males because neuropsychological disorders and childhood-onset condu ct problems are more preval ent among males (Rutter, 1983; Singer et al., 1968; Werry, 1986). Female delinque ncy also tends to be more erratic when 6 There were self-reported measures that existed in the da ta for yes/no responses to these life events; however, in order to ensure that the actual life transition occurred with in the observation period of the data available for both cohorts, the experience of these lif e transitions after age 25 for the 1942 cohort were not included. 7 It is important to note up front that this measure is limited in that Sampson and Laub have argued that mere participation in life transitions may not be as important as the quality of the attachment to the various social institutions (i.e., marriage, work, etc.). Laub et al. (1998) also indicate that the inhibiting effects of life transitions are cumulative over time and it takes a while for individu als to express the necessa ry level of cohesion or attachment to affect their behavioral pathway of offending, or affect their desistance from crim e. In the end, according to Piquero et al. (2002) it is likely that just being married or having a job is some kind of indication that individuals have developed some affective ties to their spou ses and/or jobs or else they would not be likely to be involved in the life circumstance to begin with.

PAGE 76

76 compared with male delinquency (Moffitt et al., 1994). Prior research has shown that females typically show less involvemen t in crime in both offender groups (adolescent-limited offenders and life-course-persistent offenders) when compared with males. Recent applications of Nagin and Lands (1993) trajectory analysis have al so shown distinct offender groups among females (see Cote et al., 2001; DUnge r et al., 2002; Fergusson & Horwood, 2002; Land, 2000; Piquero et al, 2005). Provided with th is theoretical framework and building on the prior research mentioned above, a dichotomous variable re presenting the sex of the 1942 and 1949 cohort members was included with males coded as 1 an d females coded as 0 as a covariate in the trajectory models as well as in the multi nomial logit models presented in Chapter 5. Black. Similar to sex, race has also been a long-standing correlate of crime (see Hawkins, 1995) and Moffitt (1994) has argued that the hi gher prevalence of both adolescentlimited and life-course-persistent offenders among Blacks may account for the racial variation in crime rates because the theoretical origins of the latter patter n of offending disproportionately affect Blacks because of the strong association between race and poverty. Prior research has shown a higher prevalence of lif e-course-persistent offenders among Blacks (Piquero & Buka, 2002), and other researchers have argued that the risk factors associated with life-coursepersistent offending typically tend to be disp roportionately concentr ated among Blacks (see Piquero & Moffitt, 2004; Piquero et al., 2005). Furthermore, macro-level research has indica ted that Blacks were more likely than Whites and Hispanics to live in poor neighborhoods and be involved in more serious forms of criminality. Prior research has also shown that Blacks who reside in these areas are disproportionately affected by th e burden of child-care and resi dential instability and these factors were strong predictors of delinquency an d violence (for review, see Gephart, 1997).

PAGE 77

77 Additional research utilizing trajectory analysis has also noted the increas ed likelihood of Blacks in chronic offending groups (see Land, 2000); therefor e, it was necessary to include race. The original nominal level variab le indicating the race of the 1942 and 1949 cohort members was recoded into a dummy variable w ith Blacks being coded as 1. Low SES Household. This prestige-based measure will be included in order to provide an individual-level SES complement to the macr o-level measures of neighborhood SES. Prior research has shown that i ndividuals who are from low SES households in low SES neighborhoods tend to have the hi ghest rates of offending and comparatively those from the high SES households in high SES communities have the lowest offending rates (Braithwaite, 1979; Reiss & Rhodes, 1961; Wikstrom, 1991a). Usin g trajectory analysis, McDermott and Nagin (1998; see also Land, 2000) have also shown that low SES is an important factor for predicting group membership, more specifically the likelihood of being classified as a chronic offender. Additionally, Brooks-Gunn and Leve nthal (2000) have noted the importance of including this family-level measure when estimating neighborhood effects. This variable was measured as a scale i ndicating the prestige of the head of the households occupation in the family in which the individual grew up. It ranged from a low of one to a high of nine with the following occ upational types represente d: (1) professional, technical, managerial; (2) clerical and sales; (3) craftsman, foreman; (4) operative; (5) maintenance and service; (6) priv ate household labor; (7) industrial labor; (8) farm, agricultural labor; and (9) unemployed. Consistent with the direction of the macro-level social disorganization index, higher valu es indicated worse situations (i.e., growing up in lower SES households).

PAGE 78

78 Family Stability. This measure was used to represen t the stability of the family dynamic in which the cohort member was socialized. Most research has shown an established relationship between family stability and delin quency/crime (Glueck & Glueck, 1950; McCord & McCord, 1959; Goetting, 1994) and crime rates (Sampson, 1987). Studies have found that adolescents living with both bi ological parents have lower rates of offending when compared with those who reside in broken homes or recons tituted families. According to Gottfredson and Hirschi (1990:104), single-parentheaded households (usually by a woman) may be unable to provide the same amount of supervision because of the time constraints imposed on them and thus may be more likely to have negative res ponses to their childrens unwanted behaviors. Furthermore, this variable along with the followi ng measure listed below (family structure) have been identified as key family-level variable s when estimating neighborhood effects (see BrooksGunn & Leventhal, 2000) Originally, this variable was based on a scale created by Shannon (1988, 1991) which combined responses to several items regarding family composition ranging from 1 to 20, where according to Shannon (1988, 1991), 1 represented the most stable family dynamic (two parents present during the entire childhood /juvenile period) and 20 the leas t stable family dynamic (lived only with grandparents, never with parents). For the purposes of this stud y and due to the little variability among the categories, the scale was collapsed into a dichotomous variable with 1 indicating the presence of two parents during the entire ch ildhood/juvenile period and 0 indicating any other type of less stable family dynamic. Family Structure. In order to account for the possi ble differential opportunities of parental supervision affecting individuals offending pathways (Gottfredson & Hirschi, 1990) along with having a complementary individual meas ure of household density to the macro-level

PAGE 79

79 household density measure (percent overcrowding) a second measure of family dynamics was included to represent family structure. This wa s a measure of the number of children within each of the cohort members family (including the cohort member).8 Differential Association. Differential association theory (Sutherland, 1939, 1947) has long stressed criminal and deviant behavior is learned through exposure to and acceptance of definitions favorable to crime. Burgess and Ak ers (1966), and in later reformulations by Akers (1977), expanded on this idea by specifying the pr ocess in which learning occurs. Favorable definitions of crime and deviance are learned in a social setting where the imitation of criminal and deviant behavior also occurs Delinquency, like any behavior, ge nerates either beneficial or adverse consequences positive or negative reinforcem ent. If delinquent or criminal behavior is reinforced either socially or nonsocially, the behavior is likely repeated in the presence of cues to which it has been paired. Or in other words, Akers (1998:50) argues th at the probability that individuals will engage in deviant behavior is in creased when they differentially associate with others who commit the behavior and express favorable definitions toward it. Thus, based on research that has produced strong evidence for wh at is now termed the perceived delinquent peer effect (Felson & Haynie, 2002; Matsueda & Anderson, 1998; Schr eck, Fisher, & Miller, 2003; Warr, 1993; Warr & Stafford, 1991), a measur e was included with 1 indicating that the 8 Overall, there was very little missing data at the individual-level (less than 1% for any of the variables used in the analyses). The only two variables that had missing data were family stability and family structure. In order to correct for the missing values, th e method of group specific means were computed. In order to do this procedure the sample was sorted into five categories: 1) White males; 2) White females; 3) Black males; 4) Black females; and 5) Chicano males (there were not any Chicano females in the sample). For the few cases (n=4) where a missing value was recorded, the sex/race specif ic group means were imputed.

PAGE 80

80 individual self-reported peer influence as th e strongest contributing factor toward their delinquent/criminal behavior in junior high and high school.9 Formal Social Control. This measure was a scale of the individuals perception of the level of police patrol in their neighborhood during their junior high and high school years.10 This indicator of institutionalized formal social control ranged from 0 repres enting not at all or no police patrol, 1 for lightly patrolled, 2 for moderately pa trolled, to a high of 3 which indicated that the individual perceived their neighborhood as be ing heavily patrolled during their early to late adolescent years.11 Moved. In order to provide a contextual measure of residential mobility to complement the compositional measure included in the social disorganization index, th is variable was coded as 1 if the individual had moved within Racine during th eir juvenile years. It is likely that moving, particularly during the adolescent years, is likely to have an adverse effect on an 9 It is worth mentioning here that this is certainly a less than ideal measure of differentia l association and more direct measure would have been prefer able; however, due to the limitations of data this measure was included as a control variable. 10 It is important to note that this measure is a weak measur e of formal social control an d that direct measures such as prior conviction or incarceration are optimal; however, there was only two items that allowed the individuals to report having received jail time as a result from their polic e contact and these questions were only in reference to their first and second police contact. Because only 1 to 4 in dividuals across the cohorts actually reported this as a response I made the determination to use the police patrol measure because there was a sufficient amount of variation in responses to this latter survey item. Additionally, this measure is also limited due to the fact that it was asking individuals to retrospectively answer this question based on their perceptions of how they felt as a middle and high-school juvenile (which is at least 10 years prior and for the 1949 cohort and 20 years prior to the date of the interviews in 1974 for the 1942 cohort) and it is also likely that those individuals who actually came into contact with the police may have had a bias toward reporting higher levels of neighborhood patrol. Future studies would benefit from incorporating more direct measures of formal social control such as those discussed by Sampson and Laub (1997) and De Li (1999), which relied on convictions and on court appearances and convictions, respectively. 11 Shannon (1988, 1991, 1994) originally developed a scale to assess formal social control, but this measure had a series of flaws, with the most serious being that the scal e was constructed by summing self-reported responses of the cohort members perceptions of the (1) level of police patrol when they were in junior high and high school (1=heavily, 2=moderately, 3=lightly; 4=not at all); (2) their and a couple of their friends views toward the police in junior and senior high, (1=positive, 2=negative, 3=indifferent) ; and (3) to report whether or not they got into trouble with the police with a couple of their friends (yes/no). Even more oddly, this scale was constructed by coding an individuals response as 1 if they had either negative or indifferent views toward the police, adding 2 points if the individual reported that they and their friends got in trouble with the police, and adding points if the individual also responded that they perceived police patrol was either heavy or moderate during their juvenile years.

PAGE 81

81 individuals pattern of offending by disrupting the ability to form /maintain prosocial bonds with peers and their schools, which ar e important factors for inhibiti ng delinquency (see Hirschi, 1969; Akers, 1985).12 Dependent Variables Delinquent/Criminal Involvement. The variable used to measure offending was the number of official police contacts for each individual from age 6 to age 25.13 The police contact measure was limited to those instances in which an individual came into contact with a Racine police officer specifically for a crime that he/she was suspected to have committed or did in fact commit. In addition, the official police contacts were also restrict ed to contacts that had true crime codes with a felony/misdemeanor distincti on as well as contacts for status offenses to measure offending (i.e., excluding tr affic offenses, and city ordinance violations) because (1) they are a more accurate official measure of t rue offending when compared with arrests and convictions; (2) they are closer in time to the actual criminal ac t than self-reported offending (see Hindelang et al., 1981; Thornberry & Krohn, 2003); a nd (3) the inclusion of status offenses increases the potential for identifying a dolescent-limited offenders (see Moffitt, 1993).14 Offending Group Membership. The results of the semiparametric mixed Poisson-based regression models will guide th e identification of the number of offending groups that exist within these data and the shapes of their offending trajectories. The identified latent classes of offenders will operate as the dependent vari able for the estimation of the multinomial 12 It is important to note that this measure is limited in that is was impossible to discern from the available data if the individual moved from a social disorganized area to a more or less disorganized area; therefore, it is important to interpret the findings with caution and keep in mind that the expectation is that moving is seen as having an adverse effect on offending patterns within this study. 13 This police contact data refers to the years that are consistently available for the 1942 and 1949 continuousresidents or up to and including age 25. 14 Prior trajectory analyses with Shannons (1988, 1991) Ra cine cohorts relied solely on felonies and misdemeanors (DUnger et al., 1998) or everything including tra ffic offenses and suspicion contacts (Land, 2000).

PAGE 82

82 (polytomous) regression models (Models 1 and 2) with the non-offender group omitted in order to serve as the reference category. The followi ng sections describe the group-based modeling technique designed by Nagin and Land (1993) that wi ll be utilized in this study for identifying the offender groups along with a discussion of th e existing debate that has been ongoing related to the usefulness of the methodology. Study Limitations There are a series of limitations of the data source itself and th e measures used in this study that are worth noting. First, the timeliness of th e data is an issue sin ce the cohorts themselves were born in the 1940s, socialized as children and/or adolescents in the 1950s and 1960s. Therefore, it is likely that these samples may not be generalizable to other cohorts from more recent time periods because they were born during WWII (e.g., 1942 cohort) or are from the baby-boomer generation (e.g., 1949 cohort) who grew up in the post-WWII era. Second, Racine, WI was a Midwestern industrial city with an estimated populati on of 89,000 residents, and it is not likely that the re sults from this study would apply to the larger metro politan urban cities of today (i.e., cities with populat ions greater than 100,000). Third, particularly during the time of data collection, Racine was a relatively homogenous city with limited numbers of minor ities, and according to Shannon (1988), although the overall population did rise to 95,000 persons by 1970, this was followed by a decline to 86,000 persons by 1980. He attributed this popula tion decline to a growing number of young people in need or work and increased female part icipation in the work force, but this was met with a decreased number of individuals leaving for retirement. Taken together, the results discovered here may not necessarily apply to ar eas with larger minority representation or dissimilar population shifts.

PAGE 83

83 Noting the limitations of the cohort members a nd the location, there is also a series of limitations with the measures used in this analysis First and foremost, the data available in the 1950 Census at the block level is very limited or in other words, there was not the typical long list of measures accessible to macro-level resear chers today for this research, and there also was a noticeable amount of missing 1950 census data for the 1942 and 1949 continuous-residents. With this is mind, the social disorganization me asure used here is relatively weak in the sense that it is only measuring two (neighborhood SES and residential mobility) of the three elements that Shaw and McKay (1942) highlight ed as important factors associated with the breakdown in informal social control and crime rates. Once again, Racine was a relatively homogeneous area, so it is likely that racial dissimilarity would be difficult to directly observe at the block level. Additionally, Sampson et al. (1997) have argued that what matters more than social disorganization is collective efficacy or wh at happens at the individual level as far as developing community ties and informal social cont rol. There were not an available individuallevel data that would allow for an investigati on of this hypothesis, so caution should be noted that these individual-level processes are not accounted for in this study. In addition, as footnoted earlier the stakes in conformity index used to assess Sampson and Laubs (1993) age-graded informal social contro l theory is limited in that it is a dichotomous index with a short range (0 to 2). More re cently, Laub et al. (1998) have argued that mere participation in local life circumstances may not be as important as the quality of the attachment and that the inhibiting effects of life transitions on desistance are cumulative and gradual over time. Therefore, while this measure does not captu re the level of attachment or cohesiveness of the relationship, it is likely that just being married or having a full-time job is indicates that some degree of attachment has already been formed or else the life transition would probably not have

PAGE 84

84 occurred in the first place (see Piquero et al., 200 2). The differential association measure is also noticeably weak. Regarding the dependent variable (police contacts), there ar e also well known strengths and weaknesses from using either official data (p olice contact, arrest, co nviction) or self-report data to measure criminal behavior. Under-reporti ng is one of the principal limitations of official statistics because the majority of police activity is a direct result of the wi llingness of citizens to report crimes (see Hindelang et al ., 1981; Thornberry & Krohn, 2003). However, there are also extensive disparities in victim reporting based on various factor s including: requirements for insurance companies, confidence or lack of confidence in police de partments, etc. Furthermore, there are many offenders who will never come to the attention of law enforcement, such as shoplifters, burglars, or white-c ollar criminals (Junger-T as & Marshall, 1999). This result might be due to the fact that more skillful and shrewd offenders are less likely of being arrested and/or may reflect the differences in pol ice discretion and priorities acro ss regions/agencies (Blumstein & Cohen, 1987). One key advantage of self-report measures is that they often offer more comprehensive information than official data. They pr ovide background and soci odemographic data on individual offenders, as well as personal characteristics. Research has indicated th at self-reports are effective in answering questi ons of causality, motivation, cha nge, and stability in criminal behavior (Junger-Tas & Marshall, 1999:296). Self-reports also tend to include both a broader range of deviant behavior and less serious types of crimes than official police records indicate (Blumstein, Cohen, Roth & Visher, 1986:32). Some examples of the types of crimes that are likely to be found in self-reports but not in offi cial data are prostitution, drug abuse, public order crimes, and drunk driving.

PAGE 85

85 Notwithstanding the mentioned benefits, there are several limitations to the self-report method. Under-reporting and poor recall of crimin al activity are disadvantages because selfreports tend to be retrospective in nature (as was the case with the interviews in these data). In addition, some respondents may even inflate their actual level of criminal activity. Cross-study comparisons of self-reports may also be hindere d due to differences in the wording of the questions, differences in the administration of the survey, or differences in time periods (Blumstein et al., 1986:33). Taki ng into account the noted limitati ons of police records and selfreported measures of offending, this study chose to measure police contacts because they are a (1) more accurate official measure of true offending than arrests and convictions, and (2) typically closer in time to the actual criminal act than self-reported offending which sometimes asks individuals to recall their pa st behavior in their lifetime, last twelve months, last six months, etc. (see Hindelang et al., 1981; Thornberry & Krohn, 2003). Within the context of all of the above mentioned limitations surrounding the data, time period, samples, and the location along with the measurement shortcomings, the following section highlights the group-ba sed methodology that will be used to estimate the group trajectories. Methodology Group-Based Modeling Longitudinal research in its most basic form (as applied to the study of crime), involves the identification of a cohort at birth and following th at group of individuals throughout their life, or at least long enough to capture the termination of the majority of criminal careers (Blumstein et al., 1986:104). The longitudinal method allows for a specified cohorts level of involvement in crime to be traced as the cohort, itself, matures (Greenberg, 1985). This knowledge is crucial for detecting causal factors or sequen ces that associate individual char acteristics or certain events to

PAGE 86

86 consequential offending (Blumstein et al., 1988). Longitudinal data also permits the use of complex behavioral models because they allo w one to examine the effects of missing or unobserved variables that may be correlated w ith crime (i.e., also known as the unobserved heterogeneity problem). It is often the case that even among people who have similar characteristics (race, age, sex) there is a s ubstantial amount of heterogeneity. Thus it is important to utilize longitudina l models or run the risk of using misspecified models and producing inconsistent paramete r estimates (see Nagin, 2005). According to Nagin (2005:1-2), longitudinal data also provide researchers with the basis to study developmental trajectories. Most of the common methods for studying developmental trajectories, however, are those th at are calculated to account for individual variability, which masks the possibility of observing meaningful su bgroups within certain samples and populations. Furthermore, this over-reliance on traditional longitudinal approaches has led to much ad-hoc theorizing and classifying individuals on the basi s of what could be random variation. Hence important developmentally distinct groups could be missed or misclassified. The particular method used in this study to model the group-based offending trajectories was originally developed by Nagin and Land (1 993) and is known by several different names including: group-based modeling, fi nite mixture models; trajectory analysis, and semiparametric Poisson-based regression analysis. It is impor tant to note that this technique shares some similarities with the traditional group-based mo deling approaches such as hierarchical linear modeling (Byrk & Raudenbush, 1987, 1992; Goldst ein, 1995) and latent curve analysis (McArdle & Epstein, 1987; Me redith & Tisak, 1990). Still, there are distinct differences. For in stance, both hierarchical modeling and latent curve analysis assume continuous distribution fu nctions. Trajectory analysis does not assume

PAGE 87

87 this continuous distribution, but ra ther assumes that there exist di stinct clusters of group-based trajectories that exhibit differe nt etiological processes, and the methodology approximates the continuous distribution by identify ing points of support (i.e., dis tinct groups or trajectories) (Nagin, 2005). These trajectory models are also based on a maximum likelihood (ML) function, which of course permits them to display the qualities of ML parame ter estimatesthey are consistent and asymptotically normally distri buted (Cramer, 1946; Greene, 1990). Given the assumptions and statistical prope rties, this method assumes that individual differences in behavioral trajectories can be described as pol ynomial functions of age or time. The following formula presents the probability fo r belonging to a particular group: J P (Yi) = j Pj (Yi) j where P(Yi) is the unconditional probability of observing individual i s longitudinal sequence of police contacts, Yi, over time. It is equivalent to the sum across J groups of the probability of Yi given i s membership in group j weighted by the probability of membership in group j (see Nagin, 2005:25). There are various forms and derivatives of th e trajectory model, but the two most common parametric forms that are likely to apply to th e data used for this study are the censored normal model (CNORM) and the zero-inflated Poisson m odel (ZIP). The CNORM model is appropriate for data where the outcome variable of interest is normally distributed bu t there exists clustering of the data at either the minimum or maximum (which typically occurs with offending data). This model can also be applied to data wit hout censoring by assigning the scale maximum and minimum values to values that are respectively sm aller and larger than any value observed in the data. The equations listed below correspond to the statis tical derivations of the cases when (1)

PAGE 88

88 the data are censored at the scale minimum; (2 ) normally distributed between the scale minimum and maximum; and (3) the data are censored at the scale maximum: pj ( yit = S min) = ( S min j xit) / pj ( yit) = 1 / ( yit j xit / ) for S min yit S max, and pj ( yit = S max) = 1 ( S max j xit) / where it is assumed that yit is normally distributed with a mean of j Xit and conditional on age ( t ) with a standard deviation of and represents the cumulative di stribution of a normal random variable. In addition, S min represents the scale maximum and S max represents the scale minimum (Nagin, 2005:30). In contrast, the parametric form of the trajectory model that is needed for modeling countbased data such as the number of police contacts is the ZIP vers ion, which was first presented in Nagin and Lands original work and is based on the generalization of what Lambert (1992) called the zero-inflated Poisson (ZIP) distributi on. Count-based models have been more frequently utilized in the social sciences in an effort to correct for th e inefficient and biased estimates that are produced when linear regres sion is employed for modeling data where the dependent variable is measured as a count or frequency of an event (for review, see Long, 1997). Although the Poisson regression model is the most basic form of the count-based models, it is often not the most suitable model due to its prim ary assumption of mean variance equivalence, or equidispersion. It is more like ly the case that, in practice a nd especially in criminological research, the distribution of the outcome variable is overdispersed, (i.e., the variance exceeds the mean); therefore, a more appropriate mode ling procedure is recommended (Long, 1997:218).

