Adulthood-limited offending: How much is there to explain?

Journal of Criminal Justice 55 (2018) 58–70

Contents lists available at ScienceDirect

Journal of Criminal Justice journal homepage: www.elsevier.com/locate/jcrimjus

Adulthood-limited offending: How much is there to explain?

T

Fredrik Sivertsson Department of Criminology, Stockholm University, Universitetsvägen 10A, SE-106 91 Stockholm, Sweden

A R T I C L E I N F O

A B S T R A C T

Keywords: Criminal career Developmental and life-course criminology Adult-onset offending Gender Population studies Growth curve

Purpose: The current study explores male and female adult-onset offending careers in a Swedish populationbased longitudinal dataset comprising five successive birth cohorts which are followed prospectively on the basis of detailed conviction data to age 50. Methods: Adult-onset offenders are compared to juvenile-onset offenders on a number of criminal career measures. Growth curve analysis is employed to visualize average trajectories for convictions during adulthood. Results: The study found that 22% of convicted males and 38% of convicted females were convicted for the first time for offenses committed between ages 25 and 50. The adult-onset males contributed 19% of all male adulthood convictions and 16% of male violent convictions in adulthood. The adult-onset females contributed 47% of all female adulthood convictions and 48% of female violent convictions in adulthood. While the adolescent-onset trajectories displayed generally decreasing trends for offending in adulthood, adult-onset females displayed increasing trends in relation to trajectories of violence and drug/alcohol-related offending as they approached middle adulthood. Conclusions: There is a need for developmental and life-course theories of crime to be explicit in explaining adult-onset offending, particularly in relation to gender disparities.

1. Introduction There has recently been an increasing interest, and growing controversy, regarding adult-onset offending within the field of developmental and life-course criminology. There remain a number of ambiguities with regard to the magnitude and seriousness of adult-onset offending and, connected to this, there is a theoretical controversy concerning how much attention should be devoted to explanations focused on this category of offenders. In essence, this controversy has concerned whether developmental theories, such as Moffitt's (1993) dual taxonomy, are sufficient to account for adult-onset offending, or whether there is a need for theories whose focus is instead directed at more proximate explanations during adulthood (Sohoni, Paternoster, McGloin, & Bachman, 2014). A recent review has noted that “(a) lthough there is a clear relationship between offending extremity and precocious onset, there is nevertheless compelling evidence of serious, and at times severe offenders who did not begin their criminal careers until well into adulthood” (DeLisi & Piquero, 2011, p. 294). Some studies suggest that adult-onset criminal careers tend to be brief and non-serious (e.g. Moffitt, 2006), while others have found that the majority of those with the most severe criminal careers were first arrested in adulthood (e.g. Delisi, 2006). As Beckley et al. (2016) have observed the ambiguities surrounding the magnitude of adult-onset offending are largely due to

methodological heterogeneity. In particular, the age cut-off used to mark the beginning of adulthood has recently been a matter of some debate. The lion's share of the onset literature has defined adult-onset as being first-time arrested or convicted at age 18 or later (Eggleston & Laub, 2002; Delisi, 2006; Delisi et al., 2018; Gomez-Smith & Piquero, 2005; Kratzer & Hodgins, 1999; Vere van Koppen, 2018). In a critique of the use of this cut-off, however, Moffitt (2006) has argued that “(a) lthough adult-onset crime begins at age eighteen in legal terms, in developmental terms for contemporary cohort samples, it begins sometime after age 25 […] In our view, the existence of individuals whose official crime record begins after age eighteen does not constitute a threat to the taxonomy” (p. 286; see also Sohoni et al., 2014). This argument was mainly based on Arnett's (2000) life-course typology, which posits that age 18 through age 24 captures a prolonged period of adolescence, or “emerging adulthood”, during which individuals are still loosely attached to adulthood markers such as stable employment and relationships. In addition, official onset may lag behind self-reported onset by a few years, which increases the risk for confusing adolescence-onset offending with adult-onset offending when using a legal definition of adulthood (Moffitt, 2006). On the basis of a “social” cut-off age for measuring adulthood, it may be argued that the bulk of previous criminal career research has overestimated adult-onset offending. At the same time, very few studies have been able to examine criminal careers in a population-

E-mail address: [email protected]. https://doi.org/10.1016/j.jcrimjus.2018.02.002 Received 20 December 2017; Received in revised form 9 February 2018; Accepted 9 February 2018 0047-2352/ © 2018 Elsevier Ltd. All rights reserved.

Journal of Criminal Justice 55 (2018) 58–70

F. Sivertsson

(Eggleston & Laub, 2002; Beckley et al., 2016). The theoretical controversy relates to what this observation means. Sohoni et al. (2014) have suggested that developmental and life-course theories may be divided into symmetrical and asymmetrical theories. Broadly speaking, asymmetrical theories posit that adulthood offending captures continuity in antisocial behavior and that the fundamental causes of adultonset offending may therefore be sought in early life. In contrast, symmetrical theories direct their attention at proximate factors preceding the criminal event relatively recently. The two most influential theories within developmental and life-course criminology are probably: Moffitt's (1993) dual taxonomy and Laub and Sampson's (2003) general age-graded theory of informal social control. According to Moffitt, persistence in crime is accounted for by the theory of life-course persistent offending. The theory posits that persistence in crime is a feature that characterizes a small and distinct group of maladaptive individuals who begin their antisocial path early in life and continue well into adulthood. The main explanation for this stability in antisocial behavior is found in traits whose roots are found in early infancy or even prenatally, and which in interaction with an often criminogenic environment cause offending across the entire life span. An early criminal record captures continuity in an antisocial lifestyle that started long before the age of criminal responsibility. While Moffitt's theory of life-course persistence has received a great deal of attention, much less attention has been given to the other part of the taxonomy, the theory of adolescence-limited offending (Moffitt, 2006). This theory sets out to explain why so many teenagers engage in normative delinquency during youth but then desist from crime in connection with the transition to adulthood. As adolescents age, the gap between biological and social maturity begins to close and adolescencelimited offenders gradually transition to adulthood status, which results in a decline both in the aggregate age-crime relationship and in the within-individual development of offending for this normative group of offenders during the juvenile phase of life. While the majority of serious and frequent adulthood offending should be accounted for by the theory of life-course persistent offending, Moffitt (2006) has argued that the main portion adult-onset offending could probably be accommodated by the adolescence-limited theory, because the criminal careers of adult-onset offenders are similar to those of adolescence-limited offenders in tending to be brief and non-serious (p. 287). In contrast to Moffitt, Laub and Sampson (2003) posit that the same mechanisms of informal social control are at work for everyone, and that they also account for both stability and change in criminal offending. The age-graded notion suggests that social institutions have different roles over the life course: the school and parents produce informal social control during the juvenile years, whereas employment and marriage are important social institutions during adulthood. Although Laub and Sampson acknowledge that childhood vulnerability is related to subsequent offending, they have strongly opposed the “kinds of people” notion that offending in adulthood is a matter of selection due to childhood risk factors. In contrast, they argue that life events in adulthood may to a substantial degree be described as a random process, and that they have an effect on subsequent offending net of childhood risk factors. In a more general sense, Laub and Sampson therefore argue that the causes of crime should primarily not be sought in the distant past but rather closer in time to the criminal event. To summarize, asymmetrical theories, here exemplified by Moffitt's dual taxonomy, do not account for the kind of “upward” change in offending behavior that is implied by the existence of adult-onset offenders. The explanations for adult-onset offending, as measured in criminal records, must therefore be sought in the ability of these individuals to persist in crime during adolescence and emerging adulthood while at the same time avoiding detection by the criminal justice system. In contrast, symmetrical theories suggest that positive change may occur for the most hardened offenders, and that vice versa, negative change may occur for individuals who lack a troubled past. The

