Life Course Criminology: Understanding Criminal Development and Transitions, Study Guides, Projects, Research of Criminology

Download Life Course Criminology: Understanding Criminal Development and Transitions and more Study Guides, Projects, Research Criminology in PDF only on Docsity!

See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/

Life Course Criminology

Article · February 2010 DOI: 10.1201/9781420085525-c3 · Source: OAI

CITATIONS 31

READS 3,

2 authors:

Some of the authors of this publication are also working on these related projects:

Academische Werkplaats bij De Nieuwe Kans - T.H.E. study View project

Rule violation in the chemical industry View project

Arjan Blokland Netherlands Institute for the Study of Crime and Law Enforcement (NSCR) 179 PUBLICATIONS 2,841 CITATIONS SEE PROFILE

Paul Nieuwbeerta Leiden University 315 PUBLICATIONS 6,096 CITATIONS SEE PROFILE

All content following this page was uploaded by Arjan Blokland on 29 December 2013. The user has requested enhancement of the downloaded file.

Life Course

Criminology

ARJAN A. J. BLOKLAND AND

PAUL NIEUWBEERTA

Contents

3.1 Introduction 52 3.2 The Roots of Life Course Criminology 54 3.3 Pillars of Life Course Criminology 55 3.4 Life Course Criminology Themes 57 3.5 Life Course Criminology Theories 60 3.5.1 Sampson and Laub's Age-graded Social Control Theory 60 3.5.2 Moffit's Dual Taxonomy 61 3.5.3 Gottfredson and Hirshi's Self-Control Theory 62 3.5.4 Summary of Theoretical Discussion 62 3.6 Life Course Criminological Data 63 3.7 Longitudinal Mefhods: Taking Observed Heterogeneity into Account 65 3.7.1 Describing Developments over fhe Course of a Person's Life 65 3.7.2 Explaining Developments over Life: Estimating Effects of Life Course Events and Interventions 67 3.8 Empirical Research in Life Course Criminology 69 3.8.1 Describing Developments over fhe Life Course 69 3.8.2 Life Course Developments-at fhe Micro Level 72 3.8.2.1 Military Service 73 3.8.2.2 Work 74 3.8.2.3 Marriage 75 3.8.2.4 Linked Lives 77 3.9 Redirecting Criminal Development 79 3.10 Conclusion 82 3.11 Future Goals 84 References 85

T AF-85514SHOHAM2-09-oB02_CQ03.indd (^51) 8/27/09 6:29:59 PM I

Life Course Criminology 53

of performing criminological research. Traditionally, criminologists have mostly been concerned with studying between-group differences in crime and had made comparisons between groups of individuals at one point in time [LeBlanc and Loeber, 1998]. Life course criminology, on the other hand, concentrates on within-individual developments in crime over time and rec- ognizes that the causal factors influencing development may shift as the individual progresses along his or her behavioral pathway. This, in turn, asks for the temporal ordering of events and of the causal processes leading up to these events to be made explicit. Theories that attempt to more fully address these developmental and life course aspects of criminal behavior have been advanced only in the last 20 years [Thornberry, 1997; Farrington, 2005]: To be able to study individual behavioral development across the entire life span, life course criminology also needed to exceed the limits of the available criminal data and find innovative ways to analyze this new information. The 1980s and 1990s witnessed a surge of new longitudinal studies, while previously initiated studies increasingly yielded data covering ever-larger parts of the life span [Thornberry and Krohn, 2003]. Together, fresh theories and newly tapped data sources acted as catalysts for new and increasingly sophisticated statistical analyses that could address the novel questions prompted. Given its broad scope and its capacity to enrapture theorists and empiricists alike, the life course approach has the potential to provide a suitable metatheoretical framework that can be used to integrate different theories explaining different aspects of crime and join existing empirical literatures on different topics that until now have been building up unconnectedly. Th.e remainder of this chapter is organized as follows. First, we consider the roots from which life course criminology has blossomed over the past two decades. Next, we explicate important elements of the life course approach and assess their repercussions in recent criminological theories. We then briefly review the available data and methods of analysis tbat have been used to study crime over the life course. In the second part of the chapter we review the empirical evidence describing the course of criminal trajectories as well as factors that have been proffered to help explain them. We also address the effects and problems associated with efforts to redirect criminal pathways, especially imprisonment. We conclude with a critical reflection of the cur- rent state of affairs in life course criminology and point to possible directions to allocate further research effort.