PAGE 89

89 The ZIP model allows for the assumption of a dual-state-system. A dual-state-system signifies two situations: one in a zero-contact state where the assumption is that a group of individuals exists that can be regarded as police-c ontact free, and the s econd state or non-zero contact state where the expected incidences of police contacts are assumed to follow some known distribution, such as the Poisson (Long, 1997; Lambert, 1992). Several other studies have also noted the practicality of the ZIP model wh en using count data with excess zeros (King, 1989; Land, McCall, & Nagi n, 1996; Zorn, 1998). The ZIP model takes the following form where the first term is the probability of no police contacts and the second term is the probability of the combined events being active and the probability of zero police contacts given jt > 0 (Nagin, 2005:35). j yit pj ( yit) = (1 t ) ( jt e jit / yit!) ( yit = 1, 2,.), Once the parametric form of the trajectory model is established it is necessary to determine the functional form of the groups trajectories and there are three possibil ities including a linear, quadratic, and cubic form (or age, age-squared, and age-cubed). Depending on the ages and the length of the observation period, it is unlikely that individual o ffending will have a linear form. Similarly, although the quadratic model is more often used in research (see Bushway et al., 2003:140), it only permits one major directional cha nge in the trajectory, and it is likely that cohort members will and have undergone periods of intermittency in their involvement with the police. Intermittency periods refer to those point s in time (i.e., years) where no official police contacts were recorded for any pa rticular cohort members. Theref ore, it is likely that the cubic model will demonstrate the best functional form of the trajectories. The CNORM and the ZIP equations that allow for all three of these possible functional forms are presented below: yit = 0 j + 1 jAgeit 1 + 2 jAgeit 2 + 3 jAgeit 3 + it (CNORM),

PAGE 90

90 ln ( t j ) = 0 j + 1 jAgeit 1 + 2 jAgeit 2 + 3 jAgeit 3 + it (ZIP), where Age1 it, Age2 it, and Age3 it are the cohort members ( i s) age, age-squared, and agecubed respectively, t represents time, and it is a disturbance parameter that is assumed to be normally distributed and the s are the parameters that determine the functional form of the trajectory (Nagin, 2005:28-33). Finally, the exact number of groups is chosen based on an examination of the Bayesian Information Criteria (BIC) in orde r to maximize model fit. This process is necessary because conventional likelihood ratio tests cannot be used to assess whether the addition of another group adds any explanatory power to the model (see DUnger et al., 1998). In the end, the exact determination of the number of groups is based on the ability of the model to find a relative convergence using all of the available informati on and by the utility of the groups for indicating distinct and homogenous trajector ies, while also having a suffici ent number of persons within each group to allow for further sta tistical analyses (Bus hway et al., 2003:140; see also Brame et al., 2001). The BIC values can be relied on to determine the numb er of groups in both nested and unnested models (see DUnger et al., 1998; Nagin, 1999), and the BIC values are estimated based on the following equation, where L is the maximum likelihood, n is the sample size, and k is the number of parameters (Nagin, 2005:64): BIC = -2 log ( L ) + log ( n ) k A Seductive or Useful Statistical Method? More than thirty years has past since Hirsch i (1969:53) initially pr ovided the word of caution to criminologists when using offender typol ogies by asserting that it begs the question of causal homogeneity by focusing exclusively on the question of behavioral homogeneity. Criminologists have come a long way since Hirschis earlier remarks regarding offender

PAGE 91

91 typologies; however, this does not mean that Nagi ns and Lands latest methodological technique has been insulated from theoretical scrutiny. The group-based modeling technique has been the source of recent criticisms and counterpoints between its origin ators and two prominent lifecourse theorists, Sampson and Laub. This debate began with the findings from an earlier study by Sampson and Laub (2003) questioning the stability of offendi ng patterns over the life course a nd more specifically with an article published in the Journal of Quantitative Criminology entitled Methodological sensitivities to latent cl ass analysis of long-term criminal tr ajectories. The authors (Eggleston, Laub, & Sampson, 2004) argued that the enthusiasm for the use of Nagin and Lands (1993) semiparametric group-based models has not b een equally supplemented with robustness and sensitivity analyses that are ofte n inherent in offending data. They proceed by summarizing the obvious advant ages of these models (see also Nagin, 2005) that include (1) the fact that they are desi gned for longitudinal data ; (2) they lead to the ability to distinguish groups of offenders that ar e homogenous within their trajectory but distinct from other trajectories; (3) th ey provide an objective system atic criterion for establishing offending groups by using the Bayesian Informa tion Criterion (BIC) rath er than presupposing subjective sub-groups of offenders; (4) posterior probabilities are used to assess the precision of group assignment; and (5) this method accommodates missing data, whereby offenders that have unavailable information for several time periods ar e not excluded from the analysis (Eggleston et al., 2004:2). Relying on 500 juvenile delinquent s data from the archives of the classic Gluecks Unraveling Juvenile Delinquency research (1950, 1968), Egglest on et al. (2004:23) concluded that the length of follow-up can affect the trajecto ry patterns observed and argue that data that

PAGE 92

92 focuses on say age 6 to 15 versus 10 to 32 may pr oduce trajectories with different curves (i.e., shapes), peak ages of offending, and possibly membership to different offending groups. Additionally, they argue that research using trajectory an alysis needs to attend to both incarceration and mortality information because these factors can also affect offending trajectories (see Piquero et al., 2001 for a similar discussion on the importance of accounting for incarceration data). This article discussed above prompted an initial response by Nagin (2004) and a subsequent response by the Sampson, Laub, and Eggl eston (2004) that were both printed in the same issue of the Journal of Quantitative Criminology Nagins (2004) initia l response indicated that he had no criticisms regarding the anal ysis conducted by Eggles ton et al. (2004) and typically agreed with their fi ndings; however, he did note some counterpoints to a few of their conclusions. Specifically, Nagin (2004:28) argu ed that the criticisms offe red by Eggleston et al. aimed at the trajectory model are limitations that plague all longitudinal data (i .e., projecting beyond the range of the data and missing data resulting from incarceration and/or deat h). Other researchers have also commented on the idea of career termination or false desistance resulting from death (see Reiss, 1989) or based on data that does not track the individua ls throughout their entire life course (see Bushway et al., 2001; Piquero et al., 2003). Nagin concluded by reminding researchers that the groups that result based on the tr ajectory models must be understood as an approximation and this warning was already previously presen ted in Nagin and Lands (1993) original work. Sampson et al. (2004) reanalyzed the Gluecks data in response to Nagins comments and subsequently provided several more suggestions for future resear chers relying on the trajectory

PAGE 93

93 method. They also felt that the majority of thei r criticisms should also be heeded by any future criminological research incorporating longitudinal data. First, using random samples of varying sizes, they found that the number of groups only began to stabilize with samples of over 200. Other research has indicated that the number of identified groups is partly a function of sample size (see Nagin & Tremblay, 2001) a nd that the number of groups te nds to be consistent with samples above this range (s ee DUnger et al., 1998). In their final remarks, Sampson et al. (2004:41) indicate that the larg er problem with the use of trajectory analysis and inferences ma de from longitudinal data with short follow-up periods concerns the causal theory for using groups. They state that future research should resort to grounding all methods in strong theory and be wary of models that provide evidence of an existence of groups just because th ey are statistically discovered or what is referred to as the reification of groups (see also Laub & Sampson, 1993; Sampson & Laub, 1993). As if this above exchange was not enough to settle the debate and prompt future longitudinal researchers to be mindful to theo ry and consider the estimation of groups in trajectory analysis as approximate, this dialogu e was continued most recently in a 2005 issue in the journal of Criminology Nagin and Tremblay (2005a:873) set forth with an attempt to debunk three commonly held misconceptions regardi ng trajectory analysis that originate from researchers failure to understa nd that the method results in an approximation of groups including: (1) the individuals actually belong to a trajectory group; (2) the number of trajectory groups is immutable; and (3) th e trajectories of group members fo llow the group-level trajectory in lock step. Regarding the first misconception, Nagin and Tr emblay (2005a:882) state that trajectory groups are referred to as is they are real/t rue entities only for the purpose of literary

PAGE 94

94 convenience in that it facilita tes the ability to clearly discu ss groupings of individuals that display similar offending patterns. Since the di scovering of trajectory groups is a statistical approximation, they are not meant to be interpre ted as factual representations of reality (a common limitation of all statistica l models). Similarly, this fals e understanding of the fact that the method results in a statistical approximation of groups also has fostered the development of the second misconception, i.e., that the number of groups in the sample is immutable. According to Nagin and Tremblay (2005a: 888), the number of groups and the shapes of their offending curves are not a fixed reality e ither. Instead, the number of groups and their respective trajectories ar e influenced by the quality of the long itudinal data that are relied on for the analysis. In this sense, the quality of the da ta refers to the sample size and the number of observation periods, and obviously larger samples and longer follow-up periods are always the most optimal. The research presented above al so showed that the number of groups and the shapes of their trajectories were relatively stable in samples that were larger than 200 cases (DUnger et al., 1998; Sampson et al., 2004). Nagin and Tremblay (2005a:892) concluded by offering a couple of comments concerning the final misconception (that indi viduals will follow their offendi ng groups trajectory in lock step). First, the shape of the groups trajecto ry within which the individual is assigned is intended to reflect a behavioral process over time versus possible individual variation across this identified pattern of offending. There will always be individual variation, just as there is variation in offending if a rese archer relies on other units of analysis (households, cities, counties, etc). It is highly unlikely that any one particular member of any these aggregations will exactly display the average offending rates of th e group. In addition, ther e are also statistical

PAGE 95

95 reasons why individuals within a group trajectory will deviate from the process captured over the entire period of observation, most not ably involving sampling variation. Notwithstanding this further attempt by Nagin and Tremblay to defend the methodology and to caution its users about the limitations of trajectory analysis this article was inevitably met with a rejoinder by Sampson and Laub (2005a). In a departure from the previous dialogue of the debate as depicted in the Journal of Quantitative Criminology this article had a much more forthright tone. Sampson and Laub (2005a: 905) basically conclude d by suggesting that trajectory analysis is just anothe r sexy statistic al tool much like path and factor analysis long ago, followed by event history analysis in the 1980s, and most recently preceded by LISREL and HLM in the 1990s. Although, they did argue that th e methodology has some uses, they believe that the larger problem is criminologys obsessi on with statistical tools without attention to theoretical underlying processes that influence behavior. In a last attempt by Nagin and Tremblay to provide a response to Sampson and Laubs criti ques, Nagin and Tremblay (2005b) themselves were surprised by the tone of the latest rejoin der, and equally cautioned criminologists to be wary of seductive theories as well as se ductive methodologies (espec ially when the critics are promoting a life-course theory th at they themselves developed). In another work, Sampson and Laub (2005b:41) did concede that they were not arguing that grouping techniques such as Nagins did no t have any place in criminology, they were just trying to warn criminologists about the mistake of the reification of gr oups as a result of the application of the method, which Nagin (2005) hi mself also mentioned as a possible outcome. They argued that this mistake is comparable to researchers making errors in interpreting

PAGE 96

96 regression results in causal versus associational terms (for furthe r discussions also see Sampson & Laub, 2005b, 2005c; Maughan, 2005; Nagin & Tr emblay, 2005c, 2005d; Raudenbush, 2005). With these criticisms and limitations in mind, this chapter concludes with identifying the software that will be used to estimate th e group-based offending trajectories. The final discussion is on the use of multinomial regression for assessing the relative strength of macrolevel versus micro-level predictors fo r predicting latent class membership. Analytic Procedure Trajectory estimation PROC TRAJ is the name of the computer soft ware that will be relied on in order to estimate the offending trajectories for the 1942 and 1949 Racine continuo us-resident samples by cohort and data source including sex as covariate. This is a special procedure that is obtainable as a macro-level function in SAS (Jones et al., 2001) and is made available for free use by the National Consortium on Violence Research. Alth ough a relatively new procedure, PROC TRAJ has become a widely used method for investigat ing trajectories of development (Piquero, 2007) and has been applied to a variet y of different research questi ons including criminal career desistance (Bushway et al., 2003), offense trajec tories (Chung et al., 200 2; Fergusson et al., 2000; Piquero et al., 2003), sex di fferences in mental health (B arrett & White, 2002), age-crime trajectories (McDermott & Nagin, 2002), and aggre ssion toward mothers (Pagani et al., 2004). While there have been two previous studies th at have used PROC TRAJ with macro-level data, they relied solely on aggregate crime rate s in Chicago census tracts (Griffiths & Chavez, 2004) or neighborhood street segments in Seattle (Weisburd et al., 2004). Therefore, this is possibly the first attempt to date (see Land, 2000 fo r a possible exception) that has utilized this method to estimate offending traj ectories and then estimate mode ls incorporating macro-level and micro-level characteristics for predicting group membership.

PAGE 97

97 The initial trajectory models will be es timated separately for the 1942 and 1949 Racine continuous-residents including sex as a covariate in the trajecto ry estimation (Models 1 and 2).15 This will be followed by two models estimated si milarly by cohort with sex as a covariate but only using the random subsamples of intervie wed Racine continuous-residents from the 1942 and 1949 cohorts (Models 3 and 4). These models will allow a preliminar y examination of the stability of offender groups across the cohorts and by data source. In other words, they offer a type of sensitivity analysis for the stability of group identification th at Nagin and Tremblay (2005a) have argued has rarely been done in the tr ajectory literature (for exceptions see DUnger et al., 1998; Sampson et al., 2004). Finally, because the purpose of this study is to examine the importance of macro-level and micro-level factors in pr edicting longitudinal trajector ies of offending, the 1942 and 1949 subsamples will be merged respectively with the sparse 1950 Census block data and an additional model will be estimated for each cohort with sex as a covariate (Models 5 and 6). These final trajectory models with the 1942 and the 1949 continuous-resident subsamples (merged with the interview and block data) are the basis for a ssessing the possible independent effects of macro-level a nd micro-level measures for predicting latent class membership (see also Table 4-1 presented above that graphically disp lays the trajectory analyses that will be estimated). Multinomial logistic regression After identifying the latent offender classes and establishing their re spective participation and frequency rates of offending, multinomial logi stic regression models will be estimated in 15 Ideally, race-specific models would also be estimated al so; however, due to the hom ogeneity of Racine in the 1940s and the small numbers of non-White s in the continuous-resident samples, race will remain a covariate in the multinomial regression models estimated in the following chapter.

PAGE 98

98 order to evaluate th e relationships among the macro-leve l and micro-level measures for predicting offender group membership within the c ontext of a multivariate model with attention to the hypotheses presented at the end of chapte r 3. This multivariate procedure will be used because the multinomial logit model is able to deal with dependent variables in which there is no hierarchical order to the responses and can handle both continuous and discrete independent variables, similar to ordinary regression models (Agresti, 1990). While de pendent variables that are interval scales are completely appropriate for the common multiple regression technique known as ordinary least squares (O LS) and although this same technique is often used in cases where the dependent variable has ordinal outcome s that range from one to four (i.e., strongly disagree, disagree, agree, and str ongly agree), this latter applica tion is typically not correct. There are a variety of assumptions associat ed with ordinary least squares regression including the assumptions of linearity, mean independence, homoscedasticity, uncorrelated disturbance (error) terms, and the assumption of normality (see Allison, 1999); however, the most important assumption that relates to the inappropriate decision to use OLS with categorical dependent variables is the assu mption of linearity. If OLS is applied to an ordinal outcome variable, then there is the expectation that an increase or decrease of one unit on the scale is interpreted as having the same influence no ma tter where you begin on the scale. Similarly, there are cases when the dependent variable out comes have no order at all in that they are nominal scales (Allison, 1999). Although it is theoretically possible to assume some inherent degree of order to outcomes such as being a non-offender, adolescent-limited offender, and a chronic offender (i.e., 1, 2, 3) these offending groups are based on nominal typologies that are theoretically and methodologically driven. There is also the poten tial outcome of where more than these three

PAGE 99

99 groups can be discovered and there is no logica l way to order these groups, for instance earlyonset adolescent-limited offenders, low-low chro nic offenders, late-ons et offenders, etc. Therefore, in this case the most appropriate modeling technique for the form of the dependent variable (i.e., group membership) in this study is what is referred to as the multinomial logit model. Fundamentally, the multinomial logit model (MNLM) can be referred to as an extension of the typical binary logit model ( 0, 1) where binary logits are si multaneously estimated for all of the potential comparisons across the available ca tegorical outcomes (offender groups in this case) (Long, 1997:149). For example, its results ar e similar to estimating binary logit models hypothetically comparing group 1 (non-offende rs) against group 2 (adolescent-limited offenders), group 3 (chronic offenders) against group 1 (non-offenders), and then finally group 2 (adolescent-limited offenders) against group 3 (chr onic offenders). This obviously can become much more work and present difficulty for expl aining the results across these models as the number of groups increase. The multinomial lo git model can be expressed in terms of the probability of observing outcome m given x (assuming y is the dependent variable with J nominal outcomes, i.e., adolescent-limited offende r, chronic offender, etc.). The MNLM as a probability model takes the fo llowing form (Long, 1997:152): Pr ( yi = m | xi) = exp ( xim) / j J =1 exp ( xij). Similarly, the MNLM can also be presented in the form of an odds model, where the odds of outcome m versus outcome n given x is illustrated as follows (Long, 1997:154): m | n ( xi) = exp ( xi m | n), and, after several derivations, the odds of an outcome m (being an adolescent-limited offender) versus outcome n (being a non-offender) are expected to change by a factor of exp ( k m | n

PAGE 100

100 multiplied by ), controlling for the eff ects of the other variables in the model (Long, 1997:169). This odds model allows for a clean er interpretation. For example, given this formula we could now hypothetically say that the odd s of being an adolescent-limite d offender relative to being a non-offender are 2.9 times greater for males, holding all other variables constant. For all of the following multinomial analyses, la tent class membership will be treated as the outcome variable with dichotomous variab les representing each offender group and the nonoffender group (group 1) will be left out as the reference group. The results of the multinomial logit models that provide the best model fit for these data will be asse ssed using the pseudo-R2 statistic which is bounded between the values of zero and one, with va lues closer to one indicating better model fit (Long, 1997). The multi nomial analyses will be estimated separately by cohort for the 1942 and 1949 Racine continuous-re sidents (total=558 indi viduals) who had all of the available information (interview and bl ock data) necessary to assess a theoretically integrated model for explaining gr oup-based offending patterns. The progression of the models will be as fo llows: Model 1 will be a baseline macro-level model where the effects of the social disorg anization index (neighborhood SES and residential mobility) for predicting group membership will be determined. Model 2 will be estimated similarly to Model 1, although this model will on ly include the individual-level independent variables (life transitions, race, and low SES hous ehold, and sex). Finally, Model 3 will present the full model where both the macro-level a nd micro-level measures with be estimated simultaneously, which would allow for an explor ation into the possible mediation effects of individual and family processes on the relationshi p between living in a social disorganized area and group membership. The results of all of thes e analyses will be discussed and elaborated on in Chapter 5.

PAGE 101

101 Study Strengths Notwithstanding the limitations noted previous ly in this chapter, there are several important strengths of this study that are worth mentioning. First, this study answers the recent calls among criminologists regardi ng moving toward theoretical in tegration of macro-level and micro-level levels of analysis (Sampson et al., 1997; Messner et al., 1989), specifically focusing on social disorganization theory (Shaw & McKay, 1942), Sampson and Laubs age-graded theory of informal social control (1993), a nd Moffitts (1993) offender taxonomy. Furthermore, it approaches this process by employing a relati vely new methodological t ool that has gained a substantial recognition in the field as an appropr iate modeling technique for identifying distinct behavioral patterns (for re view, see Piquero, 2007). Second, as an addition to the trajectory literat ure, this study provides an exploration into the importance of sex for distingu ishing group-based trajectories. Third, prior studies that have estimated trajectories with the Shannons (1988, 1991) Racine cohorts have relied on only felonies and misdemeanors as the measure of offending (see DUnger et al., 1998); therefore, this research also includes status offenses in the hopes to better capture adolescent-limited offending patterns. This study also improves by Lands (2000) study by estimating a series of trajectory models as a sensitivity analysis to discover if similar offe nding trajectories are produced by cohort and data source. This is es pecially of concern on ce the Racine continuousresident subsamples are merged with the interv iew level data, and subsequently, once the even fewer of the individuals are merged with the limited amount of available census data characterizing the community in which they were socialized. Finally, th is study is perhaps the first one (see Land, 2000 for a possible exception) to use trajectory analysis along with multilevel regression models to predict the likelihood of exhibiting a distinct pattern of offending over the life-course.