representative cohort of individuals up to midlife, which may have led to an underestimation of adult-onset offending. In their review, Beckley et al. (2016) suggest that “studies beyond the CSDD and outside of the USA would help to address generalizability of descriptive data about adult-onset offending” (p. 67). They also note the lack of research on female adult-onset offending and suggest that “(l)arger populationbased samples and offender-based samples will be needed to study the prevalence and correlates of adult-onset crime among women” (p. 79). Building on the social age typology developed by Arnett (2000), the aim of the current study is to examine the adult criminal careers of males and females who were convicted for the first time for crimes committed at age 25 or later, and to compare these to the adult criminal careers of those who were convicted for the first time in adolescence (aged 15–17) and emerging adulthood (aged 18–24). I employ a Swedish dataset comprising five successive birth cohorts of males and females born between 1960 and 1964, which follows these cohorts prospectively through 2015, and which contains detailed information on convictions, such as the timing and type of offending. The data set also allows for controls for mortality and migration. The research questions examined are: 1) How large a proportion of convicted offenders were convicted for the first time for offenses they committed between age 25 and age 50? 2) How do criminal career measures such as the number of lifetime convictions and career length differ between adult-onset offenders and juvenile-onset1 offenders? 3) Is there an association between age of onset and recidivism among adult-onset offenders? 4) How large a proportion of adulthood convictions, for different offense types, are accounted for by adult-onset offenders? 5) How does the average adult-onset offending trajectory differ from the average adulthood offending trajectories of juvenile-onset offenders? 6) How do the average violent-, property-, and drug/alcohol-related offending trajectories in adulthood differ between adult-onset offenders and juvenile-onset offenders? These research questions are systematically examined by gender. The examination of adult-onset offending careers should be informative in relation to both crime prevention policies and developmental and life-course theories of offending. Although the discovery of early risk factors provides knowledge on the precursors of criminal behavior, the first conviction often presents the first opportunity for the criminal justice system to intervene (Svensson, 2002). Given that there is some relevant proportion of offenders who are convicted for the first time for offenses they committed in adulthood, it is therefore important to evaluate their risk for recidivism and how much they contribute to the total volume of adulthood crime. The importance for developmental and life-course theories of being able to account for officially recorded adult-onset offending should be related both to the magnitude of this phenomenon and to the seriousness of the remainder of their criminal careers (van Koppen, 2018). Particularly little is known about adultonset offending among females, although there is an increasing interest within developmental and life-course criminology in studying gendered patterns of continuity and change in offending (Macmillan & McCarthy, 2014). Because many developmental theories, such as Moffitt's taxonomy, are gender-neutral (see Moffitt, Caspi, Rutter, & Silva, 2001), the systematic examination of adult-onset offending careers within and across gender ought to bring an extra dimension to the controversy surrounding the adult-onset offender. 1.1. Adult-onset offending in theory Any longitudinal study that follows individuals from the age of criminal responsibility up through some part of their adulthood will find that some proportion of these individuals are registered for offenses for the first time in relation to an offense committed in adulthood 1 By juvenile-onset offenders are here meant those who were first-time convicted for crimes they committed prior to age 25 (see also Beckley et al., 2016).

59

Journal of Criminal Justice 55 (2018) 58–70

F. Sivertsson

born during the first half of the 1960s, the majority were in upper secondary school until earliest age 18. The Swedish level of living surveys show that in these cohorts by age 25 around 90% of males and females had experienced their first employment, around 65% of males and 75% of females had entered into their first cohabitation, and around a quarter of males and 35% of the females had entered into parenthood. These adulthood markers were substantially lower by age 18; around half of the males and females had had their first employment, below 10% of males and around 20% of females had entered into their first cohabitation, and only a small percentage had entered into parenthood (Hörnqvist, 1994). In relation to marriage, register-based records from Statistics Sweden show that for cohorts born in the early 1960s around 30% of females were ever-married by age 25 while only a very small percentage were married by age 18 (Ohlsson-Wijk, 2011). These figures are lower among males since they tend to marry somewhat later than females. Hence, for Swedish cohorts born during the first half of the 1960s, age 25 appears to be a more reasonable cut-off than age 18 in marking the transition to adulthood, at least in developmental terms. In addition, the official age of criminal onset may lag behind the true age of onset by a couple of years, which constitutes yet another reason for using a more restrictive age cut-off for measuring adulthood (Moffitt, 2006, p. 285). Very few studies have explicitly adopted a social cut-off age for adulthood in order to examine the magnitude and seriousness of adultonset offending. Sohoni et al. (2014) examined adult-onset offending among the South London boys in the CSDD, as measured by conviction data up to age 50, and among the members of the Rochester Youth Study (RYDS) on the basis of arrest records up to age 31 and 32. They found that among all officially recorded males, 17% in the CSDD and 13% in the RYDS were social adult-onset offenders. These figures became substantially higher when they included those “off-time” offenders whose first convictions were for offenses committed during emerging adulthood. In the Dunedin cohort, Beckley et al. (2016) found that only around 10% of convicted males were convicted for the first time for offenses committed at or after age 25 as compared to a third of the convicted cohort when age 20 was employed as the cut-off age. Thus these two studies both showed that the magnitude of adult-onset offending was highly dependent on the adoption of a legal or a social age cut-off. When it comes to the seriousness of adult-onset offending, previous research has generally found that adult-onset offenders commit fewer offenses and display shorter career durations than juvenile-onset offenders (Beckley et al., 2016; Farrington, Piquero, & Jennings, 2013; Jennings, Loeber, Pardini, Piquero, & Farrington, 2015; but see Delisi, 2006). At the same time, even the most high-risk juvenile-onset offenders appear to be most active during the period prior to the transition to adulthood, suggesting a convergence in offending trajectories as middle adulthood approaches (Laub & Sampson, 2003; Sivertsson & Carlsson, 2015). In one of the few studies that has compared the adulthood offending rates of juvenile-onset and adult-onset offenders, Beckley et al. (2016) found that when the period between age 20 and 40 is held constant, the difference in the crime rates between juvenile-onset and adult-onset offenders becomes much smaller than when the comparison relates to the full study period. In the CSDD, McGee and Farrington (2010) found that the adult-onset men were convicted for proportionally more sex offenses, frauds, thefts from work, and vandalism than juvenile-onset offenders, whereas Beckley et al. (2016) found that property and fraud crimes, driving under the influence and other criminal driving violations were significantly more common among adult-onset offenders.

main explanations for adult-onset offending can therefore be sought in adulthood. 1.2. Adult-onset offending in practice The ambiguities surrounding the phenomenon of adult onset have been summarized in the following points by Krohn, Gibson, and Thornberry (2013): the use of only official data, the question of how much and what types of crime should be counted, and the selection of a particular age at which to distinguish adult-onset offending from juvenile-onset offending (p. 187). The general conclusions drawn by studies that have utilized both self-reported and official data to examine adultonset offending are that some proportion of those who are arrested and convicted for the first time in adulthood had engaged in some level of undetected delinquency, as measured by self-report questionnaires, in their juvenile years (Beckley et al., 2016; McGee & Farrington, 2010; Sohoni et al., 2014; Wiecko, 2014). At the same time self-report studies indicate that some degree of delinquent behavior in adolescence is so common that it is to be considered normative. McGee and Farrington (2010) have noted that “virtually all of the boys in the CSDD reported some level of involvement in offending” (p. 540). Similarly, in the Dunedin cohort, a population-representative cohort of individuals, only 10% of the males had abstained completely from antisocial behavior during childhood and adolescence (Moffitt, Caspi, Dickson, Silva, & Stanton, 1996). More restrictive definitions of juvenile offending have therefore been used to measure true adult-onset. For example, McGee and Farrington (2010) have argued that two-thirds of the official adultonset offenders in the Cambridge study were true adult-onset offenders because one-third had self-reported offending that exceeded the average of the juvenile-onset offenders. In another study, Kratzer and Hodgins (1999) examined categories of offenders in a Swedish birth cohort and found that only 19% of the male adult-starters and 24% of female adultstarters had unofficial records of delinquency that had been recorded by the child welfare agency between age 13 and 18. Perhaps the strongest, and least recognized, argument for questioning the phenomenon of adult-onset offending has to do with the chronological age used to define adulthood (Sohoni et al., 2014). Most of the adult-onset literature has employed age 18 as the cut-off on the basis that this is the age that marks the legal transition from juvenile to adult court in the US, and because it occurs subsequent to the normative peak of offending according to the aggregate age-crime curve (Eggleston & Laub, 2002; Delisi, 2006; Delisi et al., 2018; Gomez-Smith & Piquero, 2005; Kratzer & Hodgins, 1999; Vere van Koppen, 2018). Some studies departed from this legal definition and used age 21 as the cut-off (McGee & Farrington, 2010; Zara & Farrington, 2009). However, a case was recently made that any study that defines adulthood prior to age 25 overstates the magnitude of adult-onset offending (Sohoni et al., 2014). The argument is that while age 18 marks the legal age in many jurisdictions, adult roles are not typically adopted until age 25 or later in contemporary Western societies (Moffitt, 2006; Sohoni et al., 2014). Age 18 through 24 may instead be defined as an unsettled period of “emerging adulthood” during which individuals are loosely attached to markers of adulthood such as stable relationships, jobs and living arrangements, and in which they feel like neither juveniles nor adults, but a little bit of both (Arnett, 2000). It is therefore reasonable to assume that problem behaviors that have typically been associated with the teenage years, such as crime, have over time also become more common within older age groups, reflecting the postponement of social adulthood (Hayford & Furstenberg, 2008). As Arnett (2000) has highlighted, emerging adulthood is not a universal concept but a highly contextual one, which may vary in length between and within countries over time.2 For Swedish cohorts

(footnote continued) & Farrington, 2010), and more relevant to choose a later cut-off, such as age 25, in a population representative cohort of individuals born in Sweden in the early 1960s or in Dunedin in the early 1970s (see Beckley et al., 2016).