. Still, adopting a life course stance does not mean that one has to discard of all theoretical ideas that "classical" or~perhaps more aptly phrased-non-life course criminological thinking has yielded. As Farrington (2005) notes, once we start to realize that many of the circumstances traditionally studied to explain group differences in crime not only differ between individuals but also vary within individual lives, these ideas can be, and have already been, successfully adapted to fit the life course perspective.

TAF-85514SHOHAM2-09-0802_C003.indd 53 8127/09 6:30:00 PM I

54 International Handbook of Criminology

3.2 The Roots of Life Course Criminology

As with many ideas in science, tracing the exact origins of the life course approach to crime proves difficult. Many texts mention the first of Wolfgang's Philadelphia birth cohort studies, published in 1972 [Wolfgang, Figlio, and Sellin, 1972]. This study analyzed delinquent development up to age 17 in a sample of nearly 10,000 boys born in Philadelphia in 1945. Their finding, that more than half of the cohort's delinquency could be attributed to only a small proportion of the boys-6 % of the cohort, 18% of the delinquents in the cohort-sparked the interest of both researchers and politicians in efforts to try to prospectively identify this highly active delinquent subgroup and subject its members to policies aimed at redirecting their delinquent develop- ment by either rehabilitation or incarceration. Others take as the point of departure the 1986 report of the Panel of Research on Criminal CareerS. organized by the National Academy of Sciences [Blumstein et aL, 1986]. One of the major tasks of the panel was to assess the extent to which future criminal development could be predicted and to speak about the possible contribution of such predictions to policies of general and selective incapacitation. The panel's two-volume report pro- vided many of the basic tools and much of the vocabulary needed to analyze longitudinal data on criminal behavior. Yet while the Blumstein report and even the Wolfgang study are important hallmarks in the life history of criminology [Laub, 2004], a number of longitudinal studies predate them, most notably the work of Sheldon Glueck and Eleanor Glueck [1950J. The Gluecks studied the delinquent development up to age 32 of 500 boys remanded to Massachusetts reform schools during adolescence: Theirworkhas recentlybeenbrought backinto the limelight, when John Laub and Robert Sampson unearthed the raw data of the Gluecks' study, digitized all information, and reanalyzed it using modern statistical methods [Laub and Sampson, 1988]. Interest in the lives of offenders goes even further back. Others may, for example, want to point out the life history approach taken by the Chicago SchooL There, Robert Park, a former newspaper reporter, sent his students out of the lecture rooms into the streets of a city in turmoil to document people's own stories. Many of the qualitative studies that were conducted by students of the University of Chicago's sociology department, like C. Shaw's [1930] The Jack-Roller and E. Sutherland's [1937] The Professiol1al Thief, have now reached the status of criminological classics. Despite differences in methodology, what these studies share is an interest in the processes and dynamics of within-individual change in criminal behavior.'

. The Gluecks performed multiple longitudinal studies, including a 15-year followup of 1,000 juvenile delinquents referred to Judge Baker Clinic and a IS-year followup of 510 reformatory inmates. t Adolphe Quetelet and other scholars had already studied the age-crime curve in the nineteenth century, but on aggregated data.