PAGE 102

102 Table 4-1. Resident Samples by Cohort, Size, Data Sources, and Offender Groups. Continuous-Resident Samples Sample Size (n) 1942 COHORT Continuous-Residents 633 Continuous-Resident s (w/ Interview Data) 333 Continuous-Residents (w/ I ndividual & Block Data) 235 1949 COHORT Continuous-Residents 1297 Continuous-Resident s (w/ Interview Data) 556 Continuous-Residents (w/ I ndividual & Block Data) 323 Table 4-2. Observational Peri ods of Data Sources by Cohort Data Sources Years 1942 Cohort 1949 Cohort Police Contact Data 1942 Cohort 1949 Cohort 1948-1967 1955-1974 6 to 25 years of age 6 to 25 years of age 6 to 25 years of age 6 to 25 years of age Interview Data 1976 34 years old 27 years old Census Data 1950 8 years old 1 year old Table 4-3. Social Disorganization Index Factor Loadings by Cohort 1950 Census Measures 1942 1949 Owner-Property Value -.774 -.818 % Renter-Occupied Units .794 .794 % Overcrowded .612 .665 Eigenvalue % Variance Explained N 1.61 53.47 235 1.75 58.04 323

PAGE 103

103 CHAPTER 5 RESULTS Introduction This chapter begins with a description of the summary statistics and mean differences across the cohorts for all of th e independent and dependent meas ures. Following this discussion is a presentation of the groupbased trajectories for all of the samples of the 1942 and 1949 Racine continuous residents (full sample, sample with interview data, and sample with interview and block data). Included in the explanation of the group-based trajecto ry results is also a discussion of the effect of sex (being male) as a covariate for increasing the likelihood of being in an offender group as opposed to a non-offend er group. This chapter concludes with an ANOVA presenting the mean covariate levels ac ross groups for all variables used in the multivariate analysis by cohort. This is followed by an illustration of the multivariate results and a description of the impact of macro-level and micro-level f actors for distinguishing between group-based trajectories. The multivariate anal ysis concludes with a presentation of the theoretically integrated model fo r predicting group membership. Descriptive Statistics Table 5-1 below presents the descriptive statis tics that correspond to all of the macro-level and micro-level variables by cohort described previ ously that will be used in the mlogit analysis that follows and the zero-order correlations among the contin uous measures can be found in Appendix A. Independent Variables As can be seen in Table 5-1 there was a vari ety of measures that significantly differed between the 1942 and 1949 Racine continuous resi dents that precluded the possibility of combining the cohorts in further analyses. Regard ing the measures that we re combined to create

PAGE 104

104 the social disorganization index, there was evid ence that while both cohorts were characterized by comparable measures of mean owne r-property value (1942=$9,374; 1949=$9,196) and roughly equivalent percentages of rente r-occupied units (1942=34.87%; 1949=34.99%), there was an indication that the 1949 c ohort (M=1.89; SD=3.83) had a sign ificantly greater percent of overcrowded units on average when compared with the 1942 cohort (M=1.48; SD=4.40). Turning to the micro-level measures, there were fairly comparable sex percentages between the cohorts with 47% of the 1942 samp le and 52% of the 1949 sample being male. While equivalent percentages of 1942 and 1949 R acine continuous reside nts reported having been employed prior to age 25 (94%), there wa s a significant mean di fference in reporting whether or not they had also been married pr ior to age 25 with more of the 1942 residents (1942=77%; 1949=66%) reporting havi ng experienced marriage. Once these measures were combined into the additive stakes in conform ity index (range 0-2 life transitions), the 1942 cohort reported having significan tly higher stakes in confor mity (M=1.71; SD=0.45) when compared with the 1949 cohort (M=1.60; SD =0.63). Comparatively, the 1949 cohort was comprised of significantly more Bl acks (11% versus 6%) and were more likely to have reported having moved at some point in their juvenile years (61% versus 49%). In addition, the 1942 cohort reported having been socialized in significantly lower SES households (1942=3.61; 1949=3.25) and were less likely to have grown up in a natural two-parent household (84% versus 89%) when compared with the 1949 cohort. Al though, the 1949 cohort repor ted slightly greater numbers of children in their families, perceived more succeptible to differential association, and had higher perceptions of formal social contro l in their neighborhoods during their juvenile years, these differences were not significantly different.

PAGE 105

105 Dependent Variable The prevalence and incidence of police co ntacts for the 1942 and 1949 Racine continuous residents can also be found in Table 5-1. While the cohorts had relatively comparable prevalence estimates of having come into co ntact with the police with 37.9% of the 1942 residents and 41.2% of the 1949 residents having an officially recorded police contact, the 1949 cohort had significantly greater m ean numbers of police contacts than did the 1942 cohort. On average, the 1949 residents had 2.33 police contac ts (SD=6.71) whereas the 1942 residents had a 1.29 average number of police contacts (S D=2.63). Additionally, the 1949 cohort had accumulated a significantly greater total num ber of police contacts by age 25 (754 police contacts) when compared with the 302 accumulated police contacts for the 1942 cohort. Trajectory Analysis In order to aid in the substantive interpreta tion of the nature of the offending trajectories identified and estimated for the 1942 and 1949 Raci ne continuous residents, the predicted and average curves for each of the trajectories implied by the model are plotted in their corresponding figures presented below. The tr ajectory results are described in order of progression and by cohort, thus the first model is that which was produced using the full sample of 1942 continuous residents, the second model disp lays the results from a model relying on data for the 1942 continuous interviewed residents, and the third model illustrates the trajectories of the 1942 continuous interviewed residents with block data This same process is then repeated for the 1949 Racine continuous residents (total =six models; 1942=3 models; 1949=3 models). Each model is discussed in term s of the various model selecti on criteria (Nagin, 2005), or more specifically: (1) the increase in the BIC value (as previously described in Chapter 4); (2) the posterior probabilities for group assignment; a nd (3) the models ability to demonstrate a

PAGE 106

106 theoretically distinct and subs tantively meaningful number of groups. Finally, the models are described with attention to examining the e ffect of sex (being male) on group membership. The model estimations began by fitting th e 1942 Racine continuous residents police contact data to a two group model and then co mparing the BIC values from a three group model to the two group model. The three-group model pr oved a better fit and th e analysis continued by adding groups and observing the improvement in the BIC values. The trajectory procedure ultimately arrived at a four-group solution beca use despite several attempts the model was unable to converge with a five-g roup model (see Table 5-2). The average and the predicted curves are di splayed below in Figure 5-1 and these plots illustrate that the predicted (smoothed) curves closely followed the true mean offending curves. The first group (G1) represents the over whelming majority of the sample or 73.1% of all of the 1942 Racine continuous residents. Individu als in this trajectory group exhibit virtually no offending (as measured by official data) across the entire twenty years of observation (i.e., ages 6-25). The second group includes 8.8% of the sample and individua ls in this group appear to demonstrate the type of offending that woul d characterize adolescent-limited offenders as described by Moffitt (1993) with their offending beginning in adolescence, reaching a peak by age 16 before declining and desisting prior to young adulthood. No i ndividual in this group accumulated more than 10 police contacts by age 25 and the average number of police contacts for this group was 3.91 (SD=1.98). The third gr oup, representing 13% of the sample, exhibit a low-level of chronic offending beginning in early adolescence (age 12) and continue to illustrate a non-zero rate of offending into adulthood and are still offending at age 25; however, no individual in this group accumu lated more than seven police c ontacts during the twenty year period. Group 4 (5.1%) presents a sharp contrast to the other groups in that individuals in this

PAGE 107

107 group begin offending relatively earl y in life (some even as early as age 7) and they maintain a rather high amount of police c ontacts throughout the remainder of the observation period but does decline in young adulthood. Although, it does appe ar that these individuals peak in late adolescence they still exhibit high levels of offending in early a dulthood. No individual in this group accumulated less than 5 police contacts and a handful of offenders in this group had even amassed more than 20 police contacts before re aching age 26. Overall, this small group of offenders was responsible fo r 352 total police co ntacts between the ages of 6 and 25. When turning to an examination of the mean posterior probabilities for group assignment (see Table 5-3), the mean assignment probabiliti es for each group are noticeably high suggesting that there was little ambiguity in the model when determining group assignments. For example, for G1, the mean posterior probability was .91, and the next highest was G3 with a mean probability of .07 of being assigned to G1; for group 2, the mean posterior probability was .84 (median=.87); for group 3 it was .85 (median= .93); and for group 4 it was .91 (median=.99).1 The results of this trajectory analysis provided evidence that sex was a highly significant covariate for determining group a ssignment (see Table 5-4). With G1 serving as the reference group (i.e., the non-offender group), be ing male significantly increase d an individuals likelihood for being assigned to any of the three offender groups (i.e., adolescentlimited/peaked, low-level chronic, and high-le vel chronic). Table 5-5 illustrates the BIC cr iteria for assisting in model selection for the trajectory analysis using the reduced random sample of in terviewed Racine continuous residents and the average and predicted group-based offending traj ectories are displayed in Figure 5-2. These 1 All average posterior probab ilities are above the standard cut-off of .70 (Nagin, 2005).While th ere is always the possibility of mis-classification of individuals to particular trajectories none of the posterior probabilities in these data were below .80 thus indicating a high degree of discrimination between groups.

PAGE 108

108 trajectories were similar to those produced in th e previous analysis, alth ough only three distinct groups were identified in this particular analysis.2 The first group represented the overwhelming ma jority of the Racine interviewed residents with 86.2% of the sample being classified as non-offenders with virtua lly no offending during the observation period. Comparativ ely, G2 represented 7.8% of th e sample and individuals in this group appeared to display an offending patt ern consistent with Moffitts (1993) adolescentlimited offender type wherein these individuals began offending in adolescence before peaking in mid-adolescence and showed evidence of desist ance near age 19 with no individual in this group having any officially recorded police cont acts after age 20. Indivi duals in this group had an average of 4.50 (SD=2.49) police contacts and no individual in this group accumulated more than 12 police contacts prior to reaching age 26. The third group, representing 6% of the sample, is recognizable by their early onset of offending (p rior to age 10) and rela tively high levels of offending throughout the rest of the observation pe riod. Although there is evidence that this groups rate of offending peaked in late adolescen ce/early adulthood they still maintained relatively high levels of offending through age 25. On average, individuals in this group had 7.85 (SD=3.42) police contacts and a handful of these offenders accumulated more than 10 police contacts between the ages of 6 and 25. Furt hermore, this small group of chronic offenders was responsible for a tota l of 157 police contacts. Similar to the results of the previous traject ory analysis, the posterior probabilities for group assignment were remarkably high providing evidence that the individual trajectories in these data could be assigned to a group-based trajectory with a relatively high degree of 2 A four group model was not able to fit these data, thus the low-level chronic offending group was not identified in this stage of the analysis.

PAGE 109

109 probability (see Table 5-6). Only in G2 did the mean posterior probabilities fall slightly below .90 but still remained well above the .70 cut-off (Nagin, 2005). Consistent with the previous model, the results in Table 5-7 indicated that being male signifi cantly increased an individuals likelihood of being assigned to either offende r group (adolescent-limited/peaked or chronic) when compared with the non-offender group. Table 5-8 presents the BIC model selection cr iteria for the trajectory analysis using the data from the 1942 Racine continuous interviewed residents with block data that provides the trajectories and groups that will be of interest in the 1942 ANOVA and mlogit analysis that follows. A three-group model once ag ain provided the best fit for these data and the average and predicted offending curves are plotted below in Figure 5-3. G1, representing 70.2% of the sample, are virt ually a non-offending group with little to no police contacts during the ages of 6 to 25. The second group, representing 20% of the sample is similar to the adolescent-limited offender type proposed by Moffitt (1993) with the individuals in this group beginning to offend in adolescence before peaking around the age of 16 before declining and showing evidence of having desist ed by 21. No individual in this group had accumulated more than 5 total police contacts during the twen ty years of follow-up. Also consistent with the earlier analyses, the third gro up exhibited an early onset of offending (prior to age 10), relatively high levels of offending throu ghout adolescence before reaching a peak at age 16; however, these individuals continued to mainta in comparatively high rates of offending into adulthood and through the age of 25, although the trajec tory was declining. No individual in this group had accumulated less than 4 police contacts and a handful of these individuals had as many as more than ten police contacts prior to reaching age 26 (M=8.09; SD=3.30). This small number of Racine resident s amassed 186 police contacts be tween the ages of 6 and 25.

PAGE 110

110 Table 5-9 presents the posterior probabilities for group assignment and consistent with the two prior trajectory analyses us ing residents from the 1942 cohort, there was relatively little ambiguity when the model was determining group assignment with mean posterior probabilities as high as .91 (median=.99) for G3 and still co mparatively high for G1 (M=.89; median=.97) and G2 (M=.89; median=.97). Additionally, the results indicated that sex (being male) significantly increased an individuals like lihood of belonging to either an adolescent-limited/peaked or a chronic offending group when compared with membership in a non-offending group (see Table 5-10). The next set of trajectory analyses i nvolves data from the 1949 Racine continuous residents following the same progression of mode ls that were estimated for the 1942 cohort: (1) continuous resident s; (2) continuous interviewed residents; and (3) continuous interviewed residents with block data The results of the BIC model selection criteria for the 1949 continuous resident sample is pr esented below in Table 5-11. This was the only sample in which a five-group model provided the best fit for these data, but this is not su rprising given that prior researchers have shown evidence that the number of groups typically increase with sample size (DUnger et al., 1998; Nagin & Tremblay, 2001; Samp son et al., 2004) and this was in fact the largest sample in this study (N=1,297). The average and predicted curves for the 1949 c ontinuous resident trajectory analysis can be found in Figure 5-4 and consistent with th e analyses produced using data from the 1942 continuous residents, the pred icted offending curves tended to closely follow the true offending curves. The first group, representing 57.1% of the sample, is purely a non-offender group as none of these individuals had any reco rded police contacts thro ughout the entire twenty years of observation. The second group, represen ting 32.7% of the sample, exhibited very low

PAGE 111

111 rates of offending over the entire observation peri od. There was a tendency for these individuals to peak at age 16 and then decline, but they still showed non-zero rate s of offending through age 25. On average, individuals in this group ha d 1.98 (SD=1.24) police cont acts and no individual in this group had more than 7 police contacts. G3, representing 2.6% of the sample, is an atypical group that is no t identified in any of the other anal yses presented in this study although this offender group has been the point of discussi on in prior research (s ee DUnger et al., 1998; Sampson and Laub, 1993). Individuals in this gr oup are unique in that they had virtually no officially recorded police cont acts until they reached adulthood (age 18) and their rates of offending increased rather noticeably in young adu lthood. No individual in this group had less than 3 police contacts (M=6.53; SD=3.89) and a handful of these offenders had actually accumulated more than 10 police contacts despite the fact that their offending did not begin until young adulthood. The fourth group, representing 5.7% of the sample, can be best described as an adolescentpeaked group (see DUnger et al., 1998; Land, 20 00) in that these individuals do not begin offending until early adolescence but offend at relatively high rates during mid-adolescence before peaking at age 15 and then virtually have desisted by age 23. No individual in this group had less than 6 police contacts and more than half of the individuals in this group had accumulated more than 10 police contacts (the majo rity of which were dur ing mid-adolescence). G5, representing the smallest percentage of th e sample (1.9%), exhibited a high-rate chronic pattern of offending that begins early on in life (prior to age 8) and their offending drastically increases well into adolescence before reaching a p eak in late adolescence. These individuals still display high levels of offending throughout the remainder of the observation period or age 25. Individuals in this group averaged an alarming 32.48 (SD=10.30) police contacts and the

PAGE 112

112 police contacts of this group ranged from a low of 22 to a high of 63 police contacts. This small group of offenders managed to accumulate 812 poli ce contacts during the ages of 6 to 25! The mean posterior probabilities for group assi gnment indicated that like the trajectory analyses using the 1942 Racine continuous resident data, this model demons trated the ability to determine group membership with a high degree of probability (see Table 5-12). G1 had the lowest mean probability (M=.84; median=.93), yet th is was still way above the standard cutoff of .70 (Nagin, 2005). The mean posterior probabilitie s were remarkably high for the low (G2) and high-rate (G5) chronic offending groups indicating th at these individuals trajectories faced little risk for mis-classification. According to the coefficients and their corresponding p-values produced for this sex covariate, there was eviden ce that for all of the offender groups except the late-onset offenders (i.e., adol escent-limited/peaked, low and hi gh-rate chronic offenders) being male significantly increased the likelihood of group membership into an offender group compared with a non-offender group (see Table 5-13). Table 5-14 presents the BIC model se lection criteria for the 1949 continuous interviewed residents and similar to the interviewed 1942 samp le only three trajectory groups were produced. This analysis failed to detect a late-onset or a low-rate chronic group as wa s the case in the prior 1949 continuous resident analysis. The average a nd predicted curves, as plotted in Figure 5-5, indicated that the following three trajectory gr oups could be identified: (1) non-offenders; (2) adolescent-peaked offenders; and (3) high-rate ch ronic offenders. G1 (75.1%) represented the overwhelming majority of the sample and exhi biting practically no offending across the twenty year time period. Comparatively, the second group (20.5%) did not be gin offending until early adolescence before reaching a peak in mid-adol escence and then desists by their early 20s. Individuals in this group recorded an average of 4.32 (SD=2.81) pol ice contacts with a range of 2

PAGE 113

113 to 14 police contacts. G3, repres enting 4.3% of the sample, exhibi ted an early-ons et of offending (prior to age 8) and increased their offendi ng patterns throughout mi d and late-adolescence before peaking at age 18. Neve rtheless, these individuals still offended at relatively high rates by age 25, although showing evidence of desistance in young adulthood. All of the individuals in this group had accumulated at least 10 police co ntacts between the ages of 6 to 25 (M=23.21; SD=12.95) and this small number of offenders were responsible for 557 police contacts prior to reaching age 26. Table 5-15 provides the mean posterior pr obabilities for group assignment and these results show a strong ability for discriminating across trajectory groups with no mean posterior probability less than .92 (G2). In fact, the mean posterior probabilities are remarkably high for G1 (M=.97; median=.99) and G3 (M=.99; median= .99). In addition, consis tent with all of the prior trajectory analyses, being male significan tly increased the likelihood for an individual being assigned to either an a dolescent-limited/peaked or high -rate chronic offending trajectory when compared with assignment to a non-offending group (see Table 5-16). Table 5-17 presents the BIC model selection criteria for the final trajectory analysis in this study using the data from the 1949 Racine continuous interviewed residents with block data that provides the trajectories and gr oups that will be of intere st in the 1949 ANOVA and mlogit analysis that follows. Similar to the prior 1949 continuous interviewed resident analysis and consistent with the results of the comparable 1942 continuous interviewed residents with block data analysis, the procedure indicated that a thre e-group model provided the best fit for these data. The average and predicted curves are displa yed in Figure 5-6 below. The first group, representing the majority of the sample (75.2%), exhibits al most no offending throughout the

PAGE 114

114 entire twenty year observation period. In contrast, G2 (20.4%) resembles the adolescentlimited/peaked group identified in the two previo us trajectory analyses using the 1949 cohort data in that individuals in th is group did not begin offending until adolescence but did not reach a peak in offending until 17 and then declined until appearing to desist by their early 20s. Furthermore, these individuals averaged approxi mately 5 (SD=3.45) police contacts and they had accumulated at least two police c ontacts between the ages of 6 and 25 with some of these group members having amassed as high as 14 police contact s. G3 (3.7%) exhibits the high-rate chronic offending patterns discussed previously in that th ese individuals begin o ffending early (prior to age 8) and increase dramatically during adoles cence before reaching a peak in young adulthood (age 18). Yet, this group displays elevated levels of offending by age 25, but showed evidence of desistance. No individual in this group had less than 14 po lice contacts and the average number of police contacts for this group was 30.25 (SD=15.89). Overall, these handful of offenders had accumulated 363 police contac ts between the ages of 6 and 25. Table 5-18 displays the mean posterior proba bilities for group assignment and Table 5-19 illustrates the effect of sex (being male) on group membership. Like all of the previous analyses, the mean posterior probabilities were noticeably high, ranging from .89 (G2) to .98 (G1) and as high as .99 for G3. This was an indication that there was little ambiguity in assigning individual patterns of offending to a group-ba sed trajectory. Also, once again the results showed evidence that being male significantly in creased an individuals likelihood of being assigned to either an adolescent-limited/peaked or high-rate chroni c offending group compared with a non-offending group. With all of the trajectory anal yses concluded, it is important to discuss these models within the context of the rela tive stability/instability of identifying offender groups across the various

PAGE 115

115 data sources (continuous residents, continuous interviewed residents, and continuous interviewed residents with block data ). A graphical display of the offe nder groups across the data source and by cohort are presented in Table 5-20. As can be s een there appears to be more similarities than differences in the offense trajectories acro ss the data sources a nd between the 1942 and 1949 cohorts. First, in both of the original 1942 and 1949 continuous resident samples with the largest sample sizes (N=633 and N=1,297) produced the great est number of groups. The four remaining trajectory analyses with reduced samples were only able to identify three groups, which is consistent with prior research that has indica ted that the number of groups increases with the sample size (DUnger et al., 1998; Nagin & Trem blay, 2001; Sampson et al., 2004). Second, in all of the analyses the non-o ffending group (G1) was always the most prevalent group in the sample (ranging from 57.1% to 86.2%) and sim ilarly, the high-rate chronic offending group was always the smallest percentage of the offende r groups (ranging from 1.9% to 9.8%). Third, the 1942 and 1949 original continuous resident samples al so both identified a group that is atypical to what Moffitts (1993) taxonomy would expect or who have b een classified as low-level chronic offenders who offend early and maintain low non-zero rates throu ghout the twenty year period of observation; however, this type of offender group has been observed elsewhere in previous trajectory analyses (s ee DUnger et al., 1998; 2002). Regarding the differences, first, the largest sa mple in this study (1949 continuous resident sample of 1,297 individuals) actu ally identified an additional at ypical trajectory group or what others have referred to as a late-onset group (see DUnger et al., 1998; Sampson & Laub, 1993) who do not begin offending until late adolescenc e/early adulthood, but offe nd at relatively high rates in adulthood. Second, while all of the anal yses displayed an adol escent-limited/peaked group of offenders who was the larg est group of offenders in four of the six analyses, in the 1949