2 In this sense, it may be perfectly justified to choose age 21 as the cut-off in a sample of working class boys born in the London suburbs at the beginning of the 1950s (see McGee

60

Journal of Criminal Justice 55 (2018) 58–70

F. Sivertsson

means that the study population has been followed prospectively from age 15 (i.e. the age of criminal responsibility in Sweden) to age 50 in the youngest cohort and age 55 in the oldest cohort. Data on mortality were available for 1960 to 2015 and migration dates were available for 1961 to 2012. Individuals who had died before age 15 were excluded from the data set, as were individuals who had emigrated and not returned before age 15 and also individuals who had migrated to Sweden after age 15. This selection was required in order to ensure that the adult-onset offenders had no previous criminal record in Sweden or abroad.

1.3. Gender and adult-onset offending There is good reason to believe that it is particularly relevant to study females with regard to adulthood offending in general and adultonset offending in particular. Although males are generally overrepresented in offending at the population level, criminal career research suggests that this overrepresentation mainly relates to the juvenile years and that differences in magnitude are much smaller during adulthood (Block, Blokland, van der Werff, van Os, & Nieuwbeerta, 2010; Wikström, 1990). This pattern appears to be partly driven by a postponed official onset among females in relation to males (Andersson, Levander, Svensson, & Levander, 2012; Bergman & Andershed, 2009; Block et al., 2010; D'Unger, Land & McCall, 2002; Estrada & Nilsson, 2012; Sivertsson, 2016; Vere van Koppen, 2018; but see Gomez-Smith & Piquero, 2005). Only one study has been found which has to some extent been able to systematically analyze female adult-onset offending up to midlife while adopting a social cut-off age for adulthood. In the Dunedin cohort, Beckley et al. (2016) were able to follow a populationrepresentative cohort of males and females up to age 40. In line with what might be expected from previous research showing a postponed age of onset among females, Beckley et al. (2016) found a gender difference in that social adult-onset offenders were more prevalent among the females. Around 18% of convicted females were first convicted of crimes at or after age 25, and this group accounted for around 10% of all female convictions. The corresponding prevalence of adult-onset males was 10%, and these individuals only accounted for 3% of the cohort males' total number of convictions (p. 72). Interestingly, the females who were convicted for the first time at age 25 or later had a higher mean lifetime conviction rate than those convicted for crimes at age 20 or later, and their conviction rate also exceeded that of the adultonset males (Beckley et al., 2016, p. 72). The authors noted that these initial findings showed that “women appear to be somewhat different from men with regard to official adult-onset offending”, but that a more thorough comparison was not possible due to lack of statistical power (p. 72). Clearly, the combination of small sample sizes and relatively short follow-up periods places a lot of restrictions on the possibilities for analyzing the phenomenon of adult-onset offending, and this is particularly so in relation to the female population. As Andersson (2013) has observed, even in datasets which included females, the analysis of criminal careers is usually carried out on the males because the restricted number of female offenders precludes the use of statistical methods (p. 27). Given this background, the availability of Swedish longitudinal population-based data for five successive birth cohorts followed prospectively to age 50 provides a unique opportunity to analyze the criminal careers of both males and females whose first convictions were for offenses they committed long after the normative peak of criminal offending.

2.1. Swedish population-based conviction data Generally, conviction records capture offenders who have committed several offenses, and/or serious offenses, and/or traditional person- and property-oriented offenses (Kyvsgaard, 2002, p. 21; see also Blumstein & Cohen, 1987). There are at least three factors that should be highlighted when using conviction data to examine adultonset offending. First, one factor that may lead to an overestimation of adult-onset offenders is that old criminal records are often removed from the registers after a time. One advantage with the use of Swedish conviction data for research purposes in this respect is that old criminal convictions are not removed from the database in this way, and the database contains every criminal conviction from 1973 up to present day. Second, the propensity for the police to arrest and prosecute juveniles in connection with the detection of criminal offenses may generally be lower than it is in relation to adults (see MacLeod, Grove, & Farrington, 2012). Although it does not fully resolve this problem, a further advantage associated with the use of Swedish conviction data is that Swedish police and prosecutors are bound by the legality principle, which means that they are required to report all offenses that come to their attention, and that there are no legal grounds for the use of discretion in the handling of the criminal offenses during this phase of the criminal justice process (von Hofer, 2014). The age of the offender is instead taken into account by the prosecutor or the judge. It may also be noted that, in contrast to comparable countries such as the US, there is no youth court system in Sweden, and that sanctioning decisions relating to both juveniles and adults are made by the same courts (Sarnecki, 2017). Moreover, the Swedish police have no sanctioning mandate other than in relation to the imposition of fines for some minor traffic offenses such as speeding. With the exception of these minor offenses, Swedish conviction data therefore have a relatively high degree of coverage in relation to the number of offenses committed (von Hofer, 2014).4 Third, because there is a time lag between the date of an offense and the conviction date, the age of offending should ideally be registered when the offense was committed and not at the time of the conviction (Farrington et al., 2013). Swedish conviction data include the exact date of the principal offense for every conviction that was based on a court decision or where the prosecutor issued a waiver of prosecution. A conviction can take three forms: a court adjudication of guilt, being issued with a waiver of prosecution by the prosecutor, or the acceptance of summary sanction order issued by the prosecutor. If a prison sentence is not the normal outcome in accordance with the penal code, the prosecutor can choose to issue a summary sanction order or a waiver of prosecution on the condition that the suspect approves. Information on the date of the offense was missing in those cases where the prosecutor had issued a summary sanction order, which involved around one-third of all the convictions recorded for the cohort members

2. Data and method The full study population was drawn from the Swedish population register and comprises every individual born during the years 1960 through 1964 and who were resident in Sweden at age 15 – a total of 554,996 individuals. Since every Swedish resident has a unique personal identification number, it has been possible to link information on convictions, dates of death and migration dates, from the Swedish convictions register, the Swedish mortality register, and the Swedish migration register respectively (for an overview of the potential of Nordic register data, see Lyngstad & Skardhamar, 2011).3 Information on convictions was available for the period from 1973 to 2015, which

4 Due to changes in registration routines, a category of convictions concerning minor traffic violations were unavailable in the data before 1995. In order to not risk inflating adult-onset offending these convictions were excluded from the analysis. This solution was reasonable given that developmental and life-course theories of crime are not typically directed towards minor traffic violations (Farrington, 2003).

3 The research has been approved by the Regional Ethical Review Board in Stockholm (protocol 2016/5:2). All the data are stored at Statistics Sweden and have been made available via their Micro Data Online Access (MONA).