TAF-85514SHOHAM2-09-0802_CO03.lndd (^54) 8127109 6:30:00 PM I

56 International^ Handbook^ of Criminology

risk factor paradigm also focuses on protective factors, tbat is, on factors that decrease the likelihood of offending (and which may not always be merely the opposite end of a risk factor scale) [Loeber and Farrington, 1998; Rutter, 1985]. While many of the variables used to try to predict offending were taken from classical criminological theories, like the criminal career paradigm the risk factor approach does not entail efforts to systematically combine the various risk factors found in empirical studies in a comprising theory of delinquent development. Consequentially, risk factor research's primary focus is not on establishing causality and explanations (e.g., parental disharmony may be a risk factor because disruptive interfamilial processes cause delinquency in children, as well as because delinquency-prone children themselves are a major source of dissonance within the family) but predominantly on merely finding correlations [Farrington, 2000].

Developmental criminology, the third pillar of DLC, entails the study

of the development and dynamics of antisocial behavior and delinquency as correlated with age and tries to identify the causal factors that influence the course that development [Loeher and LeBlanc, 1990]. Elaborating on the criminal career paradigm, developmental criminology quantifies a number of dynamic concepts capturing change and stability over time, distinguishing activation (the initiation and stimulation of offending), aggravation (the increasing seriousness of offending), and desistance (the winding down of offending) as distinct processes in the course of offending [LeBlanc and Loeber, 1998]. Compared to both the criminal career paradigm and the risk factor approach, developmental criminology is more theoretically driven. Influenced by developmental psychology, developmental criminologists have proposed stage models, wherein different types of offending follow a relative orderly sequence [e.g., Loeber and Hay, 1994]. Also, as in developmental psychology, the age period under scrutiny in developmental criminology mostly spans from birth to young adulthood. Given that from a developmental point of view the minimum age oflegal responsibility is an arbitrary marker, developmental criminologists mostlyrely on self- or third-party reports on the level and type of delinquent involvement. Furthermore, its focus on children and adolescents has lead developmental criminologyto also include among its dependent variables nondelinquent problem behaviors like bullying, truancy, or lying. Taken as a whole, developmental criminology aims to provide new insights about the etiology and precursors of offending that are of relevance to prevention and intervention programs aimed at terminating or diverting an individual's future criminal development [Farrington, 2002].

The fourth DLC pillar is life course criminology. At first glance, life course

criminology differs most conspicuously from developmental criminology in that life course criminology focuses mainly on criminal development in adult offenders. Instead offocusing on parenting styles or delinquent peers, life course criminological research deals with the effects of such life events as

1 TAF-85514SHOHAM2-09-oa~COO3.indd 56 8/'ZlI09 6:30:00 PM I

Life Course Criminology 57

work and marriage. However, behind these superficial differences are more fundamental theoretical discrepancies. First, the life course view holds that change is not limited to some "sensitive period" during early life bnt occurs throughout the entire life span. As a result, life course criminologists go through great efforts to follow their subjects far into adulthood. Laub and Sarnpson [2003], for instance"traced the original boys from the 1930 Glueck study, who by then had reached the age of 70, and ended up interviewing 52 of them. Second, life course criminology opposes to the lawlike notion of individual development-like the idea of developmental stages-that is present in much of developmental criminology [e.g., Loeber, Keenan, and Zhang, 1997]. Life course criminology, in turn, emphasizes variability and change in individual development that cannot be explained by preexisting differences between individuals. Within the life course paradigm one's criminal career is just one of many pathways one can embark on, be it one that has the poten- tial to strongly affect development in many others. Recognizing the interde- pendence of developmental pathways in different life domains, researchers understand that hitches in other pathways may bring about radical changes in the individual's criminal career. To these four pillars may be added a fifth, namely, that of life history narratives. Life history narratives reveal criminal development from "the inside out," because they offer personal accounts of the many processes that developmental and life course theories presume underHe patterns of criminal behavior over time. Following in the footsteps of the Chicago School pioneers, a number of researchers have opted for a qualitative take on development and have documented detailed life histories from a wide range of different offenders [e.g., Klockars, 1978; Shover, 1996; Steffensmeier and Ulmer, 2005]. While some of these studies only cover a demarcated period, others have had the opportunity to let offenders reflect on practically their entire life span. Life histories let offenders reflect on their lives in their own words. This approach uot only encompasses the many different pathways a person embarks on but also lets offenders themselves speak on the interdependency of these pathways both within and between individuals. Furthermore, quali- tative research, particularly among adolescents, may reveal aspects of social timing that dictate important transitions in their worlds and may even detect developmental stages that go unnoticed by merely using quantitative data. Finally, life history narratives may offer more detailed accounts of key criminal career characteristics like escalation or desistance.