PAGE 116

116 trajectory analyses this group te nded to resemble an adolesce nt-peaked group (DUnger et al., 1998; Land, 2000) in that they stil l exhibited low rates of offendi ng into early adulthood before desisting in the early 20s. The discussion in th e next section now moves toward examining the results from the ANOVAs in order to further in vestigate what macro-level and micro-level factors are significant for disti nguishing between offender groups. Analysis of Variance The results from the analysis-of-variance te sts where the macro-level and micro-level risk/protective factors are compared across group s and by cohort are presented in Table 5-21. The group means of offending (number of police cont acts) are also displayed in Table 5-21. The trajectory models displayed in Figure 5-3 (1942 cohort) and Figure 5-6 (1949 cohort) are those that all of the remaining analys es are based on. The first most obvious findings is that relatively all of the risk factors are concentrated among the offender groups and the protective factors are primarily found in the non-offender group (G1). Mo re specifically, the ri sk factors tend to be heightened in the adolescent-limite d/peaked groups (G2) and then further elevated in the highrate chronic groups (G3) (G3 for both th e 1942 and 1949 cohorts will be referred to interchangeably as high-rate or chronic offende rs hereon). For the 1942 and 1949 cohorts, social disorganization, sex (being male), race (being Black), living a low SES household, living with natural parents, and perceived formal social control are distinguishing features of group membership. Additionally, whether an individua l had moved during their juvenile years was a distinguishing feature for 1942 continuous re sidents, whereas stakes in conformity3 and family 3 In order to further explore the possible influence of marriage and employment on group-based offending trajectories, the mean age at time of marriage and first full-time job were computed and compared across groups and by cohort. Regarding the 1942 cohort, 83.0% of the non-offenders (G1), 70.2% of the adolescent-limited/peaked (G2), and 86.9% of the chronic offenders (G3) reported being married by age 25 and of those that reported being married, the mean age of marriage was 20 .8 (G1), 21.6 (G2), and 20.7 (G3). In contrast, 97.6% of G1, all of G2, and 95.6% reported having their first full-time job by age 25 and the mean age of first full-time employment was 18.4 (G1), 18.4 (G2), and 18.0 (G3). When turning to the 1949 cohort, 72.7% of the non-offenders (G1), 63.6% of

PAGE 117

117 structure (number of children in family) were also distinguishing features for group membership among the 1949 continuous residents. While the preliminary results from the f-test in the ANOVAs revealed the concentration of risk factors among the offender groups, it was nece ssary to conduct Tukeys B post hoc tests in order to determine which (if any) of the gr oups significantly differed from one another. Concerning the 1942 cohort, individuals who were in either offender group lived in significantly more socially disorganized communities and both the adolescent-limited/peaked groups (G2) and the high-rate chronic groups (G 3) had significantly more males and G3 was characterized by a significantly greater number of Black s. G1 also had a significantly greater number of individuals reporting having grown up in a two natural parent household when compared with both G2 and G3. In addition, all of the groups mean number of police contacts signif icantly differed, or in other words G3 had significantly more police cont acts than G1 and G2, and G2 had significantly more police contacts than G1. When turning to the Tukeys B post hoc tests results for the 1949 cohort, while individuals in G3 and G2 lived in more socially disorgan ized areas than G1 the groups means did not significantly differ. Comparativel y, there was a host of individual level factors that significantly varied particularly when comparing G3 and G2 to G1. The offender groups had significantly more males and Blacks in addition to having a si gnificantly greater number of children in the household and had significantly higher perceptions of formal social control (police presence) when compared with G1 (although G2 and G3 did not significantly differ from each other). The the adolescent-limited/peaked (G2), an d only 33.3% of the chronic offenders (G3) experienced ma rriage by age 25 and of those that reported being married, the mean age of marriage was 21.0 (G1), 21.1 (G2), and 21.3 (G3). Comparatively, 96.7% of G1, 95.5% of G2, and 91.6% reported having their first full-time job by age 25 and the mean age of first full-time employment wa s 19.0 (G1), 18.8 (G2), 17.3 (G3). Thus it appeared that typically fewer adolescent-limited/peaked offenders and fewer chronic offenders reported having experienced marriage and/or employment by age 25, yet there was not consistent eviden ce that the mean age of having experienced these life transitions varied greatly across groups and by cohort.

PAGE 118

118 high-rate chronic offenders (G3) also had signi ficantly lower household SES and less stakes in conformity than both G1 and G2. Furthermor e, G1 was comprised of significantly more individuals that reported coming from a two natural parent hous ehold when compared with G2 and G3. Overall, similar to the 1942 ANOVA resu lts, G3 had significantly more police contacts than G1 and G2, and G2 had significantly more po lice contacts than G1. It is now important to investigate if these same risk/protective factor s that were shown to significantly distinguish group membership in these bivariate analyses are also significan t for predicting group membership when estimated simultaneously, which now becomes the focus of the next section. Multivariate Analysis The following set of multinomial regression analys es are estimated in order to compare the ability of the macro-level and micro-level fact ors to differentiate among the offender groups. Each cohort-specific analysis (beginning with th e 1942 cohort) will be divided into the following three separate stages: (1) macro-level level; (2) micro-level level; and (3) the integrated model. The first two regression analyses will assist in is olating the relative strength of prediction for the different levels of analysis (macro-level and micro-level) and the latter will allow for an assessment of the proposed integrated theoretic al model for differentiating the non-offenders from the adolescent-limited/peaked and high-rate chronic offenders.4 According to the results from the first macr o-level-only model, using data from the 1942 Racine cohort (see Table 5-22) there was support for the effect of soci al disorganization in differentiating the offender groups. Specifically, resi ding in a socially disorganized area nearly doubled the odds of an individua l being assigned to an adol escent-limited/peaked group and 4 Models were estimated using clustered robust standard errors in Stata 9.2.

PAGE 119

119 more than doubled the odds of an individual being assigned to a high -rate chronic offending group relative to a non-offender group.5 Comparatively, Table 5-23 presents the results fro m the micro-level-only model. First, the findings showed a robust effect for sex on the odds of being assigned to ei ther type of offender group with being male nearly tripling the odds for being assigned to an adolescentlimited/peaked offender group and increasing the odds by more than six times for an individual being assigned to a high-rate offender group. Ha ving been socialized in a natural two parent household yielded a significant protective effect on the odds of being assigned to adolescentlimited/peaked offender group by reducing the od ds by 38%, and it produced a similar reduction in the odds (41%) of being assigned to a high -rate chronic offender group relative to the nonoffender group. In addition, having moved during th e juvenile years more than doubled the odds of being assigned to a high-rate chronic offendi ng trajectory. Although the majority of the other individual level risk/pro tective factors were in the expected direction insofar as increasing or decreasing the odds for being in a particular group, being Black, living in a low SES household, stakes in conformity, number of children in the household, percei ved delinquent peer influence, and perceptions of formal social control did not achieve statistic al significance for distinguishing across trajectories. Table 5-24 displays the multivariate results for the integrated theoretical model for predicting offender group membership. The findi ngs suggest that once the macro-level and micro-level factors are estimated simultaneously, so cial disorganization is left only marginally 5 Percent non-White was originally included in the analyses as a control variable, but it failed to exert any significant effects for predicting offender group membership. Models were also estimated using backwards deletion and the results confirmed that percent non-White did not influence trajectory group membership or increase the model fit, so it was dropped and not included in the analyses presented here. According to Sampson (1997) there has been little support for this construct in relation to offending wh en included with neighborhood poverty and residential instability and given the relatively high degree of homogeneity in these data, it is not surprising that this variable was non-significant.

PAGE 120

120 significant (p<.10) for increasi ng an individuals odds for bei ng assigned to an adolescentlimited/peaked offender group in relation to a non-offender group. In contrast, the effect of social disorganization on th e likelihood of being assigned to a chronic offender group was completely mediated by the addition of the mi cro-level factors. Additionally, upon comparing the pseudo R-square statistics, it appeared that the addition of social disorganization did relatively little for providing a better model fit (McFaddens in creased .23 to .25; Nagelkerke increased .39 to .41). Overall, the same three micro-level variables that achieved statistical significance in the micro-level only model discussed above (s ex and natural two parent household for both offender groups and having moved for the highrate chronics only) still remained statistically significant dis tinguishing factors for group assignment. The results of the multivariate analysis for di stinguishing across trajectories based on data from the 1949 cohort are discussed below. Simila r to the prior mlogit models presented above, the first model illustrates the e ffect of social disorganization alone, the second demonstrates the effect of only the micro-level factors, and th e final model displays the findings from the theoretically integrated model for distinguishing across offender groups. Consistent with the results of the macrolevel-only model for the 1942 cohort, social disorganization statistica lly increased the odds for an indivi dual being assigne d to either an adolescent-limited/peaked offender group or a hi gh-rate chronic offender group in relation to a non-offender group (albeit only ma rginally significant for the high-rate chronic offender group, p=.07) for the 1949 cohort. More specificall y, living in a socially disorganized area approximately increased an indi viduals odds one and a half time s for being assigned to either offender group rather than to a nonoffender group (see Table 5-25).

PAGE 121

121 Table 5-26 reveals the results of the microlevel factors for predicting group membership for the 1949 cohort. There were several similari ties as well as differences for these indivdual level factors for distinguishing across trajecto ries in the 1949 cohort an alysis when compared with the 1942 mlogit analysis discussed previousl y. The key similarity is that a rather robust effect for sex was again found with being male mo re than doubling an i ndividuals odds of being assigned to an adolescent-limite d/peaked group and increasing an individuals odds nearly two and a half times for being assigned to a highrate chronic offending group in relation to a nonoffender group. Additionally, both family-level factors (family st ability and family structure) were found to be statistically significant di stinguishing features of group assignment to an adolescentlimited/peaked offender group with residing with na tural parents resulting in a 46% reduction in the odds and each additional child increasing an individuals odds 1.36 times for being assigned to an adolescent-limited/peaked offender group. Be ing Black also marginally increased (p=.06) an individuals odds of being assigned to an adolescent-limited/peaked offender group. Comparatively, although th ese two family factors nearly ach ieved statistical significance for increasing the odds of being assigned to a high -rate offender group [family stability (p=.14); family structure (p=.11)], residing in a low SES household was the more important distinguishing feature of group assignment to this particular groupbased trajectory. An individuals odds of being assi gned to a high-rate chronic o ffender group in relation to a nonoffender group increased more than two and a half times if they reported having been socialized in a low SES household. The remaining micro-le vel factors (stakes in conformity, differential association, perceptions of formal social control, a nd having moved during the juvenile years) were not statistically significant fo r distinguishing across trajectories.

PAGE 122

122 The findings from the final mlogit model in dicating the results of the theoretically integrated model for predicting group memb ership using the 1949 cohort (see Table 5-27) revealed similar results as to what was produced for the 1942 cohort. While social disorganization remained marginally signifi cant for the adolescent-limited/peaked offender group in the 1942 analysis (p=.06), once the macrolevel and micro-level factors were estimated simultaneously for the 1949 cohort, the effect of social disorganization for distinguishing across trajectories was no longer statis tically significant for either offender group (although the odds were still in the expected direction). Comp aratively, the micro-level factors that were statistically significant for pred icting group membership in the previous micro-level-only model still maintained their significance in the integrated model. More specifically, being male more than doubled an individuals odds of being assigned to either an adolesce nt-limited/peaked or a high-rate chronic offender group and family stabi lity and structure were significant determinants of adolescent-limited/peaked offender group memb ership. Additionally, having resided in a low SES household more than doubled the odds of be ing assigned to a high-rate offender group. Overall, upon inspecting the change in the pseudo-R-squares as indi cators of model fit, there was little gained from using a theoretically integrat ed model when compared with the micro-levelonly model (McFaddens increased from .19 to .20; Nagelkerkes increased from .31 to .32). Conclusions The results from the series of analytical pro cedures yielded several substantively important findings. First, the six trajectory analyses (three for each cohort) provided evidence that nonoffender, adolescent-limited/peaked offender, and high-rate chronic offender groups were recognized across all data sources (Racine continuous residents, Racine continuous interviewed residents, and Racine continuous interviewed residents with block data ) and in both the 1942 and 1949 cohorts. Although the group percentages va ried, there was support for hypothesis #1 in

PAGE 123

123 that the semiparametric Poisson-based regression analyses produced these three distinct classes of offenders and that (1) the non-offenders we re the most prevalent of the groups; (2) the adolescent-limited/peaked offenders were the most prevalent of the offending groups; and (3) the high-rate offenders were the least prevalent gr oup of offenders but we re characterized by a higher rate of mean offending across the observa tion period when compared with the adolescentlimited/peaked offender group. However, there were a few groups that were recognized in the larger sample trajectory analyses (low-level ch ronics and late-onsetters ) who are not consistent with Moffitts taxonomy but have been discusse d elsewhere (see DUnge r et al., 1998; Piquero, 2007; Sampson & Laub, 1993). Support was also shown for the second hypothesi s with sex being a significant covariate for distinguishing group-based trajectories. Specif ically, in all of the six trajectory analyses, being male significantly increased th e probability that an individual would be classified as either an adolescent-limited/peaked or chronic offender with the no n-offender group serving as the reference group. The only exception was in the 1949 continuous resident sample (N=1,297). While there were more males in the late-onset offender group for the cohort, being male did not significantly increase the probability of assignment to this particular group-based trajectory. Concerning hypothesis #3 that significant li fe transitions including marriage and employment can affect desistance from crime, there was limited support for Sampson and Laubs age-graded theory of informal social control. There was descriptive evidence at the bivariate level suggesting that individuals in either offender group (adolesc ent-limited/peaked or high-rate chronic offender group) scored lower on the stakes in conformity index when compared with those individuals assigned to the non-offende r group. The results from the ANOVA with the 1949 cohort also indicated that the stakes in conf ormity index was statistically associated with

PAGE 124

124 group membership to a high-rate chronic offendi ng trajectory with indi viduals in this group having significantly lower mean scores on the stak es in conformity index when compared with those in the adolescent-limited/peaked group and individuals in th e non-offender group. Qualified support was also shown for hypothesis #4 regarding the expectation that social disorganization would be a signif icant distinguishing feature for group assignment. Being from a socially disorganized area signi ficantly increased the odds that an individual would be assigned to either an adolescent-limited/peaked offender group or a high-rate offender group in relation to a non-offender group in the 1942 and 1949 cohort ml ogit analyses. The effect, however, was only marginally significant (p=.07) for disti nguishing high-rate chronic offenders in the 1949 analysis. The results indicated that social diso rganization did not signifi cantly distinguish the adolescent-limited/peaked offenders from the chronic offenders. Finally, there was little support provided for hypothesis #5 or the expectation that macrolevel and micro-level effects on the participa tion and frequency of offending will still be observed net of the effects of each other. A lthough the results did i ndicate that once the multilevel effects were considered simu ltaneously, the effect of social disorganization still maintained a marginally significant effect for distinguishing the adolescent -limited/peaked offenders from the non-offenders in theoretically integrated mlog it analysis of the 1942 cohort, the majority of the effect appeared to be indir ect. While the direction of the odds were in the expected direction for both the 1942 and 1949 mlogits (with being from a socially disorganized area increasing the odds of assignment to an offender group), an examination of the R-squares (McFaddens and Nagelkerkes) showed that the a ddition of the social disorganizat ion index contributed relatively little for improving model fit. Once again, social disorganization was unable to significantly differentiate the adolescent-limited/peaked group from the chronic offending group.

PAGE 125

125 Chapter 6 will provide a further discussion of the results from the series of analytical procedures and elaborate on the possible implicati ons of the research findings. The chapter will also highlight the limitations of th e current study, offer suggestions as to how future research can improve upon and extend the examination of integr ated theoretical models and close with a few final thoughts.

PAGE 126

126 Table 5-1. Descriptive Statistics for the Ma cro-level and Micro-level Measures by Cohort Variables 1942 Mean (SD) 1949 Mean (SD) Independent Variables Macro-level Social Disorganization Index 0.00 (1.00) 0.00 (1.00) Owner-Property Value 9374.30 (2895.29) 9196.40 (2951.55) % Renter-Occupied Units 34.87 (20.83) 34.99 (21.43) % Overcrowded 1.48 (3.83)* 1.89 (4.40)* Micro-level Male 0.47 (0.50) 0.52 (0.50) Life Transitions Stakes in Conformity Index 1.71 (0.45)* 1.60 (0.53)* Employment 0.94 (0.22) 0.94 (0.23) Marriage 0.77 (0.42)* 0.66 (0.47)* Black 0.06 (0.24)* 0.11 (0.31)* Low SES Household 3.61 (2.08)* 3.25 (1.91)* Family Stability (Natural Parents) 0.84 (0.37)* 0.89 (0.31)* Family Structure (No. of Children) 3.74 (2.21) 4.02 (2.17) Differential association 0.22 (0.42) 0.25 (0.43) Formal Social Control 1.24 (0.96) 1.32 (1.03) Moved 0.61 (.0.49)* 0.49 (0.51)* Police Contacts 1.29 (2.63)* 2.33 (6.71)* Prevalence 37.9% 41.2% Incidence 302* 754* Note. The asterisk signs denotes si gnificant mean differences between the 1942 and 1949 cohorts (p<.05). Listed below the social disorganization and stak es in conformity indexes are the values of the measures that were used to create each index prior to the data mani pulations (principal components analysis for the social disorganization index and the additive index for experiencing either none, 1, or 2 of the life transitions). Table 5-2. Tabulated Bayesian Information Criteria (BIC) for 1942 Racine Continuous Residents (N=633) Optimal # of Groups BIC Cubic ZIP Model 2 -2453.58 3 -2437.23 4 -2399.77 5 False Convergence

PAGE 127

127 Table 5-3. Mean (Median) Posterior Proba bilities for Group Assignments: 1942 Racine Continuous Residents (N=633) Group Prob (G1) Prob (G2) Prob (G3) Prob (G4) 25th Percentile 75th Percentile 1 (n=463) .91(.99) .02(.00) .07(.01) .00(.00) .89 .99 2 (n=56) .02(.00) .84(.87) .12(.08) .03(.01) .71 .98 3 (n=82) .11(.03) .04(.00) .85(.93) .01(.00) .75 .98 4 (n=32) .00(.00) .04(.00) .05(.00) .91(.99) .91 .99 Note: Minimum and Maximum values presented for the group/prob(group) combination. Table 5-4. Coefficient Estimates for Eff ect of Sex (being Male) on Trajectory Group Membership: 1942 Racine Continuous Residents (N=633) Groupsa coefficient SE p-value Cubic ZIP Model 2 2.861 0.566 <0.001*** 3 2.721 0.508 <0.001*** 4 2.642 0.642 <0.001*** *p<.05 **p<.01 ***p<.001 a Group 1 (the lowest rate group) is the reference category of comparison. Table 5-5. Tabulated Bayesian Information Criteria (BIC) for 1942 Racine Continuous Interviewed Residents (N=333) Optimal # of Groups BIC Cubic ZIP Model 2 -1164.02 3 -1148.61 4 False Convergence

PAGE 128

128 Table 5-6. Mean (Median) Posterior Proba bilities for Group Assignments: 1942 Racine Continuous Interviewed Residents (N=333) Group Prob (G1) Prob (G2) Prob (G3) 25th Percentile 75th Percentile 1 (n=287) .98(.99) .02(.00) .01(.06) .99 .99 2 (n=26) .06(.00) .89(.97) .06(.02) .77 .99 3 (n=20) .00(.00) .02(003) .98(.99) .99 .99 Note: Minimum and Maximum values presented for the group/prob(group) combination. Table 5-7. Coefficient Estimates for Eff ect of Sex (being Male) on Trajectory Group Membership: 1942 Racine Continuous Interviewed Residents (N=333) Groupsa coefficient SE p-value Cubic ZIP Model 2 2.873 0.711 <0.001*** 3 2.123 0.638 <0.001*** *p<.05 **p<.01 ***p<.001 a Group 1 (the lowest rate group) is the reference category of comparison. Table 5-8. Tabulated Bayesian Information Criteria (BIC) for 1942 Racine Continuous Interviewed Residents with Block Data (N=235) Optimal # of Groups BIC Cubic ZIP Model 2 -960.81 3 -940.76 4 False Convergence Table 5-9. Mean (Median) Posterior Proba bilities for Group Assignments: 1942 Racine Continuous Interviewed Residents with Block Data (N=235) Group Prob (G1) Prob (G2) Prob (G3) 25th Percentile 75th Percentile 1 (n=165) .89(.97) .11(.03) .00(.00) .78 .97 2 (n=47) .06(.00) .89(.97) .05(.01) .80 .99 3 (n=23) .00(.00) .09(.01) .91(.99) .85 .99 Note: Minimum and Maximum values presented for the group/prob(group) combination.

PAGE 129

129 Table 5-10. Coefficient Estimates for Eff ect of Sex (being Male) on Trajectory Group Membership: 1942 Racine Continuous Interviewed Residents with Block Data (N=235) Groupsa coefficient SE p-value Cubic ZIP Model 2 2.262 0.507 <0.001*** 3 3.103 0.768 <0.001*** *p<.05 **p<.01 ***p<.001 a Group 1 (the lowest rate group) is the reference category of comparison. Table 5-11. Tabulated Bayesian Information Criteria (BIC) for 1949 Racine Continuous Residents (N=1,257) Optimal # of Groups BIC Cubic ZIP Model 2 -7123.17 3 -6851.59 4 -6769.89 5 -6694.99 6 False Convergence Table 5-12. Mean (Median) Posterior Proba bilities for Group Assignments: 1949 Racine Continuous Residents (N=1,297) Group Prob (G1) Prob (G2) Prob (G3) Prob (G4) Prob (G5) 25th Percentile 75th Percentile 1 (n=740) .84(.93) .15(.07) .00(.00) .00(.00) .00(.00) .73 .93 2 (n=424) .00(.00) .97(.99) .02(.00) .02(.00) .00(.00) .98 .99 3 (n=34) .00(.00) .08(.01) .89(.97) .03(.00) .00(.00) .84 .99 4 (n=74) .00(.00) .04(.00) .01(.00) .93(.99) .02(.00) .91 .99 5 (n=25) .00(.00) .00(.00) .02(.00) .01(.00) .98(.99) .99 .99 Note: Minimum and Maximum values presented for the group/prob(group) combination.