61

Journal of Criminal Justice 55 (2018) 58–70

F. Sivertsson

2.3. Analytical strategy for the trajectory analysis

between the ages of 15 and 55. In these cases, the date of the conviction was used instead. This is reasonable since this type of conviction is determined more swiftly and is only used when the offense is considered relatively minor and the offender has pleaded guilty. The time between the offense and the conviction decision in these cases is therefore usually short. Finally, besides the timing of the offense and the conviction date, the Swedish convictions register also contains information on the offense type. Nine offense categories were coded: Violent crimes (assault, aggravated assault, violence against a public servant, homicide, robbery, unlawful threat, unlawful compulsion, gross violation of integrity, unlawful persecution, molestation), Sex crimes (rape, sexual coercion, sexual exploitation), Property (theft, grand theft auto, shoplifting), Fraud (forgery, corruption, receiving stolen goods, perjury), Vandalism, Traffic crimes (recklessness in traffic, driving without a license, unauthorized deviation from traffic accident), Driving under the influence, Narcotics crimes (production of drugs, possession of drugs, distribution of drugs, use of drugs), and Other crimes.5 Since a conviction may include several offenses, all of the offenses in the conviction were examined when coding the offense type categories. Coding only the principal offense (i.e. the most severe offense type according to the penal code) would have resulted in an underrepresentation of less serious offenses. As a consequence, however, it is important to bear in mind that when a conviction included more than one of the above-listed offense types, it was counted as a conviction for any of the offenses included in the conviction.

Growth curve analysis is typically used to model the change in crime as a function of age and other covariates that may vary within (level 1) or between (level 2) individuals (Kreuter & Muthén, 2008). In the current study, growth curve analysis is employed within the general framework of multilevel modeling to analyze differences in the level and shape of average adulthood crime trajectories (age 25 to 50) by age-of-onset category and gender.6 Hierarchical prevalence models were fitted using Logistic regression and hierarchical count models were estimated using Poisson regression (see also Laub & Sampson, 2003).7 The outcome variable in the Poisson models measured the count of convictions related to offending at a particular age, while the outcome variable in the prevalence models was dichotomized to measure whether an individual had (1) or had not (0) committed an offense at a particular age. The dataset was collapsed into two-year age-bands to facilitate stability in the trajectories (see also Kreuter & Muthén, 2008). Robust standard errors were estimated to account for the nested data structure. Individuals were right censored at the age of death or the age of the first emigration event that occurred after age 15. To allow for non-linearity in the effect of age on crime, a quadratic term and a cubic term for age were included in the models in addition to the linear age covariate. As Grimm, Ram, and Hamagami (2011) have noted, a drawback of such non-linear growth curve models is that it becomes difficult to interpret time-invariant predictors (p. 1366). The focus was therefore directed at producing a visualization of growthcurve-specific crime trajectories (see also Kim & Bushway, 2017; Laub & Sampson, 2003). The first step involved modeling the average adulthood offending trajectories by age-of-onset category and gender. The next step was to analyze crime-specific trajectories relating to the typical domain of street crimes, as emphasized in developmental and lifecourse criminology (Farrington, 2003; Laub & Sampson, 2003): Violent crimes, Property crimes, and Drug/Alcohol crimes.8 Since these models were nested within every strata (i.e. growth-curve-specific) the likelihood ratio test was employed to evaluate whether the model was improved by the added age polynomial (see Rabe-Hesketh & Skrondal, 2012).

2.2. Analytical strategy for analyzing the age of offense onset and recidivism The complete birth cohorts were initially used to analyze how offense onset varied by age. Every individual born in 1960 to 1964 who reached the age of criminal responsibility in Sweden was thus included in the risk set. Non-parametric methods for continuous event history data were employed (see Allison, 2014). The clock started at age 15 and stopped at the age when a first offense was recorded. The observation period was right censored at the age of follow-up, the age of death or the age at the first emigration event that occurred after age 15. Cumulative probability functions were estimated to illustrate the proportion of individuals who committed a first offense for which they were convicted up to a certain age. Cumulative probability functions in this case constitute a very useful means of analyzing the age of onset because they illustrate the speed of the event process while also being able to account for censoring due to mortality and migration. Event history analysis was also used to analyze the association between age of onset and recidivism. A bivariate Cox proportional hazard model was fitted to predict recidivism tendencies with categorical age of onset as the independent variable. In contrast to fully parametric models, Cox regression makes no assumptions as to the specific distribution of duration dependency (Allison, 2014). The clock was started at the age of offense onset and stopped at the next offense date, or was right censored at the first emigration event or at the time of death. The follow-up period was fixed at five years. This is a reasonable truncation given that most of those who recidivate do so within the first few years (see Nygaard Andersen & Skardhamar, 2014). Age of onset was included as a biannual categorical variable with age 15 to 16 as the reference category, and was truncated at age 45 to ensure that every ageof-onset category was followed for an equal period of time.

6 It should be emphasized that conventional growth curve analysis differs in many respects from the influential group-based framework presented by Nagin and Land (1993). In particular, the growth curves estimated in the current study represent averages of offending trajectories among the a priori defined groups, and not latent classes of individual offending trajectories. This choice to depart from many of the recent trajectory studies stems from the research question examined here, which involves comparing groups of offenders who are defined by their age of onset. Because these categories represent broad groups of offenders, there will naturally be heterogeneity around the average trajectory. For a comparison of different approaches employed in the study of offending trajectories, see Kreuter and Muthén (2008). 7 It should be noted that crime data are usually overdispersed and zero-inflated which may cause concern with respect to the results obtained from the Poisson models. As a check for overdispersion the Poisson distribution was replaced with the negative binomial distribution (see Rabe-Hesketh & Skrondal, 2012). The results for the negative binomial models indicated overdispersion but yielded substantially similar rate parameters. With regard to zero-inflation, such a modeling approach introduces the assumption that some individuals are ineligible for a crime conviction which may be debated with respect to the controversy between general and typological theories of crime. As previously mentioned growth curve analysis was mainly employed to describe and compare the rate of change among predefined subcategories of offenders (see also Laub & Sampson, 2003) and not to explain these differences. 8 A separate model was estimated for each category of individuals defined by their ageof-onset category and gender, that is, one model for adolescent-onset males, one model for adolescent-onset females, one model for emerging-adult-onset males and so on. The same procedure was carried out in relation to offense-specific trajectories. Drug and alcohol-related offending are usually combined into one category of offenses (e.g. Laub & Sampson, 2003). Because conviction data is here used to measure drug use, it is likely to capture more serious substance misuse (Nilsson, Estrada, & Bäckman, 2014). DUI may be argued to capture similar substance dependency issues as narcotic offending, but in terms of alcohol misuse (Oksanen, Aaltonen, & Kivivuori, 2015).

5 Because this category included every type of offense that was not included in the other categories, it was also highly heterogeneous, and included for example the illegal import of goods, refusing mandatory military service, and the illegal use of public space for commercial purposes.

62

Journal of Criminal Justice 55 (2018) 58–70

F. Sivertsson

Fig. 1 illustrates the age of onset by means of cumulative probability functions in the full birth cohorts of males and females who were living in Sweden at age 15. Age 25 is marked to indicate the transition between emerging adulthood and adulthood. The figures show that both males and females were convicted for the first time at a relatively faster pace (i.e. steeper curve) during adolescence and through emerging adulthood than in adulthood. Around one-third of the male population had been convicted for the first time by age 25 while the corresponding proportion for the female population is around 8%. Still, there are clearly substantial proportions of previously non-convicted males and females who are convicted after age 25. The pace of new first-time convictions after age 25 is greater among the females, which indicates that adult-onset is, relatively speaking, more prevalent in the female population. Again, this is very much in line with previous research on gender differences in criminal careers (e.g. Block et al., 2010). Importantly, the patterns are almost identical across the five birth cohorts among males and females respectively. The following analyses are therefore based on the pooled dataset of convicted individuals. Fig. 2 presents the distributions of annual offense-onset ages among convicted males and females. The onset distributions replicate the aggregate age-crime curve and show that the peak ages of the first offense among convicted males and females most commonly occur during adolescence. This pattern is particularly evident among the males. There is a marked drop in the proportion of conviction debuts during adolescence and emerging adulthood, while in adulthood, the decline is much less pronounced. The long tail of the two distributions shows that there are individuals who are convicted for the first time for offenses they committed beyond emerging adulthood and all the way to age 50. In the convicted population of individuals 22% of the males and 38% of the females were convicted for the first time for offenses committed in adulthood. For a substantial proportion of both males and females the first conviction related to a traffic offense (around 33% of the males and 24% of the females), but as can be seen, the age-of-onset distributions remain virtually the same when these individuals are excluded from the analysis (lower panel).