3.4 Life Course Criminology Themes

Within life course criminology, the term life course reters to the wide array of age-graded trajectories people engage in during their lives, like work careers

TAF-85514SHOHAM2-09-0802_C003.indd 57 8/27/09 6:30:00 PM I

Life Course Criminology (^59)

have been claimed to go through a new developmental phase dubbed "emergiug adulthood," which neither simply extends adolescence because it is freer of parental control, nor can br viewed as early adulthood, since most young people have not begun to make the transitions traditionally associated with adulthood [Arnett, 2004]. In turn, this prolonged period of exploration may AQ partly explain why even individuals with nonproblematic childhoods are found to postpone desistance and remain criminally active until later ages [Moffitt et aI., 2002]. The third theme, of linked lives, draws attention to people's social embeddedness. Development does not take place in social isolation: individual action has consequences for others, and the actions of others relate back to the individual. The family is an obvious example of a social institution that connects the lives of many people over many generations, linking partners to each other, partners to in-laws, parents to children, children to siblings, and grandparents to grandchildren. The strength of these ties is again dependent on age as well as developments in other life domains. Research by Meeuws, Branje, and Overbeek [2004], for example, found that AQ the importance of parental support in curbing delinquency decreased as adolescents acqUired a stable romantic partner. At the same time, however, with age, parental support became increasingly important for those ado- lescents and young adults who never had a romantic partner, suggesting a reaffirmation of the parent-child relationship as prospects of other rela- tionships diminished. Fourth and finally, human agency recognizes that individuals are capable of continuously making choices that can influence the course of their development. People do not float n,dderless across life's sea, victim to the macrosociological currents. Instead, the life course perspective sees people as actively constructing their own development, given the historical constraints and opportunities available to them. The idea of human agency goes beyond mere selection-the notion that people actively seek out environments that fit their qualities-in that agency itself is variable and is influenced by environmental factors. The importance of agency is illustrated by the narratives of offenders in the Liverpool Desistance Study. Maruna [2001] compared the self·narratives of30 desisting and 20 persisting offenders matched as closely as possible on age, gender, family background, and criminal histories. He found that persistent offenders were more likely to subscribe to a "pawn" story of self; feeling that life outcomes were largely dependent on circumstance and change events and lacking a language of agency or self·initiative. Desisters, on the other hand, actively made efforts to regain what they saw as their true self and often referred to the desisting

process as will fully "going against the grain" [see also Laub and Sampson,

2003:55].

I TAF-85514SHOHAM2-09..Q802.-COO3.indd 59 8/27/09 6:30:00 PM I

60 International Handbook of Criminology

3.5 Life Course Criminology Theories

Over the years various explanations-typkally based on classic criminological theories-have been offered to account for important aspects of individual criminal development. Many of these explanations, however, had somewhat of a post hoc character, and often uncomplicatedly generalized results from aggregated, cross-sectional data to individual, longitudinal behavioral pat- terns. Over the past decades, though, there has been substantial progress in life course criminological theory. Scholars have formulated their theories more explicitly and integrated several developmental issues into a common theoretical framework. For an overview of some of the influential life course criminologkal theories, we refer to recent systematic reviews by Farrington [2005, 2006]. For the sake of brevity here we suffice with discussing what are without doubt the three most influential and most often cited reviews: Sampson and Laub's age-graded social control theory, Moffit's dual taxon- omy, and Gottfredson and Hirshi's self-control theory. Besides being influen- tial, these theories also conveniently illustrate the different ways the various life course themes have found their way in to criminological theory.