PAGE 130

130 Table 5-13. Coefficient Estimates for Eff ect of Sex (being Male) on Trajectory Group Membership: 1949 Racine Continuous Residents (N=1,297) Groupsa coefficient SE p-value Cubic ZIP Model 2 1.272 0.132 <0.001*** 3 0.508 0.353 0.150 4 2.700 0.432 <0.001*** 5 2.714 0.741 <0.001*** *p<.05 **p<.01 ***p<.001 a Group 1 (the lowest rate group) is the reference category of comparison. Table 5-14. Tabulated Bayesian Information Criteria (BIC) for 1949 Racine Continuous Interviewed Residents (N=556) Optimal # of Groups BIC Cubic ZIP Model 2 -2979.90 3 -2848.07 4 False Convergence Table 5-15. Mean (Median) Posterior Proba bilities for Group Assignments: 1949 Racine Continuous Interviewed Residents (N=556) Group Prob (G1) Prob (G2) Prob (G3) 25th Percentile 75th Percentile 1 (n=418) .97(.99) .03(.00) .00(.00) .98 .99 2 (n=114) .06(.01) .92(.99) .02(.00) .87 .99 3 (n=24) .00(.00) .01(.00) .99(.99) .99 .99 Note: Minimum and Maximum values presented for the group/prob(group) combination. Table 5-16. Coefficient Estimates for Eff ect of Sex (being Male) on Trajectory Group Membership: 1949 Racine Continuous Interviewed Residents (N=556) Groupsa coefficient SE p-value Cubic ZIP Model 2 1.712 0.272 <0.001*** 3 2.774 0.738 <0.001*** *p<.05 **p<.01 ***p<.001 a Group 1 (the lowest rate group) is the reference category of comparison.

PAGE 131

131 Table 5-17. Tabulated Bayesian Information Criteria (BIC) for 1949 Racine Continuous Interviewed Residents with Block Data (N=323) Optimal # of Groups BIC Cubic ZIP Model 2 -1816.65 3 -1747.74 4 False Convergence Table 5-18. Mean (Median) Posterior Proba bilities for Group Assignments: 1949 Racine Continuous Interviewed Residents with Block Data (N=323) Group Prob (G1) Prob (G2) Prob (G3) 25th Percentile 75th Percentile 1 (n=245) .98(.99) .02(.00) .00(.00) .98 .99 2 (n=66) .11(.01) .89(.98) .01(.00) .82 .99 3 (n=12) .00(.00) .00(.00) .99(.99) .99 .99 Note: Minimum and Maximum values presented for the group/prob(group) combination. Table 5-19. Coefficient Estimates for Eff ect of Sex (being Male) on Trajectory Group Membership: 1949 Racine Continuous Interviewed Residents with Block Data (N=323) Groupsa coefficient SE p-value Cubic ZIP Model 2 1.493 0.364 <0.001*** 3 1.887 0.790 0.016* *p<.05 **p<.01 ***p<.001 a Group 1 (the lowest rate group) is the reference category of comparison.

PAGE 132

132Table 5-20. Trajectory Gr oups by Sample and Cohort Continuous-Resident Samples Sample Size (n) Non-offenders AdolescentLimited/Peaked Offenders Low-Level Chronic Offenders High-Rate Chronic Offenders Late-Onset Offenders 1942 COHORT Continuous-Residents 633 463 (73.1%) 56 (8.8%) 82 (13.0%) 32 (5.1%) -Continuous-Residents (w/ Interview Data) 333 287 (86.2%) 26 (7.8%) -20 (6.0%) -Continuous-Residents (w/ Individual & Block Data) 235 165 (70.2%) 47 (20.0%) -23 (9.8%) -1949 COHORT Continuous-Residents 1297 740 (57.1%) 74 (5.7%) 424 (32.7%) 25 (1.9%) 34 (2.6%) Continuous-Residents (w/ Interview Data) 556 418 (75.2%) 114 (20.5%) -24 (4.3%) -Continuous-Residents (w/ Individual & Block Data) 323 245 (75.9%) 66 (20.4%) -12 (3.7%) -

PAGE 133

133Table 5-21. Mean Covariate Levels by Trajectory Group Membership and Cohort Variables 1942 G1 G2 G3 Mean Mean Mean F-test p-value 1949 G1 G2 G3 Mean Mean Mean F-test p-value Macro-level Social Disorganization -.18 .41 .50 10.65 <.001*** -.09 .28 .29 4.38 .01* Micro-level Male .31 .79 .91 33.79 <.001*** .43 .79 .83 17 .10 <.001*** Black .04 .08 .17 3.84 .23 .02 .23 .33 10.93 <.001*** Low SES Household 3.30 4.29 4.35 5.99 <.01** 2.99 3.76 5.58 14.34 <.001*** Stakes in Conformity 1.76 1.62 1.65 2.00 .13 1.62 1.59 1.2 5 2.89 .05* Family Stability (Natural Parents) .90 .72 .70 6.49 <.01** .94 .76 .67 12.96 <.001** Family Structure (No. of Children) 3.69 3.57 4.48 1.46 .24 3.80 4.16 5.17 6.11 .01* Differential association .22 .24 .17 0.74 .84 .24 .29 .17 0.57 .57 Formal Social Control 1.14 1.43 1.61 3.56 .03* 1.21 1.62 2 .00 7.12 <.001*** Moved .59 .60 .83 2.47 .08+ .48 .53 .58 0.38 .68 Number of Police Contacts .12 2.04 8.09 492.99 <.001*** .25 4.98 30.25 481.58 <.001*** +p<.10 *p<.05 **p<.01 ***p<.001

PAGE 134

134 Table 5-22. 1942 Racine Continuous Interviewed Residents with Block Data : Odds Ratios and Probabilities for Macro-Level Predic tors of Offending Group Membership Variables Adolescent-Limited/Peaked Offenders High-Rate Chronic Offenders Odds Prob. Odds Prob. Macro-level Social Disorganization 1.97 <.001*** 2.18 <.001*** Model Diagnostics Chi-Square= 18.30*** -Log Likelihood= -176.95 McFaddens R2= .06 Nagelkerke= .11 +p<.10 *p<.05 **p<.01 ***p<.001 Note. Reference category is group 1 (the non-offenders ). Table 5-23. 1942 Racine Continuous Interviewed Residents with Block Data : Odds Ratios and Probabilities for Micro-Level Predic tors of Offending Group Membership Variables Adolescent-Limited/Peaked Offenders High-Rate Chronic Offenders Odds Prob. Odds Prob. Micro-level Male 2.99 <.001*** 6.12 <.001*** Black 1.01 .97 1.27 .30 Low SES Household 1.38 .13 1.10 .72 Stakes in Conformity 0.90 .57 0.96 .89 Family Stability (Natural Parents) 0.62 <.01** 0.59 .05* Family Structure (No. of Children) 0.96 .80 1.37 .11 Differential association 1.01 .97 0.84 .57 Formal Social Control 1.13 .55 1.12 .71 Moved 1.11 .61 2.30 <.01** Model Diagnostics Chi-Square= 53.14*** -Log Likelihood= -143.38 McFaddens R2= .23 Nagelkerke= .39 +p<.10 *p<.05 **p<.01 ***p<.001 Note. Reference category is group 1 (the non-offenders ).

PAGE 135

135 Table 5-24. 1942 Racine Continuous Interviewed Residents with Block Data : Odds Ratios and Probabilities for Macro-level and Micr o-level Predictors of Offending Group Membership Variables Adolescent-Limited/Peaked Offenders High-Rate Chronic Offenders Odds Prob. Odds Prob. Macro-level Social Disorgan ization 1.50 .06+ 1.36 .38 Micro-level Male 2.87 <.001*** 5.79 <.001*** Black 1.01 .99 1.24 .44 Low SES Household 1.30 .27 1.04 .87 Stakes in Conformity 0.92 .65 0.98 .92 Family Stability (Natural Parents) 0.64 .01* 0.61 .06+ Family Structure (No. of Children) 0.95 .76 1.33 .14 Differential association 1.01 .99 0.83 .57 Formal Social Control 1.06 .79 1.06 .85 Moved 1.03 .88 2.15 .03* Model Diagnostics Chi-Square= 55.47*** -Log Likelihood= -141.56 McFaddens R2= .25 Nagelkerke R2= .41 +p<.10 *p<.05 **p<.01 ***p<.001 Note. Reference category is group 1 (the non-offenders ).

PAGE 136

136 Table 5-25. 1949 Racine Continuous Interviewed Residents with Block Data : Odds Ratios and Probabilities for Macro-Level Predic tors of Offending Group Membership Variables Adolescent-Limited/Peaked Offenders High-Rate Chronic Offenders Odds Prob. Odds Prob. Macro-level Social Disorganization 1.50 <.01** 1.48 .07+ Model Diagnostics Chi-Square= 7.96* -Log Likelihood= -207.61 McFaddens R2= .02 Nagelkerke R2= .04 +p<.10 *p<.05 **p<.01 ***p<.001 Note. Reference category is group 1 (the non-offenders ). Table 5-26. 1949 Racine Continuous Interviewed Residents with Block Data : Odds Ratios and Probabilities for Micro-Level Predic tors of Offending Group Membership Variables Adolescent-Limited/Peaked Offenders High-Rate Chronic Offenders Odds Prob. Odds Prob. Micro-level Male 2.31 <.001*** 2.40 .01* Black 1.38 .06+ 1.20 .64 Low SES Household 1.26 .18 2.52 <.001*** Stakes in Conformity 1.09 .62 1.13 .50 Family Stability (Natural Parents) 0.64 <.001*** 0.67 .11 Family Structure (No. of Children) 1.36 .04* 1.61 .14 Differential association 1.25 .16 1.01 .99 Formal Social Control 1.14 .43 1.14 .64 Moved 0.89 .47 0.79 .51 Model Diagnostics Chi-Square= 76.67*** -Log Likelihood= -169.86 McFaddens R2= .19 Nagelkerke R2= .31 +p<.10 *p<.05 **p<.01 ***p<.001 Note. Reference category is group 1 (the non-offenders ).

PAGE 137

137 Table 5-27. 1949 Racine Continuous Interviewed Residents with Block Data : Odds Ratios and Probabilities for Macro-level and Micr o-level Predictors of Offending Group Membership Variables Adolescent-Limited/Peaked Offenders High-Rate Chronic Offenders Odds Prob. Odds Prob. Macro-level Social Disorganization 1.19 .28 1.09 .54 Micro-level Male 2.33 <.001*** 2.36 .02** Black 1.30 .13 1.20 .64 Low SES Household 1.25 .21 2.53 <.001*** Stakes in Conformity 1.10 .60 0.79 .53 Family Stability (Natural Parents) 0.63 <.001*** 0.68 .15 Family Structure (No. of Children) 1.33 .05* 1.62 .13 Differential association 1.24 .19 1.01 .99 Formal Social Control 1.10 .58 1.17 .64 Moved 0.86 .37 0.78 .53 Model Diagnostics Chi-Square= 79.26*** -Log Likelihood= -169.32 McFaddens R2= .20 Nagelkerke R2= .32 +p<.10 *p<.05 **p<.01 ***p<.001. Note. Reference category is group 1 (the non-offenders ).

PAGE 138

138 0 0.5 1 1.5 2 678910111213141516171819202122232425 AgePolice Contacts G1 pred G2 pred G3 pred G4 pred G1 avg G2 avg G3 avg G4 avg Figure 5-1. 1942 Racine Con tinuous Residents Group-Base d Trajectories (N=633)

PAGE 139

139 0 0.2 0.4 0.6 0.8 1 678910111213141516171819202122232425 AgePolice Contacts G1 pred G2 pred G3 pred G1 avg G2 avg G3 avg Figure 5-2. 1942 Racine Continuous Interviewed Residents Group-Based Tr ajectories (N=333)

PAGE 140

140 0 0.5 1 1.5 678910111213141516171819202122232425 AgePolice Contacts G1 pred G2 pred G3 pred G1 avg G2 avg G3 avg Figure 5-3. 1942 Racine Continuous Interviewed Residents with Block Data Group-Based Trajectories (N=235)

PAGE 141

141 0 2 4 6 678910111213141516171819202122232425 AgePolice Contacts G1 pred G2 pred G3 pred G4 pred G5 pred G1 avg G2 avg G3 avg G4 avg G5 avg Figure 5-4. 1949 Racine Con tinuous Residents Group-Base d Trajectories (N=1,297)

PAGE 142

142 0 0.5 1 1.5 2 2.5 3 678910111213141516171819202122232425 AgePolice Contacts G1 pred G2 pred G3 pred G1 avg G2 avg G3 avg Figure 5-5. 1949 Racine Continuous Interviewed Residents Group-Based Tr ajectories (N=556)

PAGE 143

143 0 2 4 6 678910111213141516171819202122232425 AgePolice Contacts G1 pred G2 pred G3 pred G1 avg G2 avg G3 avg Figure 5-6. 1942 Racine Continuous Interviewed Residents with Block Data Group-Based Trajectories (N=323)

PAGE 144

144 CHAPTER 6 DISCUSSION AND CONCLUSION Discussion This research began by seeking to investigate th e viability of integrated multi-level theory for explaining offense trajectories. It was expect ed that residing in a socially disorganized area early on in development would increase an indi viduals likelihood for being in an offender group as opposed to a non-offender group. Particularly, liv ing in a socially disorganized area should increase the odds for exhibiting a chronic-offending trajectory. Second, it was anticipated that certain individual and family-level factors would also elevate an individuals likelihood for being an offender. Similar to the expectations re garding the influence of living in a socially disorganized area, the belief was that the micro-level factor s would significantly distinguish adolescent-limited/peaked offenders and ch ronic offenders from the non-offenders. Furthermore, in accordance with Sampson and La ub (1993), it is was assumed that individuals who were more likely to repor t having experienced life tran sitions such as marriage and employment would be more li kely to be in the non-offe nder group and the adolescentlimited/peaked group. Third, it was hypothesize d that, once these factors were estimated simultaneously (macro-level and micro-level toge ther in the same model), they would each contribute to predicting an indivi duals odds for being in an a dolescent-limited/peaked offender and chronic offender group relative to a non-offe nder group. In addition, sex (being male) was expected to be significant for amplifying an individuals likelihood for being found in an offender group. Overall, the findings from this analysis yielded varying leve ls of support for the various theoretical frameworks used. Support was shown Moffitts (1993) taxonomy of offending in that all of the trajectory analyses identified a non-offender, an adol escent-limited/peaked offender,

PAGE 145

145 and a chronic offender group (with a declining tr ajectory in young adulthood). However, the semi-parametric Poisson based models also disc overed two groups that we re inconsistent with Moffitts prediction when relying on police cont act data from the largest samples of the 1942 (n=633) and 1949 (n=1,297) Racine continuous residents. A lo w-level chronic offending group was identified in both cohorts who were char acterized by an early onset of offending and maintained a fairly low level of offending (y et a non-zero rate) throughout the twenty year observation period (age 6-25). Comparatively, a late-onset offender group was discovered among the 1949 cohort who had virtually no invovlement in offending until late in adolescence/early adulthood but showed re latively high rates of offending throughout young adulthood. In addition, the adoles cent-limited/peaked group found in these analyses slightly differed from the group as theoretically discussed by Moffitt (1993) in that they were classified as adolescent-peaked and their offending patterns did not app ear to desist until around age 21. While these groups are atypical inso far as Moffitts expectations, they have been discovered and the source of discussion elsewhere (see D Unger et al., 1998; La nd, 2000; Piquero, 2007; Sampson & Laub, 1993). Also consistent with Moffitt and others (Rutter, 1983; Singer et al., 1968; Werry, 1986) that have argued that life-course-persistent o ffending may be observed almost exclusively among males and the prior research has shown that fema les typically show less i nvolvement in crime in either type of offender group (adolescent-limited offenders and life-course-persistent offenders) (see Piquero, 2007), all of the traj ectory results indicated that be ing male significantly increased an individuals likelihood of bei ng classified into any of the offender groups across all data sources (continuous re sidents, continuous interviewed residents, and continuous interviewed residents with block data ) and by cohort. The one exception was the late-onset offender group

PAGE 146

146 identified in the 1949 continuous resident anal ysis (N=1,297) where although there were more males in the group, sex was not found to be a si gnificant covariate for distinguishing this offender group from the non-offender group. The mlogit results also re-affirmed the importance of sex for predicting group membership through the consistent and robust effect of being male on increasing the odds by at least two times for an individual exhibiting either an adolescentlimited/peaked or a chro nic offending trajectory. The results only revealed limited support for Sampson and Laubs age-graded theory of informal social control and these findings were on ly evident at the bivariat e level. In both cohort analyses, the non-offender group ha d the highest mean score on the stakes in conformity index when compared with either offender group, but only in the 1949 cohort analys is was the stakes in conformity index significantly asso ciated with group membership. In line with the expectation of the influence of social disorganization, there was a minimal amount of evidence that living in a socially disorganized area in creased an individuals odds for being in an offender group. The bivariate findings suggested that Racine conti nuous residents, whose individual offending pattern could be clas sified into a adolesce nt-limited/peaked or chronic offending group-based trajec tory, tended to live in more soci ally disorganized areas than the non-offenders. The multivariate results c onfirmed that social disorganization was a significant predictor of group membership fo r both the 1942 and 1949 cohort. Although, living in a socially disorganized area failed to disti nguish the adolescent-limited/ peaked offenders from the chronic offenders. Despite the fact that the bivariate and mlog it (macro-level-only) results indicated that social disorganization was a di stinguishing feature of group memb ership, the findings from the mlogits, where both the macro-level and microlevel factors were estimated simultaneously,

PAGE 147

147 failed to find much support for the usefulness of an integrated theoretic al model for explaining group-based offending trajectories. Once the micr o-level factors (particu larly sex, natural two parent household, having moved, and low SES ho useholds depending on which cohort mlogit is being referred to) were added to the model, soci al disorganization failed to exert a significant independent effect on predicting group membership. Although, the various indicators of model fit (chi-square, log likelihood, and the psuedo R-square statistics) showed that th e integrated model was a better fit, it appeared that the infl uence of social disorgan ization on increasing an individuals likelihood for being in either an adolescent-limited/peaked or chronic offender group tended to be indirect in that it operated through individual and family-level factors. Overall, these findings tend to mirror those produced in prior multi-level studies that have shown that the addition of neighborhood le vel processes typically only in crease the variance explained by 1% to 4% (see Cheong & Raudenbush, 2000; Duncan & Raudenbush, 1999; Elliot et al., 1996; Simcha-Fagan & Schwartz, 1986), although ot her studies have begun to find more support for school contextual effects ranging from a 6% to 7% increase in the variance explained (Anderson, 2002; Welsh et al., 1999). Limitations As with all research, there are a series of data limitations that are worth mentioning. First, the most apparent limitation relates to the era wherein these individuals were raised. These individuals were either bor n during WWII (1942 cohort) or baby-boomers (i.e., 1949 cohort) who grew up in the immediate pos t-WWII era. They were living their early years during the 1960s, which is a time characterized by large social change includi ng the Civil Rights and Womens Movement, Vietnam, riots, etc. Thus, it is likely that some of these findings may not necessarily generalize to individu als who were socialized in diffe rent periods of time, say the 1980s or 1990s. Second, the area of Racine itself was a Midwestern industrial city with a

PAGE 148

148 relatively small homogenous population of predomin antly White residents, and it is not likely that the results from this community context a nd demographic would appl y to the larger more metropolitan areas of today or in communitie s/neighborhoods with an ethnically diverse population. Third, the substantial va riety of the census-level meas ures that are typically used currently in macro-level analyses were not av ailable at the block le vel in 1950. Among the measures that were accessible there was a rath er noticeable amount of completely missing data which precluded the ability to use mean substituti on to retain a greater number of individuals for the integrated theoretic al models. Extending from this lim itation, the generali zability of these findings may not be applicable or reproduced if different conceptualizat ions or aggregations were used to define the community/neighborhood such as cities, SMSAs, counties, police districts, etc. Fourth, the social disorganization index used in this study was limited in that it did not include a measure for ethnic heterogeneity as discussed by Shaw and McKay (1942). Although some research has shown that rates of delinquenc y are lower in ethnically homogenous areas (see Lander, 1954), there is typically less support for the measure wh en compared with poverty and residential instability (see Sampson, 1997). Additionally, Bursik (1988) and Sampson and Groves (1989) have argued that various structural dimensions such as local friendship ties, organizational density, and particip ation in local formal and informal voluntary organizations are among the most important factors that affect de linquency rates. More recently, Sampson et al. (1997) have argued that collective efficacy and the ability to form social capital matters more than social disorganization, and Sampson (1997) has also indicated th at one of the most important structural dimensions for control ling neighborhood delinquency is the communitys ability to supervise and control teenage peer groups, particularly gangs. Sampson and Groves

PAGE 149

149 (1989) also identified additional factors includi ng family disruption, urbanization, and rapid population change that can all impe de the ability of a community to create informal networks of social control for supervising teenage peer gr oups. Unfortunately, th ere were not any data available to measure these important processes, so caution should be taken when interpreting these results. Similarly, the stakes in conf ormity index used to assess Sampson and Laubs (1993) agegraded theory of informal social control wa s limited since it relied on a dichotomous index ranging from 0 to 2. Laub et al. (1 998) have argued that the quality or cohesiveness of the adult social bonds is what is more important than mere participation in these life transitions. Others, however, have argued that it is likely that being married a nd/or having a full-time job are probably indicators that some le vel of attachment has occurred in that it seems doubtful that these events would have happened at all if this was not the cas e (see Piquero et al., 2002). It is also important to note th at this research relied on offici al data of delinquency/crime as measured through police contacts. Despite the kn own strengths and weaknesses of official and self-report data when investigating criminal be havior (see Junger-Tas & Marshall, 1999), police contacts may be the better measure of tru e officially-based offending because they are typically closer in time to the actual criminal event than self-reported offending (see Hindelang et al., 1981; Thornberry & Krohn, 2003). Finally, there are some limitations of the traj ectory modeling procedur e as it specifically pertains to this study. There was evidence provided from the sensitivity analysis conducted that data loss did not equally affect all of the trajectory groups. Fo r instance, it appeared that once the multiple sources of data were merged toge ther that the model only discovered three groups and that the high-rate offending grou p (also referred to as chronics ) actually showed evidence of