Table 1 Description of the full study population. Individuals born 1960 to 1964 and residing in Sweden at age 15. Females

N = 284,456

N = 270,540

42.4 22.1

12.0 25.5

4.9 39.1

2.7 41.2

6.7 32.7

6.9 30.6

3. Results The first section below describes the age of offense onset distribution for the complete birth cohorts as well as for the offender populations. Next, age of onset is related to criminal career measures and recidivism over the full study period. The subsequent section examines how much adult-onset offenders contribute to the total number of adulthood convictions and to adulthood convictions for the different offense categories examined. Finally, adulthood crime trajectories are analyzed and compared across the age-of-onset categories and gender. 3.1. Age of onset from middle adolescence to middle adulthood

3.2. Age of onset and the criminal career Table 2 describes the offender population in terms of a number of criminal career measures by age of offense onset and gender over the full study period, from age 15 to age 50. As can be seen, the proportion of recidivists was substantially larger among juvenile-onset offenders. Almost three-quarters of the males whose first convictions occurred during adolescence were eventually reconvicted for a new offense, as

Failure rate 0.06 0.09 0.03

Failure rate 0.20 0.30 0.10 0.00

15

B. Females

0.12

0.15

A. Males

0.40

0.50

Table 1 presents an overview of the full study population with regard to the proportions and mean ages of males and females who were convicted over the follow-up period, and also the proportions and mean ages who either emigrated or died after age 15. As expected, the proportions of convicted males far outnumber the proportions of convicted females; 42,4% of the males and 12% of the females were convicted over the follow-up period. In line with previous research, the females had a later mean age of first convicted offense than the males (e.g. Block et al., 2010). The proportions of males who died during the follow-up period were also larger than the corresponding proportions of females; 4.9% of the males died between ages 15 and 55 as compared to only 2.7% of the females. The decease occurred on average around two years earlier among the males. Finally, the proportions of males and females who were residents at age 15 but who then emigrated were very similar; 6.7% of the males and 6.9% of the females, although males on average emigrated a few years later.

20

25

30

1960

35 Age

40

45

1961

50

55

0.00

Convicted Percent convicted Mean age at first offense Deceased after age 15 Percent deceased Mean age at decease Emigrated after age 15 Percent emigrated Mean age at emigration

Males

15

Birth cohort 1962

20

25

30

1963

35 Age

40

45

50

55

1964

Fig. 1. Age of offense-onset illustrated with cumulative probability functions. Males and females residing in Sweden at age 15. Note different scales.

63

Journal of Criminal Justice 55 (2018) 58–70

F. Sivertsson

B1. Females, any offense

15 20 25 30 35 40 45 50

15 20 25 30 35 40 45 50

10 8 0

2

4

6

Percent

12

14

16

A1. Males, any offense

Age of onset B2. Females, non-traffic offense

15 20 25 30 35 40 45 50

15 20 25 30 35 40 45 50

10 8 0

2

4

6

Percent

12

14

16

A2. Males, non-traffic offense

Age of onset Fig. 2. Distribution of annual offense-onset ages among first-time convicted males and females.

differ between individuals with different ages of onset in a cohort study which follows individuals to a certain year. As a way of accounting for this, and also as a means of exploring how recidivism tendencies vary by age of onset in a more nuanced fashion, a bivariate Cox model was fitted to estimate the hazard of recidivism over the five years subsequent to a first conviction using categorical (biannual) age of onset as the independent variable. Fig. 3 presents the predicted hazard ratios for males and females respectively, with age 15 to 16 as the reference category. A reference line marks the division between juvenile-onset and adult-onset offenders. Looking to the males in Panel A, there is a substantial decrease in the recidivism risk following a later age of onset during the juvenile years. Compared to the earliest onset category (age 15 to 16), the hazard of recidivism is almost 80% lower among those whose first conviction came at age 25 to 26. There is a relatively low and stable recidivism risk from early adulthood up to the mid-40s, showing that age 25 captures quite well the time when the association between age of onset and recidivism started to flatten out among the males. Turning to the females in Panel B, it may first be noted that there is an inverse association between age of onset and recidivism over the juvenile years, although this is not as strong as it is among the males. Compared to the earliest onset females, the hazard of recidivism is not quite 60% lower among those whose first convictions came at age 25 to 26. Interestingly, in contrast to the males, the association between age of onset and recidivism follows a u-shaped trend among the females,

compared to half of the emerging-adulthood-onset offenders and a quarter of the adult-onset offenders. The same inverse association is seen when career length or the mean numbers of life-time convictions are compared between the three groups. In terms of conviction characteristics, the males accumulated a total of near 445,000 convictions and the majority of these convictions, around 61%, were accumulated by adolescence-onset offenders, while only around 9% were acquired by adult-onset offenders. In general, these results are in line with previous research on male samples showing that there is an inverse association between age of onset and criminal career severity (DeLisi & Piquero, 2011). However, as can be seen from a comparison of Panels A and B these associations are substantially less pronounced among the females. In terms of recidivism, 45% of the adolescence-onset females recidivated compared to 21% of the adult-onset females. Adolescentonset females accumulated an average of 3,6 convictions per person while the corresponding figure among adult-onset females was 1,6. The females accumulated a total of over 72,000 convictions. In relation to the males, the volume of convictions was more evenly distributed between the different age-of-onset groups. The largest proportion of convictions was accumulated by emerging-adult-onset offenders (43%), while around 31% and 27% of convictions respectively were acquired by adolescence-onset offenders and adult-onset offenders. Because the criminal career, by definition, starts at the age of the first offense, the exposure time for measuring subsequent crime will 64

Journal of Criminal Justice 55 (2018) 58–70

F. Sivertsson

Table 2 Criminal career measures by age-of-onset group and gender. Age 15–50. Panel A. Males

Age-of-onset group Adolescence (15–17)

Individual-level description Convicted individuals One-time offenders Recidivists Age at first offense Age at last offense Career length Lifetime convictions per person Conviction-level description Total number of convictions Percentage of the total number of convictions

Panel B. Females

Emerging adulthood (18–24)

Adulthood (25–50)

n

%

n

%

n

%

42,403 11,412 30,991 Mean 16.6 27.8 11.2 6.4

100.0 26.9 73.1 SD 0.7 11.5 11.6 10.5

51,721 26,061 25,660 Mean 20.7 26.9 6.2 2.6

100.0 50.4 49.6 SD 1.8 9.2 9.2 3.9

25,995 19,424 6571 Mean 33.8 36.0 2.3 1.5

100.0 74.7 25.3 SD 7.4 8.0 5.2 1.5

269,274 60.5

136,574 30.7

39,093 8.8

Emerging adulthood (18–24)

Adulthood (25–50)

Age-of-onset group Adolescence (15–17)

Individual-level description Convicted individuals One-time offenders Recidivists Age at first offense Age at last offense Career length Lifetime convictions per person Conviction-level description Total number of convictions Percentage of the total number of convictions

n

%

n

%

n

%

6057 3334 2723 Mean 16.7 23.6 6.9 3.6

100.0 55.0 45.0 SD 0.8 10.8 10.9 7.0

13,982 9736 4246 Mean 21.0 24.9 3.9 2.2

100.0 69.6 30.4 SD 1.9 8.2 8.1 4.2

12,182 9588 2594 Mean 35.1 36.9 1.8 1.6

100.0 78.7 21.3 SD 7.6 8.1 4.6 2.4

21,961 30.5

31,036 43.0

A. Males

19,121 26.5

1 .9 Predicted hazard ratio .4 .5 .6 .7 .8 .3 .2 .1

.9 .3 .2 .1

Predicted hazard ratio .4 .5 .6 .7 .8

1

B. Females

15-16 19-20 23-24 27-28 31-32 35-36 39-40 43-44 Age of onset (biannual)

15-16 19-20 23-24 27-28 31-32 35-36 39-40 43-44 Age of onset (biannual)

Fig. 3. Bivariate association between age of onset and recidivism following a first conviction. Hazard ratios with onset at age 15–16 as reference group. 95% CI.

be as homogenous in relation to recidivism as is the case among the males, but it is those females who are first convicted of offenses later on in adulthood who display the strongest recidivism tendencies among the adult-onset females.

with the risk for recidivism being relatively stable by age of onset during early adulthood, but then increasing during the late 30s to the early 40s. Compared to the earliest onset females, the hazard of recidivism is around 40% lower among those whose first convictions came at age 39 to 40.9 Hence, the adult-onset females do not appear to

9 18.9% of females with an offense onset at ages 39–40 recidivated into a new convicted offense over a five year-follow up period, compared to 13.4% of females with an offense onset at ages 25–26. Over a ten-year follow-up period (the longest possible followup time for the 39–40 category) 26.0% of females with an offense onset at ages 39–40 recidivated into a new convicted offense, compared to 18.1% of females with an offense onset at ages 25–26. Hence, although individuals continued to recidivate after five years

(footnote continued) from the first convicted offense the ratio of reconviction risk between the two compared groups remained the same.