3.5.1 Sampson and Laub's Age-graded Social Control Theory

Main theorists in life course criminology have been Robert Sampson and John Laub. In numerous articles and two influential books, Sampson and Laub report the empirical results of their 60-year followup of 500 ex-reform school pupils first studied by the Gluecks and disclose their age-graded theory of informal social control. The basic argument of their theory is that criminal involvement results from a lack of informal social controls and that what constitute appropriate sources of social control varies throughout life [Sampson and Laub, 1993; Laub and Sampson, 2003]. Important transitions in other life course domaius yield changes in the level of informal control and

can therefore act as turning points for crime [Sampson and Laub, 1993; Laub

and Sampson, 2001]. Although Sampson and Laub do recognize that indi- viduals may select themselves in certain transitions, the researchers claim that many of these transitions are to a large extent "chance" events, occur- ring or not occurring independently of the individual's characteristics [Laub and Sampson, 2003:45]. The theory views both desistance and persistence in crime as outcomes of the same general causal process. While increasing ties to conventional society aid desistance from crime, criminal behavior itself may deteriorate already weak ties or cut off opportunities for conven- tional development, resulting in even lower levels of social control and mak- ing future crimes even more likely. Sampson and Laub [1995] refer to this

downward spiral-offending, weakened ties, more offending-as cumulative

continuity of disadvantage.

TAF-85514$HOHAM2-09-0802_COO3.indd (^60) 8/27109 6:30:00 PM I

62 International^ Handbook of Criminology

trajectories in constituting potential turning points for behavior. Especially Moffitt's adolescence limited theory is also closely linked to a particular location in time and place. In her view, the mismatch between biological and social maturation and the delinquent patterns that result from it, are by-products of modernization and therefore ouly characteristic for the twentieth-century Western world. The theory of adolescence limited offending also incorpo- rates the notion of timing oflives. Part of why delinquent behavior is increasingly costly is because with age it is increasingly perceived as "off-timed," that is, as part of the wild oats that already had to have been sowed.

3.5.3 Gottfredson and Hirshi's Self-Control Theory

The (re)introduction of the life course approach in criminology in general that has been taking place over the last 20 years and the work of Sampson and Laub and Moffitt, in particular, has not gone uncontested. In fact, it has evolved against the backdrop of theorists arguing that personal characteristics draw so heavily on individuals' choice-making abilities, that they drive them to choose social contexts to match these characteristics. These theories argue that if personal characteristics underlie behavioral outcomes as well as choice of social context, attributing causal power to contextual factors in shaping behavioral development would be fallacious. Representatives of this ontoge- AQ5 netie view are Gottfredson and Hirschi [1993J, who not only argue that this line of reasoning is valid for a small minority of adolescent offenders but also maintain that it is applicable to the entire population. In Gottfredson and Hirschi's view, criminal development is inversely linked to a person's level of self-control. Self-control also shapes individuals' agency in such a way that individuals tend to select themselves into environ- ments that match their self-control levels, rendering the causal relationship between certain social contexts and crime spurious [Hirschi and Gottfredson, AQ6 1995]. Gottfredson and Hirschi [1983, 1993J go on to argue that individual development mirrors the shape of the aggregate age-crime relationship, which they claim to be similar across individuals, places, and historical times. They further state that no variable other than age can offer an explanation for the way criminal involvement waxes and wanes. In the eyes of these scholars, the invariability of the age-crime curve in turn releases criminologists from the need to do longitudinal research, rendering Gottfrerlson and Hirschi important antagonists of the life course perspective on crime both theoretically AQ7 and methodologically [Gottfredson and Hirschi, 1986, 1987].