PAGE 150

150 a sharp declining trajectory in early adulthood and even perhap s an indication of desistance by age 25. Unfortunately, given the data limitations and the reliance on only official measures of offending to model the trajectories this research is unable to an swer the question of whether or not this is true evidence of de sistance. Furthermore, there is also the possibility that these individuals in the high-rate offending group whic h accumulated a rather large number of police contacts over time had become incarcerated for some period of time by their young adulthood years, thus not providing them an opportunity to offend. This is also an empirical question that cannot be explored with this data. Lastly, once the individual and ne ighborhood level factors were estimated in multivariate models, the results indicated that while these risk/protective factors significantly distinguished the adoles cent-limited/peaked offenders and the chronicoffenders from the non-offenders, these same variab les (with the ex ception of sex) were not able to significantly distinguish the adolescent-limited/ peaked offenders from the chronic-offenders. Once again, this is likely the result of the data lo ss that appeared to diffe rentially affect offender groups. Directions for Future Research The findings from this research provided mixe d results for the various theories including Moffitts developmental typology and the saliency of sex when distinguishing offender groups, Sampson and Laubs age-graded theory of informal social control, and social disorganization theory. This research did generate some prelimin ary findings that can dire ct future research on integrated multi-level theoretical models. Sex was found to be rather robust for si gnificantly affecting group assignment when estimated as a covariate in the trajectory analyses. It also exerted a consis tent independent effect for significantly increasing the odds for group assignment to either an adolescent-limited/peaked offender group or a chronic offender group. Future research would benefit from estimating sex-

PAGE 151

151 disaggregated trajectories and then estimating ml ogit models to see whether or not the macrolevel and micro-level factors, such as those examined here, significantly differ across sex. Unfortunately, there were not enoug h individuals in the analyses re lied on in this study to permit the estimation of sex-specific trajectories and make use of all of the ri sk/protective factors in the mlogit models. Further studies employing considerab ly larger samples would permit this type of analysis. It is worth noting that there are other theories that may account for sex differences as well. One such example may be power-control theo ry (Hagan et al., 1985). It is possible that parents and society socializes males and females differently and this may have an impact on sexspecific pathways of offending. This po ssibility should be further explored. Larger samples with multi-level data would also allow future researchers to examine these effects on distinguishing between offender groups fo r groups that were unc overed in some of the trajectory analyses that were not consistent with Moffitts typology su ch as the low-level chronics and the late-onset offende rs. It may be the case that soci al disorganization matters more so for influencing low-level chronic offenders with low yet relatively stable offending rates over the life-course. Similarly, it seems reasonable to assume that the absence of the various adult social bonds identified by Sampson and Laub ( 1993) or a low level of attachment to the institutions of marriage and employment may serv e as explanations for a late-onset pattern of offending. Additionally, future studies investig ating the chronic patterns of offending should rely on data from longer observational periods pa rticularly through middle and late adulthood to see if these group-based offendi ng patterns identified in th is study continue beyond age 25. This research relied solely on an official m easure of crime to estimate offense trajectories and more specifically, police contacts It would be intere sting if future studies would attempt to reproduce these analyses using al ternative official measures of offending such as arrest or

PAGE 152

152 conviction records. Comparatively, it may also prove fruitful to examine the possible multi-level effects on life-course patterns of offending ba sed on self-reported criminal involvement. Arguably, the best test would be to include some type of compos ite measure that represented a combination of official and self-reported offending. As far as the relatively weak found support for the macro-level effects, it would likely be the case that future studies accounting for the processes identified by prior researchers (see Bursik, 1988; Sampson, 1997; Sampson & Groves, 1989; Sampson et al., 1997) such as family disruption (percentage of divorced or single-pa rent households), collective efficacy, social capital, supervision of teenage peer groups, and local participation in voluntary organizations may yield stronger support for macro-level influen ces on offending patterns and especially net of the effects of micro-level factors. Future mu tli-level studies should also make use of other macro-level theoretical constructs to investigate their relative impact on trajectories of criminal behavior including routine activities theo ry (see Cohen & Felson, 1979; Kennedy & Baron, 1993) and urban disadvantage perspectives (see Krivo & Peterson, 1996; Parker et al., 2005). With attention to these aforementioned suggess tions, it would seem that perhaps the ideal research design for best examining the hypothese s highlighted in this research would be a prospective longitudinal design which would identif y individuals at birth in different social contexts and follow them throughout their life-co urse and well into adulthood. Not only would official and self-report data be gathered on the individuals at various periods over their lifecourse, the optimal design would include both parent, teacher, and peer surveys throughout different developmental periods. Additionally, surveys administered to community residents over time would also allow for m easuring the informal networks of social control, collective efficacy, and social capital. Fina lly, if spousal surveys in add ition to self-reports were also

PAGE 153

153 adminstered in adulthood this would provide wort hwhile information in order to gauge the level of cohesiveness or attachment in the adult social bonds to better determine their effect in the desistence process. Nevetheless, it is obvious that collecting resear ch in this magnitude, for a lengthy period of time, involving a numerous amount of actors woul d require an enormous amount of monetary, individual, political, and community support, yet there are early indi cations that research similar to this magnitude (in a lesser degree) is possi ble and the preliminary findings leave hope for future research in integrative mutli-level mode ling (see related publications from the Project on Human Development in Chicago Neighborhoods). Implications Taken together, the findings from this research are intended to be preminary and to offer a starting point to guide and direct future research on theoretically integrated approaches. Given the minimal significance of social disorganization in the baseline models, it is important that future studies further investigate other social processes that have been recently discussed by neighborhood-level researchers as perhaps bei ng more important when estimating neighborhood effects (for review, see Sampson et al., 2002) befo re any concerted effort can be made toward imforming policy. Nevertheless, it is important here to at least highlight some of the possible implications that research in this area may provide in the future. Programs such as the Rochester Nurse-Home Visitation Program (see Olds et al., 1986) are likely to provide the most benefit in target ing the cumulative eff ects of both neighborhood disadvantage and factors particul arly associated with chronic offending (i.e. low birth weight, birth trauma). The Rochester program involved random assignement of expectant mothers in a small semi-rural community to one of three conditi ons: (1) pre and post natal visits; (2) pre-natal only visits; and (3) minimal standard services. The results indicated that the mothers who

PAGE 154

154 received the home visits faired significantly better than those that received the minimal amount of services and had higher levels of employment and fewer cases of child abuse and neglect. In an outcome evaluation of a similar program, Seitz et al. (1985) also found that pre and post-natal home visits to disadvantaged mo thers reduced the impact of adve rse risk factors and the children whose mothers participated in the program were rated as less aggressi ve and disruptive than children who were in the nonexperimental group (see Provence and Naylor, 1983 for an extended discussion of the Yale Ch ild Welfare Research Program). Programs targeting poor urban children during the preschool years ha ve shown to have a significant impact on reducing delin quent outcomes. Head Start is one of the most well-known of these type of initiatives and subsequent eval uations (see Darlington et al., 1980 ; Lee et al., 1990) have shown that the program was successful in producing short-term intellectual and social gains during the early preschool years a nd that Head Start youth were more likely to graduate high school. Other programs such as the Perry Preschool Project, which focused on early childhood education (ages 3 to 5), parent meetings, and home visits, have yielded a host of beneficial outcomes for program participants in cluding better grades, higher standardized test scores, higher graduation and employment rates, and lower delinquency and arrest rates (see Berreuta-Clement et al., 1984; Schweinhart & Weikart, 1988). In addition, programs implemented in the later elementary/early adol escent years specifically directed at improving problem-solving and social competency among poor youth have been associated with decreased levels of aggression and probl em behaviors (see Weissberg et al., 1981; Hawkins et al., 1991; Lochman et al., 1984; Sarason & Sarason, 1981; Hammond, 1991). Family intervention programs facilitating th e development of parental support networks and increasing the link between parents and comm unity services as well as improving parent-

PAGE 155

155 child interactions have provided mixed results for reducing delinquent involvement (see Dumas, 1989; Johnson & Walker, 1987; Hawkins et al., 1991) but Guerra (1997:290) has argued that these programs may not be as effective for urban minority disadvantaged youth because of cultural differences and the multiple stressors that plague the residents in these areas. Peer-based interventions like peer mentori ng (Goldstein, 1978) and redirec ting gang members toward more prosocial community activ ities (see Klein, 1971) have also yiel ded varying results in influencing an adolescents invovleme nt in delinquency. Residential relocation programs including the Moving to Opportunity for Fair Housing Demonstration (or MTO) and the Gautreaux Prog ram have been shown to be beneficial. Gautreaux youth who moved to the suburbs substa ntially increased their educational attainment and job opportunities (see Rosenbaum, 1995; Rosenbaum & Popkin, 1991) and recent MTO evaluations have indicated that residential relocation was significantly related to decreased involvement in violence, better overall health, and lower victimiz ation rates (see Katz et al., 2001; Ludwig et al., 2001). Leventhal and Br ooks-Gunn (2003) have also highlighted the significantly lower levels of depression, anxiety, and/or dependency problems among MTO youth when compared with youth who remained in the disadvantaged areas. Other initiatives that focus on increasing neighborhood supervisio n of teenage peer groups and promoting civic participation are likely to provide noticeable benefits as well (see Sa mpson, 1997). Overall, these programs reviewed above provide sugge stions for how future research targeting disadvantaged youth in disadvant aged areas can likely alleviat e the conditions that plague residents of these communities and decrease th e prevalence of life-course-persistent chronic patterns of offending.

PAGE 156

156 Final Thoughts Guerra (1997:300-301) perhaps best summarizes the direction for programmatic and policy initiatives in the following statement presented below if we as a community have any hopes for reducing the risk factors associated with both types of offending patterns identified by Moffitt, but particularly for prevention/inte rvention with chronic offenders: It is likely that the most effective interventions will address multiple causal factors, be carried out in multiple contexts, and be extende d in time over the course of development. Families, peers, schools, and communities must work in concert to provide opportunities for children to learn prosocial beha vior and to create environments where prosocial behavior is modeled and reinforced. It is important for researchers, policymakers, and even individuals to be aware that all of the abovementioned elements do not operate in isol ation from one another. There are individual, family, and community factors that all have infl uence on life-course pathways of development. Sampson et al. (2002:474) also challenges neig hborhood level researcher s to move beyond mere structural compositional measures of neighborhood effects and gi ve greater attention toward understanding "cultura l, normative, and collective action persp ectives that attach meaning to how residents frame their committment to places." Spatial dynamics are also important to investigating the influence of neighborhoods on offending patterns. It is possible that while youth may reside in a disdvantaged area, they may spend more time in an adjacent neighborhood where there are stronger informal social ties and collective supervision of teenage peer groups. Thus, while compositional measures of disadvant age may be related to crime rates, without measuring the influence of spatial proximity to more advantaged areas the effect of neighborhoods might be overestimated.

PAGE 157

157 It is clear that community and individualistic perspectives have there own shortcomings and limitations. Yet, it is also the case that perhaps by highlighting the strengths of each approach and seeking an unde rstanding of their interac tive influence on life-course developmental patterns of offending, future multilevel theoretical models can gain greater support within the discipline. Al though the results here failed to offer much evidence of the saliency of this approach, research in this ar ea is still an under-explor ed phenomenon. Future studies that are able to bette r capture the complex nature of macro-level and micro-level processes are needed before the usefulness of theoretical integration can be discounted.

PAGE 158

158 APPENDIX ZERO-ORDER CORRELATIONS BY COHORT Table A-1. 1942 Cohort 1 2 3 4 1. Social Disorganization Index 1.0 2. Stakes in Conformity Index -.07 1.0 3. Low SES Household 30 .06 1.0 4. Family Structure .12 -.01 19 1.0 Note. Significant effects (p<.05) are in italics. Table A-2. 1949 Cohort 1 2 3 4 1. Social Disorganization Index 1.0 2. Stakes in Conformity Index -.07 1.0 3. Low SES Household .29 -.22 1.0 4. Family Structure .18 .10 .16 1.0 Note. Significant effects (p<.05) are in italics.

PAGE 159

159 REFERENCES Aber, J.L., Gephart, M.A., Brooks-Gunn, J., & Conne ll, J.P. (1997). Development in context: Implications for studying neighborhood effect s. In J. Brooks-Gunn, G. Duncan, & J.L. Aber (eds.), Neighborhood poverty: Context and c onsequences for children. New York: Sage Foundation. Aguilar, B., Sroufe, L.A., Egeland, B., & Carl son, E. (2000). Distinguishing the early-onsetpersistent and adolescent-onset antisocial behavior types: From birth to 16 years. Developmental and Psychopathology, 12, 109-132. Akers, R.L. (1985). Deviant behavior: A social learning approach Belmont, CA: Wadsworth. Akers, R.L., Krohn, M.D., Lanza-Kaduce, L., & Radosevich, M. (1979). Social learning and deviant behavior: A specific test of a general theory. American Sociological Review, 44, 635-655. Akers, R.L. (1977). Deviant behavior: A social learning approach 2nd edition. Belmont, CA: Wadsworth. Allison, P.D. (1999). Multiple regression: A primer Thousand Oaks, CA: Sage Publications. Anderson, E. (1990). Streetwise: race, class, and change in an urban community Chicago: University of Chicago Press. Anderson, A.L. (2002). Individual an d contextual influences on delinquency: The role of the single-parent family. Journal of Criminal Justice 30 575. Agresti, A. (1990). Categorical data analysis New York: John Wiley and Sons. Bandura, A. (1982). The psychology of chance encounters and life paths. American Psychologists, 37, 747-755. Barrett, A.E., & White, H.R. (2002). Trajectories of gender role orientations in adolescence and early adulthood: A prospectiv e study of mental health effects of masculinity and femininity. Journal of Health and Social Behavior, 43, 451-468. Bartusch, D.R., Lynam, D.R., Moffitt, T.E., Silv a, P.A. (1997). Is age important? Testing a general versus a developmental theory of antisocial behavior. Criminology, 35, 13-48. Bellair, P.E. (1997). Social interaction and co mmunity crime: Examining the importance of neighbor networks. Criminology, 35 : 677-703. Berreuta-Clement, J.R., Schweinhart, L.J., Barne tt, W.S., Epstein, A.S., Weikart, D.P. (1984). Changed lives: The effects of the Perry Pr eschool Program on youths through age 19 Ypsilanti, MI: High/Scope Press. Berry, B.J., & Kasarda, J.D. (1977). Contemporary urban ecology. New York: Macmillan.

PAGE 160

160 Blackwell, B.S., & Piquero, A.R. (2005). On the relationships between gender, power control, self-control, and crime. Journal of Criminal Justice, 33, 1-17. Blumstein, A., & Cohen, J. (1979). Estimation of individual crime rates from arrest records. Journal of Criminal Law and Criminology, 70, 561-585. Blumstein, A., Cohen, J., & Hsieh, P. (1982). The duration of adult criminal careers. Final Report submitted to NIJ. Pittsburgh, PA: School of Urban and Public Affairs, CarnegieMellon University. Blumstein, A., Cohen, J., Roth, J. A., & Visher, C.A (eds.). (1986). Criminal careers and career criminals, Vol 1. Report of the Panel on Criminal Ca reers, National Research Council. Washington, D.C.: National Academy of Sciences. Blumstein, A. & Cohen, J. (1987). Ch aracterizing criminal careers. Science, 237 985-991. Blumstein, A., Cohen, J., & Farrington, D.P. (1988) Criminal career re search: Its value for criminology. Criminology, 26 1-35. Braithwaite, J. (1979). Inequality, crime, and public policy London: Routledge and Kegan Paul. Brame, B., Nagin, D., & Tremblay, R. (2001). Deve lopmental trajectories of physical aggression from school entry to late adolescence. Journal of Child Psychol ogy and Psychiatry, 42, 503. Broidy, L.M., Nagin, D.S., Tremblay, R.E., Bate s, J.E., Brame, R., Dodge, K.A., Fergusson, D.M., Horwood, J.L., Loeber, R., Laird, R., Lynam, D.L., Moffitt, T.E., Petit, G.S., & Vitaro, F. (2003). Developmental trajector ies of childhood disruptive behaviors and adolescent delinquency: a si x-site, cross-national study. Developmental Psychology, 39, 222-245. Bronfenbrenner, U., Moen, P., & Garbarino, J. (1984). Child, family, and community. In R. Parke (ed.), Review of child developmental research: Vol. 7: The family Chicago: University of Chicago Press. Brooks-Gunn, J., Duncan, G.J., Klebanov, P.K ., & Sealand, N. (1993). Do neighborhoods influence child and adolescent development? American Journal of Sociology, 99, 353395. Bursik, R.J. (1984). Urban dynamics a nd ecological studies of delinquency. Social Forces, 63, 393-413. Bursik, R.J. (1986). Ecological stability and dynami cs of delinquency. In A.J. Reiss & M. Tonry (eds.), Crime and Justice (vol. 66) Chicago: University of Chicago Press. Bursik, R.J., & Grasmick, H.G. (1993). Neighborhoods and crime: The di mensions of effective community control. New York: Lexington Books.

PAGE 161

161 Bushway, S., Piquero, A.R., Broidy, L.M., Cauffm an, E., & Mazerolle, P. (2001). Anempirical framework for studying desistance as a process. Criminology, 39, 492. Burgess, R., & Akers, R. (1966). A differential association-reinforcement theory of criminal behavior. Social Problems, 14, 128-147. Bushway, S.D., Thornberry, T.P., & Krohn, M.D. (2003). Desistance as a developmental process: A comparison of st atic and dynamic approaches. Journal of Quantitative Criminology, 19, 129-153. Byrk, A.S., & Raudenbush, S.W. (1987). Application of hierarchical linear models to assessing change. Psychology Bulletin, 101, 147-158 Byrk, A.S., & Raudenbush, S.W. (1992). Hierarchical Linear Models for social and behavioral research: Application a nd data analysis methods. Newbury Park, CA: Sage Publications. Caspi, A., Elder, G.H., & Bem, D.J. (1987). Mo ving against the world: Life-course patterns of explosive children. Developmental Psychology, 23, 308-313. Caspi, A, & Moffitt, T.E. (1993). The continuity of maladaptive behavior: From description to understanding in the study of antisocial beha vior. In D. Cicchetti & D. Cohen (eds.), Manual in the development of psychopathology. New York: John Wiley & Sons. Caspi, A., Moffitt, T.E., Stouthamer-Loeber, M., Krueger, R.F., & Schmutte, P.S. (1994). Are some people crime-prone? Replications of the personality-crime relationship across countries, genders, races, and methods. Criminology, 32, 163-195. Catalono, R.F., & Hawkins, J.D. (1996). The soci al development model: A theory of antisocial behavior. In D. Hawkins (ed.), Delinquency and crime: Current theories. Cambridge: Cambridge University Press. Chaiken, J., & Chaiken, M. (1982). Varieties of criminal behavior. Report R-2814-NIJ. Santa Monica, CA: Rand Corporation. Chase-Lansdale, P.L., Gordon, R.A., Brooks -Gunn, J., & Klebanov, P.K. (1997). Neighborhood and family influences on the intellectual and behavioral competence of preschool and early school-age children. In J. Brooks-G unn, G.J. Duncan, & J.L. Aber (eds.), Neighborhood poverty: Vol. 1. Context and consequences for children. New York: Sage Foundation. Cheong, Y.F., & Raudenbush, S.W. (2000). Measurem ent and structural models for childrens problem behaviors. Psychological Methods, 5 477. Chung, I., Hill, K.G., Hawkins, J.D., Gilchrist, L.D., & Nagin, D.S. (2002). Childhood predictors of offense trajectories. Journal of Research in Crime and Delinquency, 39, 60-90. Cleckley, H. (1976). The mask of sanity (5th ed.). St. Louis, MO: Mosby.

PAGE 162

162 Cohen, L.E., & Felson, M. (1979). Social changes and crime rate trends: A routine activities approach. American Sociological Review, 44 588-608. Cohen, L.E., & Land, K.C. (1987). Age structure and crime: Symmetry versus asymmetry and the projection of crime rates through the 1990s. American Sociological Review, 52 170183. Cohen, L.E., & Machalek, R. (1988). A ge neral theory of e xpropriative crime. American Journal of Sociology, 94, 465-501. Coleman, J.S. (1988). Soci al capital in the crea tion of human capital. American Journal of Sociology, 94 95-120. Coleman, J.S. (1990). Foundations of social theory Cambridge, MA: Harvard University Press. Cote, S., Zoccolillo, M., Tremblay, R.E., Nagi n, D., & Vitaro, F. (2001). Predicting girls' conduct disorder in adolescence from childhood trajectories of disruptive behaviors. Journal of the American Academy of Child and Adolescent Psychiatry, 40, 678-684. Cote, S., Tremblay, R.E., Nagin, D., Zoccolill o, M., Vitaro, F. (2002). The development of impulsivity, fearfulness, and helpfulness during childhood: Patterns of consistency and change in the trajectories of boys and girls. Journal of Child Psyc hology & Psychiatry & Allied Disciplines, 43, 609-618. Cramer, H. (1946). Mathematical models of statistics Princeton, NJ: Princeton University Press. Crane, J. (1991). The epidemic theory of ghettos and neighborhood eff ects on dropping out and teenage childbearing. American Journal of Sociology, 96, 1126-1159. Daly, K. (1998). Gender, crime, and criminology. In M. Tonry (ed.), The handbook of crime and punishment New York: Oxford University Press. Darlington, R.B., Royce, J.M., Snipper, A.S ., Murray, H.W., & Lazar, I. (1980). Preschool programs and later school competence of children from low-income families. Science, 208 202-204. De Li, S. (1999). Legal sanctions and youths status achievement: A longitudinal study. Justice Quarterly, 16 377-401. Donnellan, M.B., Ge, X., & Wenk, E. (2000). Cogn itive abilities in adolescent-limited and lifecourse-persistent criminal offenders. Journal of Abnormal Psychology, 109, 396-402. Dumas, J.E. (1989). Treating antis ocial behavior in children: Child and family approaches. Clinical Psychology Review, 9 197-222. Duncan, G.J. (1994). Families and neighbors as sources of disadvantage in the schooling decisions of White and Black adolescents. American Journal of Education, 103, 20-53.