65

Journal of Criminal Justice 55 (2018) 58–70

F. Sivertsson

Table 3 Conviction characteristics between age 25 and 50 by age-of-onset group and gender. Panel A. Malesa

Total By offense typec Violent crimes Sex crimes Property Fraud Vandalism Traffic crimes (excl. DUI) DUI Narcotics crimes Other crimes

Age-of-onset group

a b c d

Emerging adulthood (18–24)

Adulthood (25–50)

N = 41,249

N = 51,105

N = 25,995

b

Rate per person

%b

Rate per person

26.8

1.07

19.2

1.50

28.0 27.4 23.7 26.0 27.9 22.2 27.5 21.9 27.2

0.22 0.01 0.24 0.12 0.06 0.25 0.19 0.14 0.23

16.0 36.8 7.8 15.8 13.5 14.2 21.5 4.1 25.3

0.24 0.03 0.16 0.15 0.06 0.31 0.30 0.05 0.41

Adult convictions

%

Rate per person

%

203,406

54.0

2.66

39,518 1957 51,981 23,901 11,643 57,512 35,682 33,535 42,607

56.0 35.8 68.5 58.2 58.6 63.6 51.0 73.9 47.6

0.54 0.02 0.86 0.34 0.17 0.89 0.44 0.60 0.49

Panel B. Femalesd

Total By offense typec Violent crimes Sex crimes Property Fraud Vandalism Traffic crimes (excl. DUI) DUI Narcotics crimes Other crimes

Adolescence (15–17)

b

Age-of-onset group Adolescence (15–17)

Emerging adulthood (18–24)

Adulthood (25–50)

N = 5897

N = 13,798

N = 12,182

b

Rate per person

%b

Rate per person

30.4

0.90

46.7

1.57

27.8 13.0 35.1 30.3 33.0 31.7 26.0 40.9 25.5

0.08 0.00 0.39 0.13 0.03 0.22 0.09 0.20 0.09

47.8 69.6 36.8 44.9 43.0 42.4 55.5 26.1 55.5

0.16 0.00 0.47 0.22 0.04 0.34 0.22 0.15 0.23

Adult convictions

%

Rate per person

%

40,928

22.9

1.59

4178 23 15,424 6055 1133 9634 4742 6859 4961

24.4 17.4 28.1 24.9 24.0 25.8 18.4 33.1 19.0

0.17 0.00 0.73 0.26 0.05 0.42 0.15 0.38 0.16

b

1770 males emigrated or died before age 25 and were therefore excluded from the analysis. Percentage figures show how the convictions within the sanction/offense type category are distributed between age of onset groups. Since a conviction may include several offenses, the number of convictions in the offense categories together exceed the sum of convictions. 344 females emigrated or died before age 25 and were therefore excluded from the analysis.

contribute to the largest share of convictions. Close to half of all female adulthood convictions (around 47%) relate to female adult-onset offenders, and this group is overrepresented in every offense category with the exception of narcotics offenses. For example, 48% of all female violent convictions, 56% of DUI convictions, and 45% of fraud convictions in adulthood were accounted for by female adult-onset offenders. Turning to the rate of convictions per person, adult-onset females and adolescence-onset females are very similar, whereas the females whose onset came during emerging adulthood generally have a lower rate of offending than the other groups. When comparing the patterns for male and female adult-onset offenders, the total rate of adult convictions per person are very similar. While the rate of violent crimes, DUI, and other crimes are higher among the male adult-onset offenders, female adult-onset offenders have higher rates of property crimes, fraud crimes and narcotics crimes.

3.3. How much do adult-onset offenders contribute to adulthood offending? The following analyses direct their attention at adulthood, from age 25 to age 50. Around 46% of the total number of male convictions accumulated in these birth cohorts, and around 57% of female convictions, were for offenses committed during adulthood. Table 3 shows how much the three age-of-onset groups contributed to adulthood crime (see also Beckley et al., 2016). Among the males, the adolescenceonset group was responsible for a majority of the adulthood convictions, 54%, while emerging-adulthood-onset and adult-onset offenders contributed to around 27% and 19% of all adulthood convictions respectively. The adolescence-onset males were also responsible for most of the convictions within the different offense categories; for example they contributed 56% of all convictions for violent crime, 69% of all property convictions, and 74% of all narcotics convictions. This indicates that the adolescence-onset male group probably captures a selected high-rate category of offenders. Still, as is indicated by the rate figures presented in Table 3, a substantial proportion of these earlyonset offenders did not contribute to adulthood offending. The rate differences between the age-of-onset categories are much smaller, and the male adult-onset offenders on average present a higher rate than the males who were first convicted of crimes during emerging adulthood. Turning to female adulthood offending, the pattern may be described as almost the reverse of that found among the males. Since adult-onset females represent the largest age-of-onset group, they also

3.4. Adulthood trajectories of offending It is well-known that early-onset offenders on average tend to have lengthier and more intensive criminal careers than adult-onset offenders (Farrington et al., 2013; Jennings et al., 2015). However, studies that have adopted a prospective approach to crime trajectories have shown that this overrepresentation mainly plays out during the juvenile years and that there is tendency towards decline even among the highest-risk delinquents as adulthood approaches (Laub & Sampson, 66

Journal of Criminal Justice 55 (2018) 58–70

F. Sivertsson

B1. Females, participation

.02

.02

Predicted participation .06 .1 .14 .18

Predicted participation .1 .14 .06 .18

.22

.22

A1. Males, participation

A2. Males, frequency

B2. Females, frequency Predicted count .04 .08 .12 .16 .2 .24 .28 .32 .36

25-26 29-30 33-34 37-38 41-42 45-46 49-50 Age (biannual)

Predicted count .04 .08 .12 .16 .2 .24 .28 .32 .36

25-26 29-30 33-34 37-38 41-42 45-46 49-50 Age (biannual)

25-26 29-30 33-34 37-38 41-42 45-46 49-50 Age (biannual)

25-26 29-30 33-34 37-38 41-42 45-46 49-50 Age (biannual)

Average trajectory

Adolescence (15-17)

Emerging adulthood (18-24)

Adulthood (25-50)

Fig. 4. Predicted participation and frequency trajectories for offending in adulthood by age-of-onset group and gender. 95% CI.

Piquero, 2010). As a final stage in the analysis, the participation trajectories have been divided into three typical offense categories, as emphasized in developmental and life-course criminology: violence, property, and drug/alcohol (Farrington, 2003; Laub & Sampson, 2003).10 In Fig. 5, a growth-curve-specific trajectory has been estimated for every strata of individuals defined by their age-of-onset category, type of onset offense, and gender. Among the males, it is clear that those who were first convicted of a crime in adolescence present higher trajectories than the males who were first convicted of a crime in emerging adulthood or adulthood, regardless of offense type. The adult-onset males and the emerging-adulthood-onset males are by contrast virtually indistinguishable in their trajectories. While the male trajectories generally display a decreasing participation rate over time, there is a slightly increasing tendency for drug/alcohol-related offending as middle adulthood approaches (Panel A3). Turning to the females, it may first be noted that the patterns are generally similar to those of the males in that adolescence-onset offenders present higher rates of property- and drug-related offending (Panels B2 and B3) than the groups whose first conviction came later. Among the females, however, there is a more marked tendency towards convergence over time, which means that adolescence-onset and adultonset offenders become increasingly similar in their participation rates

2003; Sivertsson & Carlsson, 2015). Additionally, little is known about adult-onset trajectories of offending. Fig. 4 show predicted participation and frequency trajectories in adulthood offending by the respective ageof-onset categories separated by gender. Looking at Panel A1, the male adolescence-onset offenders have a slightly higher participation rate than those who were first convicted for crime in adulthood. The male adult-onset offenders in turn have a higher participation rate than the emerging-adulthood-onset offenders. While the latter two groups present relatively parallel trends, there is a tendency towards convergence when the adolescence-onset offenders are compared with the adultonset offenders. Turning to the females, Panel B1 shows that the adultonset group has a higher rate of participation than the adolescenceonset offenders over the entire study period. As with the males, the adolescence-onset offenders have a higher participation rate than the emerging-adult-onset offenders. A comparison between Panels A1 and B1 shows that, with the exception of the adolescence-onset males, there are quite small gender differences in adulthood offending, particularly when male adult-onset offenders are compared with their female counterparts. Turning to the frequency trajectories presented in Panels A2 and B2, three things may be noted. First, the difference between the adolescence-onset males and the other male groups becomes even more distinct, indicating that this group probably comprises the most highfrequency offenders. Second, there is no longer a marked difference between the adult-onset females and the adolescence-onset females. Third, the participation and frequency patterns follow similar trends over time, that is, when participation in crime decreases so does the frequency of crime and vice versa (see also Petras, Nieuwbeerta, &

10 The same analysis was conducted on frequency trajectories but revealed no substantial differences in the shapes of the trajectories or in the level-ordering of the three onset groups and is therefore not shown.