3.5.4 Summary of Theoretical Discussiou

Although differences in opinion continue to exist on what, for example, counts as persisting in crime and whether or not persistent offenders constitute a

TAF-85514SHOHAM2-09-Q802_C003.indd 62 8127/09 6:30:00 PM I

Life Course Criminology 63

separate group that^ can be meaningfully distinguished at an early stage,^ in hindsight much of the dust raised around the life course perspective seems to have whirled around whether the proverbial glass was either half full or half empty. Most life course criminologists acknowledge the intluence of stable individual characteristics on development and have sought ways to incorpo- rate such differences in their theoretical and empirical work. Similarly, many adversaries still consent to the notion that some exogenous influences are involved in development. If anything, the counterpressure of Gottfredson and Hirschi and others has inspired life course criminologists to refine their research questions and improve their methods of investigation in their continuous effort to make even better sense ofllves. Aiming to adequately describe and explain devel- opment in criminal behavior over time, lougitudinal data have now become the hallmark oflife course criminology.

3.6 Life Course Criminological Data

Much of the empirical foundation underlying "classical" criminological theory comes from cross-sectional studies, which by definition examine many different individuals at one point in time. Although these studies have yielded important insights on the correlates of delinquency and crime, cross-sectional studies are less suited to speak on devel- opmental issues. For instance, many cross-sectional studies are based on offender samples drawn from such databases as police records or prison populations. Reports on why and how these people started their criminal careers are therefore exclusively retrospective and possibly biased because the offenders arepresently involved in crime. In sampling only active offenders, these cross-sectional studies are also unable to reveal why people may desist from crime. Finally, since all information is gathered at one point in time, the causal nature of the associations revealed remains opaque [Menard, 2002J. AQ Liberman [2008J readily highlights the importance of a longitudinal approach by stating that when planning interventions, what we really want to know is whether we are capable of making changes to the constellation of risk factors an individual is exposed to. Will this same individual behave any differently than before? This is a question that can best be answered by following the same individual over time. More generally put, cross-sectional studies have a hard time establishing precise temporal order; consequently, causal inferences from such studies need to be interpreted with due caution. While different types of longitudinal studies continue to be initiated, the pith of empirical life course criminology seems to move from cohort studies, through panel studies, to multiactor studies.

T AF-85514SHOHAM2-09-0802_COQ3.lndd 63 8/27/09 6:30:00 PM I

Life Course Criminology 65

While panel studies gather rich data, sometimes using multiple informants, they tend to view individual development in relative isolation. That is, while panel studies involve information on the individual's social context, includ· ing family and friends, they typically do not link developmental pathways of different individuals. Multiactor studies do just that, thereby paying homage to the life course idea of linked lives. The Cambridge Study in Delinquent Developmeut, for example, was augmented with data on the criminal histo· ries of both the parents and children of the 411 boys originally sampled, thus linking criminal development of family members across three generations [Farrington, Coid, and Murray, 2008]. Similarly, starting out as a cohort AQ study, the Criminal Career and Life Course Study now includes information on the criminal careers of those convicted in 1977 and also on the criminai careers of all their lifetime marriage partners and their children [Van der Rakt, Nieuwbeerta, and De Graaf, 2008]. Gathering longitudinal data on individuals tied by social relationships helps to more fully appreciate the way individual development is embedded in social relationships that form and dissolve across the life span.

3.7 Longitudinal Methods: Taking Observed

Heterogeneity into Account

With more and more longitudinal data becoming available, life course-oriented criminologists began to look for adequate ways of analyzing this information. Within the context oflife conrse criminology's aims, stated in the introduction of this chapter-describing, explaining, and redirecting developments in criminal behavior over the life course-we discuss some of these method· ological developments.