PAGE 163

163 Duncan, G.J., Brooks-Gunn, J.P., & Klebanov, P. K. (1994). Economic deprivation and earlychildhood development. Child Development, 65, 296-318. Duncan, G.J., & Raudenbush, S.W. (1999). Assessing th e effect of context in studies of child and youth development. Educational Psychologist, 34 29. D'Unger, A.V., Land, K.C., McCall, P.L, & Na gin, D.S. (1998). How many latent classes of delinquent/criminal careers? Results from mixed Poisson regression analyses. American Journal of Sociology, 103, 1593-1630. D'Unger, A.V., Land, K.C., & McCall, P.L. ( 2002). Sex differences in age patterns of delinquent/criminal careers: Results from Poisson latent class analyses of the Philadelphia Cohort Study. Journal of Quantitative Criminology, 18, 349-375. Eggleston, E.P., Laub, J., & Sampson, R.J. (2004). Methodological sensitiviti es to latent class analysis of long-term criminal trajectories. Journal of Quantitative Criminology, 20, 126. Elder, G.H. (1975). Age differentiation and the life course. Annual Review of Sociology, 1, 165190. Elder, G.H. (1985). Perspectives on the lif e course. In Glen H. Elder, Jr. (ed.), Life course dynamics: Trajectories and transitions, 1968-1980 Ithaca: Cornell University Press. Elliott, D.S., Ageton, S.S., Huizinga, D., Knowles, B.A., & Canter, R.J. (1983). The prevalence and incidence of delinquent behavior: 1976. Boulder, CO: Behavioral Research Institute. Elliot, D.S., Huizinga, D., & Morse, B. (1986) Self-reported violent offendingA descriptive analysis of juvenile violent offenders and their offending careers. Journal of Interpersonal Violence, 1, 472-514. Elliot, D.S. (1994). Serious violent offenders: Onset, developmental course, and termination The American Society of Criminology 1993 address. Criminology, 32, 1-21. Elliott, D.S., Wilson, W.J., Huizinga, D., Sampson, R.J., Elliott, A. & Rankin, B. (1996). The effects of neighborhood disadvantag e on adolescent development. Journal of Research in Crime and Delinquency, 33 389. Ensminger, M.E., Lamkin, R.P., & Jacobson, N. (1996). School leav ing: A longitudinal perspective including neighborhood effects. Child Development, 67, 2400-2416. Entwisle, D.R., Alexander, K.L., & Olson, L.S. (1994). The gender gap in math: Its possible origins in neighborhood effects. American Sociological Review, 59, 822-838. Farrington, D.P. (1986a). Age and crime. In M. Tonry & N. Morris (eds.), Crime and Justice, Vol. 7 Chicago, IL: University of Chicago Press.

PAGE 164

164 Farrington, D.P. (1986b). Stepping stone s to adult criminal careers. In D. Olweus, J. Block, & M. Radke-Yarrow (eds.), Development of antisocial and prosocial behavior New York: Academic Press. Farrington, D.P. (1995). The development of o ffending and antisocial behaviour from childhood: Key findings from the Cambridge Study in Delinquent Development. Journal of Child Psychology and Psychiatry, 360, 929-964. Farrington, D.P. (1999). A crim inological research agenda for the next millennium. International Journal of Offender Therapy and Comparative Criminology, 43 154-167. Farrington, D.P. (2003). Developmental and lif e-course criminology: Key theoretical and empirical issues-The 2002 Sutherland Address Award. Criminology, 41 221-255. Farrington, D.P., & West, D.J. (1990). The Cambridge Study in deli nquent development: A prospective longitudinal survey of 411 male s, in H. Kerner & G. Kaiser (eds.), Criminality: Personality, behavior, and life history Berlin: Springer-Verlag. Farrington, D.P., Loeber, R., Elliott, D.S., Hawk ins, J.D., Kandel, D.B., Klein, M.W., McCord, J., Rowe, D.C., & Tremblay, R.E. (1990) Advancing knowledge about the onset of delinquency and crime. In B.B. Lahey & A.E. Kazdin (eds.), Advances in clinical child psychology New York: Plenum Press. Farrington, D.P., Sampson, R.J., & Wikstrom, P.H. (1994). Integrating individual and ecological aspects on crime Stockholm, Sweden: Nationa l Council for Crime Prevention. Felson, R.B., & Haynie, D.L. (2002). Pubertal deve lopmental, social factors, and delinquency among adolescent boys. Criminology, 40 967-988. Fergusson, D.M., Horwood, L.J., Caspi, A., Moffitt, T.E., & Silva, P.E. ( 1996). The (artefactual) transmission of reading disability: Psychom etric lessons in the study of stability and change in behavioral development. Developmental Psychology, 32, 132-140. Fergusson, D.M., Horwood, L.J., & Nagin, D.S. ( 2000). Offending trajectories in a New Zealand birth cohort. Criminology, 38, 525-551. Fergusson, D.M., & Horwood, L.J. (2002). Ma le and female offending trajectories. Development and Psychopathology, 14, 159-177. Gephart, M.A. (1997). Neighborhoods and communities as contexts for development. In J. Brooks-Gunn, G. Duncan, & J.L. Aber (eds.), Neighborhood poverty: Context and consequences for children New York: Sage Foundation. Glueck, S., & Glueck, E. (1950). Unraveling juvenile delinquency. Cambridge, MA: Harvard University Press. Glueck, S., & Glueck, E. (1968). Delinquents and nondelinquents in perspective. Cambridge, MA: Harvard University Press.

PAGE 165

165 Goetting, A. (1994). The parenting crime connection. Journal of Primary Prevention, 14, 167184. Goldstein, A.P., Sherman, M., Gershaw, N.J ., Sprafkin, R.P., & Glick, B. (1978). Training aggressive adolescents in prosocial behavior. Journal of Youth and Adolescence, 7 7392. Goldstein, H. (1995). Multilevel statistical models 2nd edition. London, Edward Arnold. Goldsmith, H.H., Bradshaw, D.L., & Reiser-Danne r, L.A. (1984). Temperament as a potential developmental influence on attachment. In J.V. Lerner & R.M. Lerner (eds.), Temperament and social interaction during infancy and childhood San Francisco: Jossey Bass. Gordon, R.A. (1968). Issues in multiple regression. American Journal of Sociology, 73, 592-616. Gottfredson, M., & Hirschi, T. (1990). A general theory of crime Palo Alto: Stanford University Press. Gottfredson, D.C, McNeil, R.J., & Gottfreds on, D.G. (1991). Social area influences on delinquency: A multi-level analysis. Journal of Research in Crime and Delinquency, 28 197-226. Granovetter, M. (1973). The strength of weak ties. American Journal of Sociology, 78, 1360-1380. Grasmick, H., Tittle, C.R., Bursik, R.J., & Arnekl ev, B. (1993). Testing the core implications of Gottfredson and Hirschis general theory of crime. Journal of Research in Crime and Delinquency, 30, 5-29. Greenberg, D.F. (1985). Age, crim e, and social explanation. American Journal of Sociology, 91 1-21. Greene, W.H. (1990). Econometric analysis. New York: Macmillan. Griffiths, E., & Chavez, J.M. (2004). Communiti es, street guns and homicide trajectories in Chicago, 1980-1995: Merging methods for exam ining homicide tren ds across space and time. Criminology, 42, 941-978. Guerra, N.G. (1997). Intervening to prevent chil dhood aggression in the inner city. In J. McCord (ed.), Violence and childhood in the inner city Cambridge, UK: University of Cambridge Press. Hagan, J. (1991). Destiny and drift: Subcultural preferences, status attainments, and the risks and rewards of youth. American Sociological Review, 56, 567-582.

PAGE 166

166 Hagan, J., Gillis, A.R., & Simpson, J. (1985). Th e class structure of ge nder and delinquency: Toward a power-control theory of common delinquent behavior. American Journal of Sociology, 90 1151-1178. Halpern-Felsher, B., Connell, J.P., Spencer, M.B., Aber, J.L., Duncan, G.J., Clifford, E., Crichlow, W., Usinger, P., & Cole, S.S. ( 1997). Neighborhood and family factors predicting educational risk and attainment in African Am erican and White children and adolescents. In J. Brooks-Gunn, G.J. Duncan, & J.L. Aber (eds.), Neighborhood poverty: Vol. 1. Context and consequences for children. New York: Sage Foundation. Hamparian, D.M., Schuster, R., Dinitz, S., & Conrad, J. (1978). The violent few: A study of dangerous juvenile offenders. Lexington, MA: Lexington Books. Hamparian, D.M., Schuster, R., Davis, J., & White, J. (1985). The young criminal careers of the violent few. Report to the U.S. Department of Jus tice, Office of Juvenile Justice and Delinquency Prevention. Harpur, J., & Hare, R.D. (1994). Assessment of psychopathology as a function of age. Journal of Abnormal Psychology, 103, 604-609. Hawkins, J.D., Von Cleve, E., & Catalano, R. F. (1991). Reducing earl y childhood aggression: Results of a primary prevention program. Journal of the American Academy of Child and Adolescent Psychiatry, 30 208-217. Hawkins, D. (ed.). (1995). Ethnicity, race, and crime: Persp ectives across time and space. Albany, NY: State University of New York Press. Heilbron, D. (1994). Zero-altered and other regression models for count data with added zeros. Biometrical Journal, 36 531-547. Heitgerd, J.L., & Bursik, R.J. (1987). Extr a-community dynamics and the ecology of delinquency. American Journal of Sociology, 92 775-787. Hindelang, M.J., Hirschi, T., & Weis, J.G. (1981). Measuring delinquency Beverly Hills, CA: Sage Publications. Hipwell, A.E., Loeber, R., Stouthamer-Loe ber, M., Keenan, K., White, H.R., & Kroneman, L. (2002). Characteristics of girls with early-onset disr uptive and antisocial behaviour. Criminal Behaviour and Mental Health, 12, 99. Hirschi, T., & Hindelang, M. J. (1977). Inte lligence and delinquency: A revisionist review. American Sociological Review, 42, 571-587. Hirschi, T., & Gottfredson, M. (1983). Age and the explanation of crime. American Journal of Sociology, 89 552-584.

PAGE 167

167 Hirschi, T., & Gottfredson, M. (1985). Age a nd crime, logic and scholarship: Comment on Greenberg. The American Journal of Sociology, 91 22-27. Horney, J., Osgood, D.W., & Marshall, I.H. (1995). Criminal careers in the short-term: Intraindividual variability in crime and it s relation to local life circumstances. American Sociological Review, 60, 655-673. Huesmann, L.R., Eron, L.D., Lefkowitz, M.M., & Wa lder, L.O. (1984). Stability of aggression over time and generations. Developmental Psychology, 20, 1120-1134. Huff-Corzine, L., Corzine, J., & Moore, D.C. (1986). Southern expos ure: Deciphering the Souths influence on homicide rates. Social Forces, 64, 906-924. Jencks, C., & Mayer, S.E. (1990). The soci al consequences of growing up in a poor neighborhood. In L. Lynn & M. McGeary (eds.), Inner-city poverty in the United States. Washington: National Academy Press. Jessor, R., Donovan, J.E., & Costa, F.M. (1991). Beyond adolescence: Problem behavior and young adult development. New York: Cambridge University Press. Johnson, D.L., & Walker, T. (1987). The primary prevention of behavior problems in MexicanAmerican children. American Journal of Community Psychology, 15 375-385. Jones, B.L., Nagin, D.S., & Roeder, K. (2001). A SAS procedure based on mixture models for estimating developmental trajectories. Sociological Methods and Research, 29, 374-393. Junger-Tas, J., & Marshall, I.H. (1999). The se lf-report methodology in crime research. In M. Tonry (ed), Crime and Justice (vol. 25) Chicago, IL: University of Chicago Press. Katz, L.F., Kling, J.R., & Liebman, J.B. (2001). M oving to opportunity in Boston: Early results of a randomized mobility experiment. Quarterly Journal of Economics, 116 607. Kazemian, L., & Farrington, D.P. (2006). Exploring residual career length and residual number of offenses for two generations of repeat offenders. Journal of Research in Crime and Delinquency, 43, 89-113. Kennedy, L.W., & Baron, S.W. (1993). Routine activ ities and a subculture of violence:A study of violence on the street. Journal of Research in Crime and Delinquency, 20 88-112. Kim, J., & Mueller, C.W. (1978). Introduction to factor analysis : What it is and how to do it. Thousand Oaks, CA: Sage Publications. King, G. (1989). Event count models for inte rnational relations: Generalizations and applications. International Studies Quarterly 33, 123-147.

PAGE 168

168 Klebanov, P.K., Brooks-Gunn, J., Chase-Lands dale, P., & Gordon, R.A. (1997). Are neighborhood effects on young children mediated by features of the home environment? In J. Brooks-Gunn, G.J. Duncan, & J. Aber (ed.), Neighborhood pove rty: Context and consequences for children. Sage: New York, NY. Klein, M.W. (1971). Street gangs and street workers Englewood Cliffs, NJ: Prentice Hall. Kling, J.R., Liebman, J.B., & Katz, L.F. (2007). E xperimental analysis of neighborhood effects. Econometrica, 75 83. Kornhauser, R.R. (1978). Social sources of delinquency Chicago: University of Chicago Press. Kratzer, L., & Hodgins, S. (1999). A typology of offenders: A test of Moffitts theory among males and females from childhood to age 30. Criminal Behavior and Mental Health, 9, 57-73. Krivo, L.J., & Peterson, R.D. (1996). Extremely disadvantaged neighborhoods and urban crime. Social Forces, 75 619-650. Kroneman, L., Loeber, R., & Hipwell, A.E. (200 4). Is neighborhood context differently related to externalizing problems and delinq uency for girls compared with boys? Clinical Child and Family Psychological Review, 7 109-122. Kupersmidt, J.B., Griesler, P.C., DeRozier, M.E., Patterson, C.J. & Davis, P.W. (1995). Childhood aggression and peer relations in the context of family and neighborhood factors. Child Development, 66, 360-375. Lambert, D. (1992). Zero-inflated Poisson regr ession, with an application to defects in manufacturing. Technometrics 34, 1-14. Land, K. (2000). Influence of neighborhood, peer, and fa mily context: Trajectories of delinquent/criminal offending across the life course, Final Report Washington, DC: National Institute of Justice. Land, K.C., McCall, P.L., & Cohen, L.E. (1990). Stru ctural covariates of homicide rates: Are there any invariances across time and social space? American Journal of Sociology, 95, 922-963. Land, K.C., McCall, P.L., & Nagin, D.S. (1996). A comparison of Poiss on, Negative Binomial, and Semiparametric Mixed Poisson Regression Models with empirica l applications to criminal careers research. Sociological Methods and Research, 24, 387-442. Land, K.C., & Nagin, D.S. (1996). Micro-level mode ls of criminal careers: A synthesis of The criminal careers and life course approach es via Semiparametric Mixed Poisson Regression Models, with empirical models. Journal of Quantitative Criminology, 12, 163-191.

PAGE 169

169 Lander, B. (1954). Toward and understanding of juvenile delinquency New York: Columbia University Press. Lash, T., Sigal, H., & Dudzinski, D. (1980). State of the child: New York City New York: Foundation for Child Development. Laub, J.H., & Sampson, R.J. (1988). Unravelling families and delinquency: A reanalysis of the Gluecks' data. Criminology, 26 355-380. Laub, J.H., & Sampson, R.J. (1993). Turning points in the life course: Why change matters to the study of crime. Criminology, 31, 301-325. Laub, J.H., Nagin, D.S., & Sampson, R.J. (1998). Tr ajectories of change in criminal offending: Good marriages and the desistance process. American Sociological Review, 63, 225-238. Laub, J.H., & Sampson, R.J. (2003). Shared beginnings, divergent li ves: Delinquent boys to age 70. Cambridge, MA: Harvard University Press. Lee, V.E., Brooks-Gunn, J., Schnur, E., & Liaw, F. (1990). Are Head Start effects sustained? A longitudinal follow-up comparison of disa dvantaged children attending Head, no preschool, and other preschool programs. Child Development, 61 495-507. Lempers, J., Clark-Lempers, D., & Simons, R. (1989). Economic hardship, parenting and distress in adolescence. Child Development, 60 25-49. Leventhal, T. & Brooks-Gunn, J. (2000). The neighborhoods they live in: Effects of neighborhood residence on child and adolescent outcomes. Psychological Bulletin, 126, 309-337. Leventhal, T., & Brooks-Gunn, J. (2003). Moving to opportunity: An experimental study of neighborhood effects on mental health. American Journal of Public Health, 93, 15761582. Lindstrom, P. (1996). Family interaction, ne ighborhood context and deviant behaviour. Studies in Crime and Crime Prevention, 5, 113-119. Lochman, J.E., Burch, P.R., Curry, J.F., & Lamp ron, L.B. (1984). Treatment and generalization effects of cognitive behavioral and goal setting interventions with aggressive boys. Journal of Consulting and Clinical Psychology, 52 915-916. Loeber, R. (1982). The stability of antisocial child behavior: A review. Child Development, 53, 1431-1446. Loeber, R., & Wikstrom, P. (1993). Individual pathways to crime in different types of neighborhoods. In Integrating individual and ecological aspects of crime edited by D.P. Farrington, R.J. Sampson, & P. Wikstrom Stockholm: Swedish National Council for Crime Prevention.

PAGE 170

170 Loeber, R., Burke, J.D., Lahey, B.B., Winters, A., & Zera, M. (2000). Oppositional defiant and conduct disorder: A review of the past ten years, part 1. Journal of the American Academy of Child and Adolescent Psychiatry, 39, 1468-1484. Loftin, C., & Hill, R.R. (1974). Regional subculture and homicide: An examination of the GastilHackney thesis. American Sociological Review, 39, 714-724. Loftin, C., & Parker, R.N. (1985). An errors-in-va riable model of the effect of poverty on urban homicide rates. Criminology, 23, 269-287. Loehlin, J.C. (1989). Partitioning environmenta l and genetic contribu tions to behavioral development. American Psychologist, 44, 1285-1292. Long, J.S. (1997). Regression models for categorical and limited dependent variables Thousand Oaks, CA: Sage. Ludwig, J., Duncan, G.J., & Hirschfield, P. (2001). Urban poverty and juvenile crime: Evidence from a randomized housing-mobility experiment. Quarterly Journal of Economics, 116, 655. Lynam, D.R., Moffitt, T.E., & Stouthamer-Loebe r, M. (1993). Explaining the relationship between IQ and delinquency: Clas s, race, test motivation, sch ool failure, or self-control. Journal of Abnormal Psychology, 102, 187-196. Lynam, D.R, Caspi A., Moffitt, T.E., Wikstr om, P.O., Loeber, R., & Novak, S. (2000). The interaction between impulsivity and neighbor hood context on offending: the effects of impulsivity are stronge r in poorer neighborhoods. Journal of Abnormal Psychology, 109, 563-574. Matsueda, R., & Anderson, K. (1998). Th e dynamics of delinquent peers and delinquentbehavior. Criminology, 36 269-308. Maughan, B. (2005). Developmental trajectory modeling: A view from developmental psychopathology. The ANNALS of the American Ac ademy of Political and Social Science, 602, 118-130. Maume, M., Ousey, G., & Beaver, K. (2005). Cut ting the grass: A reexamination of the link between marital attachment, delinquent pe ers and desistance from marijuana use. Journal of Quantitative Criminology, 21, 27-53. Mazzerole, P, Brame, R., Piquero, A.R., & D ean, C. (2001). Onset age, persistence, and offending versatility: Comparisons across gender. Criminology, 38, 1143-1172. McArdle, J.J., & Epstein, D. ( 1987). Latent growth curves within developmental structural equation models. Child Development, 58, 110-113. McCord, W., & McCord, J. (1959). Origins of crime: A new evaluation of the CambridgeSomerville study. New York, NY: Cambridge University Press.

PAGE 171

171 McDermott, S., & Nagin, D.S. (2001). Same or different? Comparing offender groups and covariates over time. Sociological Methods and Research, 29, 282-318. McDermott, S., & Nagin, D. (1998). Same or different over time? In S.O White (ed.), Handbook of Law and Social Science: Youth and Justice New York: Plenum Press. Meredith, W., & Tisak, J. (1990). Latent curve analysis. Psychometrika, 54, 557-585. Messner, S., Krohn, M., & Liska, A. (1989). Theoretical integration in the study of deviance and crime: Problems and prospects Albany, NY: State University of Albany Press. Moffitt, T.E. (1990). Juvenile delinquency a nd attention-deficit disorder: Developmental trajectories from age 3 to 15. Child Development, 61, 893-910. Moffitt, T.E. (1993). Adolescentlimited and life-course-persist ent antisocial behavior: A developmental taxonomy. Psychological Review, 100, 674-701. Moffitt, T.E. (1994). Natural histories of deli nquency. In E.G.M. Weitekamp & Hans-Jurgen Kerner (eds.), Cross-national longitudinal resear ch on human development and criminal behavior. Dordrecht, the Netherlands: Kluwer Academic. Moffitt, T.E. (1997a). Neuropsychology, antisocia l behavior, and neighborhood context. In J. McCord (ed.), Violence and childhood in the inner city Cambridge, UK: University of Cambridge Press. Moffitt, T.E. (1997b). Adolescence-limited and life-course-persistent offending: A complementary pair of developmenta l theories. In T.P Thornberry (ed.), Developmental theories of crime and delinquency: Advances in criminological theory, Volume Seven New Brunswick, NJ: Transaction Publishers. Moffitt, T.E. (2003). Life-course-persistent a nd adolescent-limited antisocial behavior: A tenyear research review and a research agenda. In B.B. Lahey, T.E. Moffitt, & A. Caspi (eds.), Causes of conduct disorder and delinquency. New York: Guilford Press. Moffitt, T.E. (2005). Life-course-persistent versus adolescence-limited antisocial behavior. In D. Cicchetti & D. Cohen (eds.), Developmental psychopathology (2nd ed.). New York: Wiley. Moffitt, T.E., & Henry, B. (1989). Neuropsychological assessment of executive functions in selfreported delinquents. Development and Psychopathology, 1, 105-118. Moffitt, T.E., & Henry, B. (1991). Neuropsychol ogical studies of juvenile delinquency and violence: A review. In J. Milner (ed.), The neuropsychology of aggression Norwell, MA: Kluwer Academic. Moffitt, T.E., Lynam, D.R., & Silva, P.A. (1994) Neuropsychological test s predicting persistent male delinquency. Criminology, 32, 277-300.