67

Journal of Criminal Justice 55 (2018) 58–70

A2. Males, property

A3. Males, drug/alcohol

Predicted participation .04 .06 .08 .02

Predicted participation .01 .02 .03 .04 .05 .06 .07 .08

0

0

A1. Males, violence Predicted participation .01 .02 .03 .04 .05

F. Sivertsson

25-26 29-30 33-34 37-38 41-42 45-46 49-50

Age (biannual)

Age (biannual)

B2. Females, property

B3. Females, drug/alcohol

.01

Predicted participation .004 .008 .012 .016

B1. Females, violence

Predicted participation .01 .015 .02 .025 .03 .035 .04

25-26 29-30 33-34 37-38 41-42 45-46 49-50

Age (biannual)

Predicted participation .02 .03 .04 .05 .06

25-26 29-30 33-34 37-38 41-42 45-46 49-50

25-26 29-30 33-34 37-38 41-42 45-46 49-50

25-26 29-30 33-34 37-38 41-42 45-46 49-50

25-26 29-30 33-34 37-38 41-42 45-46 49-50

Age (biannual)

Age (biannual)

Age (biannual)

Adolescence (15-17)

Emerging adulthood (18-24)

Adulthood (25-50)

Fig. 5. Predicted participation trajectories for offending in adulthood by offense type, age-of-onset group and gender. 95% CI. Note different scales.

found that while there is a non-trivial proportion of adult-onset offending to be explained among males, there is a considerably more substantial proportion of adult-onset offending to be explained among females. In the convicted subpopulations of males and females, around 22% of males and 38% of females were first convicted for crimes they committed in adulthood (between age 25 and 50). While the adultonset males contributed to 19% of all male adulthood convictions and to 16% of male violent convictions, the corresponding figures among the adult-onset females were 47% of all female adulthood convictions and 48% of female violent convictions in adulthood. These figures should be viewed in light of the fact that almost half (46%) of the male convictions and 57% of the female convictions accumulated by these birth cohorts were for crimes committed during adulthood. Still, it must be emphasized that in terms of criminal careers, the findings from the current study are in line with contemporary wisdom in the sense that those who make their debut in officially recorded offending earliest, on average display higher-frequency and lengthier subsequent criminal careers than their later-onset counterparts (DeLisi & Piquero, 2011). However, this inverse association between age of onset and criminal career severity is nowhere near as strong among the females as among the males. The strength of this association among the males meant that the adolescence-onset offenders continued to be the dominant offending group in adulthood, particularly with regard to the typical forms of street criminality that have been emphasized by developmental and life-course criminology (e.g. Laub & Sampson, 2003). In contrast, the adult-onset offenders and the adolescence-onset offenders competed in terms of adulthood offending among the females. Thus although the early-onset females may be considered a particularly vulnerable group (e.g. Estrada & Nilsson, 2012), their adulthood criminal careers do not, on average, trump those of the female adultonset offenders. The age cut-off for adulthood used in the current study captured

as middle adulthood approaches. There is also a gender difference in that the adult-onset offenders generally have a higher participation rate than the emerging-adulthood-onset offenders. An interesting pattern can also be seen when looking at violent and drug/alcohol-related careers among adult-onset offenders (Panels B1 and B3). In contrast to the other trajectories, these display increasing rates of offending during the study period. This upward trend indicates that violent- and drug-related offending become increasingly important for an understanding of female offending as middle adulthood approaches. When it comes to violence, the adult-onset females have higher rates than the adolescence-onset offenders from their late 30s and onwards, and from the mid-40s the same is true with regard to drug/alcohol-related offending. 4. Discussion The current study has utilized longitudinal population-based data for five successive birth cohorts born in 1960 to 1964 and followed the cohort members to midlife on the basis of detailed conviction data. The dataset has provided a unique opportunity to make systematic comparisons across both males and females who initiated their criminal careers at quite different phases of the life course. While recognizing the critique advanced by the recent adult-onset literature that earlier adultonset studies have confused adulthood with emerging adulthood and have therefore overstated the phenomenon of adult-onset offending (Moffitt, 2006; Sohoni et al., 2014), the current study shows that the prevalence of adulthood-onset offending is likely to be underestimated in studies that do not follow individuals over a substantial segment of the life course. The initial analysis conducted on the full birth cohorts of males and females showed that previously non-convicted individuals continued to enter the population of convicted individuals all the way to age 55. Returning to the question posed in the title, the current study has 68

Journal of Criminal Justice 55 (2018) 58–70

F. Sivertsson

As has previously been highlighted by Andersson et al. (2012) in their study of a birth cohort up to age 30, a particular problem arises for asymmetrical theories in explaining the larger proportion of female adult-onset offenders found in comparisons with male adult-onset offenders. The current study provides further support for this conclusion in showing that those females who are convicted for their first time for offenses committed way up in adulthood display the highest recidivism risk among adult-onset females, and in showing that there is an upward trend in trajectories of violence and drug/alcohol-related offending as adult-onset females approach middle adulthood. An alternative to the gender-neutral approach to criminal careers is provided by the gendered pathways perspective (Belknap, 2014). Qualitative narratives provide some hypotheses of the mechanisms associated with female adult-onset offending. Carr and Hanks (2012), for example, interviewed eight incarcerated females who came into contact with the criminal justice system after age 20. The narratives provide an insight into possible mechanisms that may have resulted in these females becoming involved in crime after having lived “fairly conventional lives through their adolescence and early adulthood” (p. 400). One important aspect was the social control associated with the work of caring and having their children at home, which made crime a less accessible option. In particular, “nest emptying” constituted a transition which increased the risk for drug use among several of the females: “Frequently, the addiction was a contributing factor to late onset criminality” (p. 407). The results in the current study resonate well with the idea from the gendered pathways perspective that substance misuse constitutes a particularly important factor in understanding female adulthood offending. It is also interesting to note the upward trend in both drug/ alcohol offending and in the strength of the association between age of onset and recidivism at a time when the “nest is emptied” for many young mothers. While such an interpretation would be highly speculative on the basis of the current study, the findings presented above nonetheless suggest this as a possible hypothesis to be tested in future work focused on developing the knowledge on female adult-onset offending. In general, while criminal career patterns display considerable regularities at the aggregate level (e.g. MacLeod et al., 2012), they are highly complex at the individual level (e.g. Carlsson, 2011). The current study has presented a broad view of the phenomenon of adult-onset offending and has therefore not been able to examine the phenomenon in more detail. Future adult-onset research should dig deeper into the life experiences of individuals who are prosecuted for the first time in adulthood, and the consequences this has for their subsequent lives. The current study is not without limitations, the main one being that the measure of crime was limited to conviction records, or what Nygaard Andersen and Skardhamar (2014) term a back-end measure in the criminal justice process. Both individuals who offend and avoid apprehension, and individuals who offend and are apprehended but not proven guilty are not included in conviction records. It must therefore be emphasized that the proportions of adult-onset offenders reported in this study would most probably have been lower had a broader operationalization of crime and delinquency been used. In a more general sense, the phenomenon of adult-onset offending is heavily conditioned on how crime is measured, and the debate between the use of selfreported questionnaires and official crime records is likely to continue (Wiecko, 2014). What can safely be concluded from the current study is that substantial proportions of individuals are convicted for their first time for offenses they committed beyond emerging adulthood. In sum, the results resonate well with Eggleston and Laub's (2002) original notion that the adult-onset offender should not be neglected in criminal career research. On the contrary, the prediction made by the current study is that it will become increasingly important for developmental and life-course theories of crime to account for adult-onset offending as contemporary longitudinal studies are updated, providing longer follow-up periods.