3.7,1 Describing Developments over the Course of a Person's Life

One of the major aims of life course criminology is to adequately describe the development of criminal behavior over the life span. The simplest way of visualizing such development is to plot the behavior ofinterest-for example, the frequency or severity of delinquency and crime-against age. This can be done for the entire sample under scrutiny or for subgroups defined by such personal characteristics as sex and ethnicity. For growth processes concerning which it seems reasonable to assume that individuals develop in a more or less-typical fashion, plotting the average developmental pathway may be well suited. However, when individual variation in development is ",,,pected, as is the case with criminal development, more complicated methods are needed that do justice to the complexities of the data gathered. That is why criminologists have turned to multilevel growth curve models or simply growth models [Bryk and Raudenbush, 1992].

TAF-85514SHOHAM2-09-0802_C003.indd (^65) 8/27/09 6:30:00 PM I

66 International Handbook of Criminology

These models allow for individual variation in the rate and shape of the developmental pathway and are able link variation in these growth parameters to individual characteristics. Still, these models assume average trends and express developmental variation in terms of variance around the average growth within the population. Theorist like Moffitt and Patterson, for example, state specifically that there are different types of offenders showing qualitatively distinct developmental pathways and that the average age-crime curve is an amalgam of very differently shaped developmental pathways. When it comes to criminal development, growth curve models therefore seem to make little intuitive sense. More recently, scholars have developed a group-based, latent class approach for modeling criminal trajectories [Nagin and Land 1993; Nagin, 2005; Muthen 2001]. These models are more in linewith theories assuming that there is no general trend in criminal behavior, but different groups of people have different trajectories. Group-based models approximate individual developmental variation by a number of discreet groups. As in growth curve models, the outcome variable is linked to age by some polynomial function; but unlike growth curve models, the group-based model allows the parameters of this function to vary freely across groups. As a result, the estimated course of development for each group can be very different both in level and time path. In addition to the trajectories themselves, the group- based model produces an a posteriori probability of membership for each individual in the sample and for each of the distinguished trajectory groups. These probabilities indicate the likelihood that an individual showing a specific behavioral pattern belongs to one of the specified groups. These probabilities can be used for creating descriptive profiles of each group by assigning each individual to the trajectory group for which his posterior probability of group membership was highest [e.g., Vau der Geest, Blokland, and Bijleveld, 2008]. Group-based models have become a popular way of analyzing crimi- nal development [for reviews, see Piquero, 2008; Van Dulmen et al., 2009]. Yet, some researchers-and the models' architect in in the lead [see Nagin, 2005]-have warned against the unintelligible use of the method and pointed at common misconceptions in interpreting its outcomes [e.g., Eggleston, Laub, and Sampson, 2004; Nagin, 2004, 2005; Nagin and Tremblay, 2005]. In the context of this chapter, we highlight three such misconceptions. First, group-based methods essentially provide a statistical tool to reduce complexity in longitudinal data sets following many individuals over time. Any trajec- tories distinguished describe long-term behavioral patterns but leave room for short-term indIvidual variability. Individuals, thus, do not necessarily develop exactly according to the trajectory they are allocated to [Nagin and Tremblay, 2005]. Second, while partly inspired by typological theories, the mere finding of developmental trajectories in a particular data set does not