PAGE 172

172 Moffitt, T.E., & Caspi, A. (2001). Childhood predicto rs differentiate life-course persistent and adolescent-limited antisocial path ways among males and females. Developmental and Psychopathology, 13, 355-375. Moffitt, T.E., Caspi, A, Rutter, M., & Silva, P.A. (2001). Sex differences in antisocial behavior: Conduct disorder, delinquency, and violen ce in the Dunedin Longitudinal Study. Cambridge, UK: Cambridge University Press. Nagin, D.S. (1999). Analyzing developmental tr ajectories: A semiparametric, group-based approach. Psychological Methods, 4, 139. Nagin, D.S. (2004). Response to Methodological sens itivities to latent class analysis of longterm criminal career trajectories. Journal of Quantitative Criminology, 20, 27-35. Nagin, D.S. (2005). Group-based modeling of development Cambridge, MA: University of Harvard Press. Nagin, D.S., & Paternoster, R. (1991). On the re lationship of past to future delinquency. Criminology, 29, 163-189. Nagin, D.S., & Farrington, D.P. (1992). The onset and persistence of offending. Criminology, 30, 301-523. Nagin, D.S., & Land, K.C. (1993). Age, crimin al Careers, and popul ation heterogeneity: Specification and estimation of a nonpa rametric, mixed Poisson model. Criminology, 31, 327-362. Nagin, D.S., Farrington, D.P., & Moffitt, T.E. (19 95). Life-Course trajector ies of different types of offenders. Criminology, 33, 111-140. Nagin, D.S., & Waldfogel, J. (1998). The effect of conviction on income through the life cycle. International Review of Law and Economics, 18, 25-40. Nagin, D.S., & Tremblay, R.E. (2001). Analyzing developmental trajectori es of distinct but related behaviors: A group-based method. Psychological Methods, 6, 18. Nagin, D.S., & Tremblay, R.E. (2005a). Developm ental trajectory groups: Fact or a useful statistical fiction. Criminology, 43 873-904. Nagin, D.S., & Tremblay, R.E. (2005b). From seduc tion to passion: A response to Sampson and Laub. Criminology, 43, 915-918. Nagin, D.S., & Tremblay, R.E. (2005c). What ha s been learned from group-based trajectory modeling? Examples from physical aggr ession and other problem behaviors. The ANNALS of the American Academy of Political and Social Science, 602, 82-117.

PAGE 173

173 Nagin, D.S., & Tremblay, R.E. (2005d). Furthe r reflections on modeling and analyzing developmental trajectories: A re sponse to Maughan and Raudenbush. The ANNALS of the American Academy of Political and Social Science, 602, 145-154. OBrien, R.M., Stockard, J., & Isaacson, L. (1999) The enduring effects of cohort characteristics on age-specific homicide rates, 1960-1995. The American Journal of Sociology, 104 1061-1095. Olds, D.L., Henderson, C.R., Tatelbaum, R., & Ch amberlin, R. (1986). Improving the delivery of prenatal care and outcomes of pregnancy: A randomized trial of nurse-home visitation. Pediatrics, 77 16-28. Orr, L., Feins, J.D., Jacob, R., Beecroft, E., Sa nbonmatsu, L., Katz, L.F., Liebman, J.B., & Kling, J.R. (2003). Moving to Opportunity: Interim impact evaluations. Washington, DC: U.S. Department of Housing and Urban Developm ent, Office of Policy Development and Research. Pagani, L.S., Tremblay, R.E., Nagin, D.S., Zoccolil lo, M., Vitaro, F., & McDuff, P. (2004). Risk factor models for adolescent verbal a nd physical aggression toward mothers. International Journal of Behavioral Development, 28, 528-537. Park, R.E., Burgess, E.W. (1925). Introduction to the science of sociology Chicago: University of Chicago Press. Parker, R.N., & Smith, M.D. (1980). Deterr ence, poverty, and type of homicide. American Journal of Homicide, 85, 614-624. Parker, K.F., Stults, B.J., & Rice, S.K. (2005). Racial Threat, concentr ated disadvantage and social control: Considering the macro-le vel sources of variation in arrests. Criminology, 43 1111-1134. Paternoster, R., & Brame, R. (1997). Multiple ro utes to delinquency? A test of developmental and general theories of crime. Criminology, 35, 49-84. Paternoster, R., Dean, C.W., Piquero, A.R., Mazerolle, P., & Brame, R. (1997). Generality, continuity, and change in offending. Journal of Quantitative Criminology, 13, 231-266. Petersilia, J., Greenwood, P.W., & Lavin, M. (1977). Criminal careers of habitual felons. Report R-2144-DOJ. Santa Monica, CA: Rand Corporation. Peterson, M.A., & Braiker, H.B. (1980). Doing crime: A survey of California prison inmates. Report R-2200-DOJ. Santa Moni ca, CA: Rand Corporation. Peeples, F., & Loeber, R. (1994). Do individu al factors and neighborhoo d context explain ethnic differences in juvenile delinquency? Journal of Quantitative Criminology, 10, 141-157.

PAGE 174

174 Piquero, A.R. (2007). Taking stock of developmental trajectories of criminal activity over the life course. In A. Liberman (ed.), Longitudinal research on crime and delinquency New York: Springer. Piquero, A.R., & Brezina, T. (2001). Testing Moffitts account of adolescence-limited delinquency. Criminology, 39, 353-370. Piquero, A.R., Blumstein, A., Brame, R., Haapanen, R., Mulvey, E.P., & Nagin, D.S. (2001). Assessing the impact of exposure time and incap acitation on longitudinal trajectories of criminal offending. Journal of Adolescent Research, 16, 54. Piquero, A.R., & Buka, S. (2002). Linking juvenile and adult patterns of cr iminal activity in the Providence Cohort of the National Collaborative Perinatal Project. Journal of Criminal Justice, 30, 259-272. Piquero, A.R., Brame, R., Mazzerole, P., & Haapanen, R. (2002). Crime in emerging adulthood. Criminology, 40, 137-169. Piquero, A.R., Farrington, D.P., & Blumstein, A. (2003). The criminal career paradigm. Crime and Justice, 30, 359-506. Piquero, A.R., & White, N. (2003). On the rela tionship between cognitiv e abilities and lifecourse-persistent offending among a sample of African Americans: A longitudinal test of Moffitt's hypothesis. Journal of Criminal Justice, 31, 399-409. Piquero, A.R., & Moffitt, T.E. (2004). Life-cour se-persistent offending. In J.R. Adler (ed.), Forensic Psychology: Concepts, debates, and practices. Portland, OR: Willan Publishing. Piquero, A.R., Brame, R., & Lynam, D. (2004). Studying career length through early adulthood among serious offenders. Crime & Delinquency, 50, 412-435. Piquero, A.R., Brame, R., & Moffitt, T.E. (2005). Extending the study of continuity and change: Gender differences in the linkage be tween adolescent and adult offending. Journal of Quantitative Criminology, 21, 219-243. Piquero, A.R., Brezina, T., & Turner, M.G. ( 2005). Testing Moffitts ac count of delinquency abstention. Journal of Research in Crime and Delinquency, 42, 27-54. Piquero, A.R., Moffitt, T.E., & Lawton, B. (2005). Race and Crime: The contribution of individual, familial, and neighborhood level risk factors to life-course-persistent offending. In D. Hawkins & K. Kempf-Leonard (eds.), Our children, your children: Confronting race and ethnicity in th e American juvenile justice system Chicago, IL: University of Chicago Press. Plomin, R. (1990). The role of inheritance in behavior. Science, 248, 183-188. Plomin, R., Chipuer, H.M., & Loehlin, J.C. (199 0). Behavioral genetics and personality. In L.A. Pervin (ed.), Handbook of personality theory and research New York: Guilford Press.

PAGE 175

175 Pratt, T.C., Turner, M., & Piquer o, A.R. (2005). Parental sociali zation and community context: A longitudinal analysis of the struct ural sources of low self-control. Journal of Research in Crime and Delinquency, 41, 219-243. Pratt, T.C., & Cullen, F.T. (2005). Assessing macrolevel predictors of theories of crime: A meta-analysis. In M. Tonry (ed.), Crime and Justice (vol. 32) Chicago: University of Chicago Press. Provence, S., & Naylor, A. (1983). Working with disadvantaged parents and children: Scientific issues and practice New Haven, CT: Yale University Press. Raudenbush, S.W. (2005). How do we study "what happens next"? The ANNALS of the American Academy of Political and Social Science, 602, 131-144. Reiss, A.J., & Lewis, A.L. (1961). The distributi on of juvenile delinquency in the social class structure. American Sociological Review, 26, 720-732. Reiss, A.J. (1986). Why are communities important in understanding crime? In A.J. Reiss & M. Tonry (eds.), Communities and Crime Chicago: University of Chicago Press. Reiss, A.J. (1989). Ending criminal careers Prepared for the final report of the DesistancePersistance Working Group of the Progr am on Human Development and Criminal Behavior. MacArthur Foundation and National Institute of Justice. Roncek, D.W. (1981). Dangerous places: Crime and residential environment. Social Forces, 60, 74-96. Rosenbaum, J.E. (1995). Changing the geograp hy of opportunity by expanding residential choice: Lessons from the Gautreaux Program. Housing Policy Debate, 6, 231-269. Rosenbaum, J.E., & Popkin, S. (1991). Employme nt and earnings of low-income Blacks who move to middle-class suburbs. In C. Jencks & P. Peterson (eds.), The urban underclass Washington, DC: Brookings Institution. Rowe, D.C., Vazsonyi, A.T., & Flannery, D.J. ( 1995). Sex differences in crime: Do means and within-sex variation have similar causes? Journal of Research in Crime and Delinquency, 32, 84. Rutter, M. (1983). Developmental neuropsychiatry New York, NY: Guilford. Rutter, M. (1989). Pathways from childhood to adult life. Journal of Child Psychology and Psychiatry, 30, 23-51. Sampson, R.J. (1987). Urban black violence: Th e effect of male joblessness and family disruption. American Journal of Sociology, 93 348-382. Sampson, R.J. (1995). The community. In J.Q. Wilson & J. Petersilia (eds.), Crime San Francisco, CA: ICS, Inc

PAGE 176

176 Sampson, R.J. (1997). The embeddedness of child and adolescent development: A neighborhood-level perspective on urban violence. In J. McCord (ed.), Violence and childhood in the inner city Cambridge, UK: University of Cambridge Press. Sampson, R.J. (2002). Transcending tradition: New directions in community research, Chicago style. Criminology, 40 213-230. Sampson, R.J., & Groves, W.B. (1989). Community structure and crime: Testing socialdisorganization theory. American Journal of Sociology, 94, 774-802. Sampson, R.J., Raudenbush, S.W., & Earls, F. (1997). Neighborhoods and violent crime: A multilevel study of collective efficacy. Science, 277, 918-924. Sampson, R.J., Laub, J.H. (1990). Crime and devian ce over the life c ourse: The salience of adult social bonds. American Sociological Review, 55, 609-627. Sampson, R.J., & Laub, J.H. (1992). Crim e and deviance in the life course. Annual Journal of Sociology, 18, 63-84. Sampson, R.J., & Laub, J.H. (1993). Crime in the making: Pathways and turning points through life. Cambridge, MA: Harvard University Press. Sampson, R.J., & Laub, J.H. (1995). Understa nding variability in lives through time: Contributions of life-course criminology. Studies on Crime and Crime Prevention, 4, 143-158. Sampson, R.J., & Laub, J.H. (1997). A life-course theory of cumulative disadvantage and the stability of delinquency. In T.P. Thornberry (ed.), Developmental theories of crime and delinquency (pp. 131-161). New Brunswi ck, NJ: Transaction. Sampson, R.J., Morenoff, J.D., & Gannon-Rowle y, T. (2002). Assessing neighborhood effects: Social processes and new directions in research. Annual Review of Sociology, 28 443478. Sampson, R.J., & Laub, J.H. (2003). Life-course desisters? Trajectories of crime among delinquent boys followed to age 70. Criminology, 41, 555-592. Sampson, R.J., & Raudenbush, S.W. (2004). Seei ng disorder: Neighborhood stigma and the social construction of broken windows. Social Psychological Quarterly, 67, 319-342. Sampson, R.J., Laub, J.H., & Eggleston, E.P. ( 2004). On the robustness and validity of groups: Response to Daniel Nagin. Journal of Quantitative Criminology, 20, 37-42. Sampson, R.J., & Laub, J.H. (2005a). Seductions of method: Rejoi nder to Nagin and Tremblays Developmental Trajectory Groups: Fact or Fiction. Criminology, 43, 905-914.

PAGE 177

177 Sampson, R.J., & Laub, J.H. (2005b). When pred iction fails: From crime-prone boys to heterogeneity in adulthood. The ANNALS of the American Academy of Political and Social Science, 602, 73-79. Sampson, R.J., & Laub, J.H. (2005c). A life-cour se view of the development of crime. The ANNALS of the American Academy of Political and Social Science, 602, 12-45. Sarason, I.G., & Sarason, B.R. (1981). Teaching cognitive and social skills to high school students. Journal of Clinical and Consulting Psychology, 49 908-918. Schreck, C.J., Fisher, B., & Miller J.M. (2004). The social contex t of delinquent peer networks and violent victimization. Justice Quarterly, 21 23-48. Schuerman, L., & Kobrin, S. (1986). Community caree rs in crime. In A.J. Reiss & M. Tonry (eds.), Crime and justice: A review of research Chicago: University of Chicago Press. Schweinhart, L.J., & Weikart, D.P. (1988). The High/Scope Perry preschool program. In R.H. Price, E.L. Cohen, R.P. Lorion, & J. Ramos-McKay (eds.), 14 ounces of prevention: A casebook for practicioners (pp. 53-65). Washington, DC: American Psychological Association. Seitz, V., Rosenbaum, L.K., & Apfel, N.H. (1985). Effects of family support intervention: A tenyear follow up. Child Development, 56 376-391. Shannon, L.W. (1988). Criminal career continu ity: Its social context New York: Human Sciences Press. Shannon, L.W. (1991). Changing patterns of delinquency and crime : A longitudinal study in Racine Boulder: Westview Press. Shannon, L.W. (1994). Juvenile delinquency and adult crim e, 1948-1977 [Racine, Wisconsin]: City ecological data [Computer file]. 2nd ICPSR versi on. Iowa City, IA: University of Iowa, Iowa Urban Community Research Center [producer], 1980. Ann Arbor, MI: Interuniversity Consortium for Political and Social Research [distributor], 1994. Shavit, Y., & Rattner, A. (1988). Age, crime, and the early life course. The American Journal of Sociology, 93 1457-1470. Shaw, C.D., & McKay, H.D. (1942). Juvenile delinquency and urban areas Chicago: University of Chicago Press. Short, J.F. (1990). Gangs, neighborhoods, and youth crime. Criminal Justice Research Bulletin, 5, 1-11. Silverthorn, P., & Frick, P.J. (1999). Developmen tal pathways to antisocial behavior: The delayed-onset pathway in girls. Development and Psychopathology, 11, 101-126.

PAGE 178

178 Silverthorn, P., Frick, P.J., & Reynolds, R. (2001) Timing of onset and correlates of severe conduct problems in adjudicated girls and boys. Journal of Psychopathology and Behavioral Assessment, 23, 171. Simcha-Fagan, O., & Schwartz, J.E. (1986). Ne ighborhood and delinquency: An assessment of contextual effects. Criminology, 24, 667-703. Singer, J.E., Westphal, M., & Niswander, K.R. (1968). Sex differences in the incidence of neonatal abnormalities and abnormal performance in early childhood. Child Development, 39, 103-112. Skogan, W.G. (1990). Disorder and decline New York: Free Press. Snyder, H. (1988). Court careers of juvenile offenders. Washington, DC: U.S. Department of Justice, Office of Juvenile Ju stice and Delinquency Prevention. Snyder, J., & Patterson, G. (1987). Family inter action and delinquent behavi or. In H. Quay (ed.), Handbook of juvenile delinquency. New York: Wiley. Steffensmeier, D.J., Allan, E.A., Harer, M.D., & Streifel, C. (1989). Age and the distribution of crime. The American Journal of Sociology, 94 803-831. Steffensmeier, D.J., Streifel, C., & Shihadeh, E.S. (1992). Cohort size and arrest rates over the life course: The Easterlin Hypothesis revisited. American Sociological Review, 57 306314. Sutherland, E.H. (1947). Principles of criminology. 4th ed. Philadelphia: Lippincott. Sutherland, E.H. (1939). Principles of criminology 3rd edition. Philade lphia: Lippincott. Taylor, R.B. (1995). The impact of crime on communities. Annals of the American Academy of Political and Social Science, 539, 28-45. Taylor, R.B., & Covington, J. (1988). Nei ghborhood changes in ecology and violence. Criminology, 26, 553-589. Thornberry, T.P., & Krohn, M.D. (2003). The self -report method for measuring delinquency and crime. In D. Duffee (ed.), Criminal justice 2000, volume 4, measurement and analysis of crime Washington, DC: National Institute of Justice. Tibbetts, S.G., & Piquero, A.R. (1999). The in fluence of gender, low birth weight and disadvantaged environment in predicting early-onset of offending: A test of Moffitt's interactional hypothesis. Criminology, 37, 843-878. Tittle, C.R. (1988). Two empirical regularities (ma ybe) in search of explanation: Commentary on the age-crime debate. Criminology, 26, 75-86. Tonry, M., Ohlin, L.E., & Farrington, D.P. (1991). Human development and criminal behavior New York: Springer-Verlag.

PAGE 179

179 Tracy, P.E., Wolfgang, M.E., & Figlio, R.M. (1990). Delinquency in two birth cohorts. New York: Plenum. Uggen, C. (2000). Work as a turning point in the life course of criminal s: A duration model of age, employment, and recidivism. American Sociological Review, 65 529-546. Visher, C.A., & Roth, J.A. (1986). Participation in criminal careers. In A. Blumstein, J. Cohen, J.A. Roth, & C.A. Visher (eds.), Criminal careers and "career criminals, Volume l. Washington, DC: National Academy Press. Warr, M. (1998). Life-course trans itions and desistance from crime. Criminology, 36, 183-216. Warr, M. (1993). Age, peers, and delinquency. Criminology, 31 17-40. Warr, M., & Stafford, M. (1991). The influence of delinquent peers: What they think or what they do? Criminology, 29 851-866. Weissberg, R.P., Gesten, E.L., Rapkin, B.D., Co wen, E.L., Davidson, E., de Apodaca, R.F., & McKim, B.J. (1981). The evaluation of a so cial problem-solving training program for suburban and inner-city third grade children. Journal of Clinical and Consulting Psychology, 49 251-261. Weisburd, D., Bushway, S., Lum, C., &Yang, S. (2004). Trajectories of crime at places: A longitudinal study of street segmen ts in the city of Seattle. Criminology, 42, 283-322. Wellman, B., & Leighton, B. (1979). Networks neighborhoods, and communities: Approaches to the study of the community question. Urban Affairs Quarterly, 14, 363-390. Welsh, W., Greene, J. & Jenkins P. (1999). School disorder: The influence of individual, institutional, and community factors. Criminology, 37 73. Werry, J.S. (1986). Biological factors. In H.C. Quay & J.S. Werry (eds.), Psychopathological disorders of childhood New York, NY: Wiley. West, D.J., & Farrington, D.P. (1973). Who becomes delinquent? London: Heinemann. West, D.J., & Farrington, D.P. (1977). The delinquent way of life. London: Heinemann. White, N., & Piquero, A.R. (2004). A preliminary empirical test of S ilverthorn and Frick's delayed-onset pathway in girls using an urban, African-American, US-based sample. Criminal Behaviour and Mental Health, 14, 291-309. Wikstrom, P.H. (1991a). Urban crime, criminals, and victims New York: Springer-Verlag. Wikstrom. P.H. (1991b). Housing tenure, social class and offending. Criminal Behaviour, 1, 6989.

PAGE 180

180 Wikstrom, P.H., & Loeber, R. (2000). Do di sadvantaged neighborhoods cause well-adjusted children to become adolescent delinquents? A study of male juvenile serious offending, individual risk and protective fa ctors, and neighborhood context. Criminology, 38, 11091142. Wilson, J.Q., & Kelling, G.L. (March, 1982). Broken windows: The police and neighborhood safety. Atlantic Monthly, 29-38. Wilson, J.Q., & Hernstein, R. (1985). Crime and human nature New York: Simon & Schuster. Wilson, W.J. (1987). The truly disadvantaged: The inner city, the underclass, and public policy Chicago: University of Chicago Press. Wirth, L. (1938). Urbanism as a way of life. American Journal of Sociology, 44 1-24. Wolfgang, M.E., Figlio, R.M., & Sellin, T. (1972). Delinquency in a birth cohort Chicago: University of Chicago Press. Wolfgang, M., Thornberry, T.P., & Figlio, R.M. (1987). From boy to man, from delinquency to crime. Chicago: University of Chicago Press. Zorn, C.J. (1998). An analytic and empirical assessment of zero-inflated and hurdle Poisson specifications. Sociological Methods and Research 26, 368-400.

PAGE 181

181 BIOGRAPHICAL SKETCH Wesley Glenn Jennings received his bachelors degree in ps ychology from the University of South Carolina in 2002. After completing hi s undergraduate education he continued on to receive his masters degree in criminal jus tice in 2004 also from the University of South Carolina. Upon finishing his masters degree he moved down to the University of Florida to pursue his doctorate in criminology. His rese arch interests specifically include offense trajectories, sex offenders, and multi-level modeling.