I TAF-85514SHOHAM2-09-oS02.....CO03.indd 66 8fZ1!09 6:30:00 PM I

68 International Handbook of Criminology

intervention net of the influence of the confounders added. The most important drawback to using regression models is that they only control for variation in observed characteristics, or observed heterogeneity. As a result, variation in unobserved characteristics, or unobserved heterogeneity, may still unsuit- ably influence any association established. In an effort to overcome problems caused by unobserved heterogene- ity, criminologists have turned to models that have allowed them to analyze within-individual change. These kinds of models make optimal use of the longitudinal data becoming available, because they compare the behavior of the individual after undergoing the transition or intervention to that of the AQlO same iudividual before undergoing the transition or intervention [Osgood, 2005]. In within-individual change analyses, the individual thus acts as his or her owu control, thereby nullifying the confounding effects of time-stable individual characteristics, both observed and unobserved. Such models have been applied in criminology to estimate the effects of such important life course transitions as finding employment, getting married, or becoming a parent [Blokland and Nieuwbeerta, 2005; Homey, Osgood, and Marshall, AQll 1995; Laub and Sampson, 2003; Layton MacKenzie and De Li, 2002]. Although within-individual change models suffer far less from omitted variable bias than do regression models, they only control for confounders that themselves do not change over time. Another way of modeling selection resulting from confounding variables that has recently found its way into life course criminology is propensity score matching. A propensity score is the conditional probability of experiencing a transition rather than the control given the observed covariates [Rosenbaum and Rubin, 1983]. In the life course context, the propensity score could, for instance, represent the conditional probability of being married at age 25, given observed covariates up to that age. If two individuals have the same propenSity score observed covariates- both time stable as well as time varying-these observed covariates will be of no further use in predicting which of these two individuals will be married at age 25. The next step in propensity score matching involves matching the individual experiencing the transition with an individual-or several individuals-with a comparable chance of experiencing the transition but who in actuality does not experience the transition. Comparing posttransition outcomes for matched individuals will yield estimates of the effect of the transition net of time- stable and time-varying observables. In the marriage example, propensity score matching would entail comparing criminal behavior in the 26-t030 age period for those who were married at age 25 with that of those who were, in important respects, similar to those married but who themselves were not married at age 25. Nevertheless, if as a result of unobserved cofounders the applied propensity model is not entirely successful in capturing selection into marriage, the resulting effect of this transition may still be compromised.

TAF-85514SHOHAM2-09-Q802_CO03.indd (^68) 8/27/09 6:30:00 PM I

Life Course Criminology 69

Sensitivity analyses can be conducted to test the robustness of the obtained effect against the presence and magnitude of unobserved confounding.

3.8 Empirical Research in Life Course Criminology

The growing amount of available longitudinal data, combined with increasingly sophisticated analytic methods, has yielded an enormous amount of empiri- cal findings. We now highlight some of these findings as they reflect the aims of life course criminology cited in the opening paragraph of this chapter: describing, explaining, and redirecting criminal behavior:

3.8.1 Describing Developments over the Life Course

The 1986 report of the Panel of Research on Criminal Careers distinguished three important dimensions that together demarcate the criminal career: participation, frequency, and duration. Participation asks how many people actually engage in criminal behavior over their life course. Using data from the first Philadelphia birth cohort study, Wolfgang, Figlio, and Sellin [1972] found that of those boys born in 1945, little over one-third had experienced a police contact by the time they reached 18. The second Philadelphia birth cohort study, following up a sample born in 1958, yielded a comparable par- ticipation rate [Tracy, Wolfgang, and Figlio, 1990]. Results from a Dutch cohort study tracking the entire Dutch population born in 1984 found that by the time cohort members had reached 22 years of age, 14% had at least one police contact [Grimbergen et al., 2008]. Age 25 participation in the AQ Swedish project Metropolitan reached 19% [Wikstriim, 1990]. Tracking the lives of 411 London working-class males, the Cambridge Study in Delinquent Development indicated that by 50 years of age, 41% of the men were convicted [Farrington et al., 2006]. Visher and Roth [1986] reviewed severallongitudi- nal studies and estimate a lifetime participation rate of 40-50%. In contrast, self-report studies in the United States and elsewhere have yielded participa- tion rates close to 100%, with the vast majority of people reporting behaviors that could have resulted in a criminal conviction. Comparing participation rates across different age periods tells us that for both official records and self-reports, the peak period of first participation or onset is before age 18. In sum, most people seem to engage in criminal acts sometime during the course of their lives; and for many, their criminal behavior results in some kind of criminal justice involvement.

. For extensive reviews: Farrington (2003) summarizes the empirical data by distinguish- ing established knowledge and contentious issuesj- Piquero, Farrington, and Blumstein (2003) also provide an elaborate overview or the empirical status of criminal career research.

TAF-85514SHOHAM2-09-0802_C003.indd 69 8/27109 6:30:00 PM I