quite well the time when the association between age of onset and recidivism started to flatten out among the males. In this sense, the adultonset males appear to constitute a rather homogenous group that presents a relatively constant and low risk for recidivism when compared to the juvenile-onset offenders. When the focus is directed at the females, a different picture emerges, in that it was the females who entered into officially recorded offending a long way into adulthood, in their early 40s, who displayed the strongest recidivism tendencies among female adult-onset offenders. This upward trend in offending is also found when the focus is directed at drug/alcohol-related offending and violence among adult-onset females. Together these findings indicate that middle adulthood would appear to be an important period of life for an understanding of female criminal careers. The overarching message communicated by the current study is that there is a need for developmental and life-course theories to provide more explicit accounts of adult-onset offending. While a case was recently made that adult-onset offending is a mere artifact of using official records (Sohoni et al., 2014; see also Wiecko, 2014), the mere association between normative delinquency and a crime conviction some two decades later appears to present a rather weak support for the falsification of symmetrical theories. As Beckley et al. (2016) noted developmental theories ought to explain how an individual may persist in and first be detected for antisocial behavior during adulthood (p. 68). In their thorough study of the Dunedin cohort, Beckley and colleagues did not find support for the hypothesis that adult-onset offenders were able to avoid detection and conviction as juveniles due to high socioeconomic status or high intelligence (Beckley et al., 2016), suggesting that more research into the link between normative delinquency and adult-onset offending is needed in favor of asymmetrical theories. Moffitt (2006) argued that if there are adult-onset offenders, they are probably accommodated by the theory of adolescence-limited offending, with the main reason for this being that their criminal careers tend to be brief and non-serious (p. 287). The findings from the current study provide some support for this argument in that the average adultoffending career was relatively short and most adult-onset offenders were only convicted once between ages 25 and 50. However, the results described above would suggest that the main mechanism for explaining adolescence-limited offending, the maturity gap, never closes for quite a substantial proportion of individuals in that individuals continue to make their debut in offending resulting in convictions from age 25 and all the way up to age 50. If an even more restrictive cut-off is employed, around 8% of the convicted males and 17% of the convicted females were convicted for the first time for offenses they committed at age 35 or later. Hence, how the adolescence-limited theory, and more specifically the maturity gap, explains adult crime is an important question in need of further clarification. Although the results in the current study does not directly support symmetrical theories such as Laub and Sampson's (2003) age-graded theory of informal social control, it resonates well with their overarching argument that the juvenile years does not capture the whole story of criminal careers. As with the offender population in general, there appears to be heterogeneity also in the criminal careers of adultonset offenders where a larger share are convicted only once and a smaller proportion continue into subsequent offending. It should therefore be of importance for future research to separate between onetime adult-onset offenders and adult-onset recidivists, both in exploring their past and their future. In general, the latter group ought to be more relevant to study in relation to criminal careers. Additionally, the agegraded notion suggests that the adult-onset literature ought to take into account a broader repertoire of crimes. While traditional property related offending is mainly distributed to the juvenile years, other utilitarian crimes ought to become more available in adulthood. For example, a recent study provided evidence for that a small proportion of low-risk individuals first started offending as they approached late adulthood, in their 60s, and that this group was primarily convicted for white-collar crimes (DeLisi et al., 2018). 69

Journal of Criminal Justice 55 (2018) 58–70

F. Sivertsson

Funding

(Eds.). Vardagens villkor: levnadsförhållanden i Sverige under tre decennier. Bromberg: Stockholm. Jennings, W. G., Loeber, R., Pardini, D. A., Piquero, A. R., & Farrington, D. P. (2015). Offending from childhood to young adulthood: Recent results from the Pittsburgh Youth Study. Springer. Kim, J., & Bushway, S. D. (2017). Using longitudinal self-report data to study the age–crime relationship. Journal of Quantitative Criminology, 1–30. http://dx.doi.org/ 10.1007/s10940-017-9338-9. van Koppen, M. V. (2018). Criminal career dimensions of juvenile-and adult-onset offenders. Journal of Developmental and Life-Course Criminology, 1–28. 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 Behaviour and Mental Health, 9(1), 57–73. Kreuter, F., & Muthén, B. (2008). Analyzing criminal trajectory profiles: Bridging multilevel and group-based approaches using growth mixture modeling. Journal of Quantitative Criminology, 24(1), 1–31. Krohn, M. D., Gibson, C. L., & Thornberry, T. P. (2013). Under the protective bud the bloom awaits: A review of theory and research on adult-onset and late-blooming offenders. Handbook of life-course criminology (pp. 183–200). New York: Springer. Kyvsgaard, B. (2002). The criminal career: The Danish longitudinal study. Cambridge University Press. Laub, J. H., & Sampson, R. J. (2003). Shared beginnings, divergent lives: Delinquent boys to age 70. Cambridge, MA: Harvard University Press. Lyngstad, T. H., & Skardhamar, T. (2011). Nordic register data and their untapped potential for criminological knowledge. Crime and Justice, 40(1), 613–645. MacLeod, J. F., Grove, P. G., & Farrington, D. P. (2012). Explaining criminal careers: Implications for justice policy. Oxford University Press. Macmillan, R., & McCarthy, B. (2014). Gender and offending in a life course context. The Oxford handbook of gender, sex, and crime (pp. 343–361). McGee, T. R., & Farrington, D. P. (2010). Are there any true adult-onset offenders? The British Journal of Criminology, 50(3), 530–549. Moffitt, T. E. (1993). Adolescence-limited and life-course-persistent antisocial behavior: A developmental taxonomy. Psychological Review, 100(4), 674. Moffitt, T. E. (2006). A review of research on the taxonomy of life-course persistent versus adolescence-limited antisocial behavior. Taking stock: The status of criminological theory. 15. Taking stock: The status of criminological theory (pp. 277–311). Moffitt, T. E., Caspi, A., Dickson, N., Silva, P., & Stanton, W. (1996). Childhood-onset versus adolescent-onset antisocial conduct problems in males: Natural history from ages 3 to 18 years. Development and Psychopathology, 8(2), 399–424. Moffitt, T. E., Caspi, A., Rutter, M., & Silva, P. A. (2001). Sex differences in antisocial behaviour: Conduct disorder, delinquency, and violence in the Dunedin longitudinal study. Cambridge, UK: Cambridge University Press. Nagin, D. S., & Land, K. C. (1993). Age, criminal careers, and population heterogeneity: Specification and estimation of a nonparametric, mixed Poisson model. Criminology, 31(3), 327–362. Nilsson, A., Estrada, F., & Bäckman, O. (2014). Offending, drug abuse and life chances—A longitudinal study of a Stockholm birth cohort. Journal of Scandinavian Studies in Criminology and Crime Prevention, 15(2), 128–142. Nygaard Andersen, S., & Skardhamar, T. (2014). Pick a number: mapping recidivism measures and their consequences. Discussion papers no. 772. Statistics Norway Research Department. Ohlsson-Wijk, S. (2011). Sweden's marriage revival: An analysis of the new-millennium switch from long-term decline to increasing popularity. Population Studies, 65(2), 183–200. Oksanen, A., Aaltonen, M., & Kivivuori, J. (2015). Driving under the influence as a turning point? A register-based study on financial and social consequences among first-time male offenders. Addiction, 110(3), 471–478. Petras, H., Nieuwbeerta, P., & Piquero, A. R. (2010). Participation and frequency during criminal careers across the life span. Criminology, 48(2), 607–637. Rabe-Hesketh, S., & Skrondal, A. (2012). Multilevel and longitudinal modeling using Stata. (3rd ed.). Categorical responses, counts and survival: Vol. 2. College Station, Tex.: Stata Press Publication. Sarnecki, J. (2017). Sweden. In S. H. Decker, & N. Marteache (Eds.). International handbook of juvenile justice. Springer International. Sivertsson, F. (2016). Catching up in crime? Long-term processes of recidivism across gender. Journal of Developmental and Life-Course Criminology, 2(3), 371–395. Sivertsson, F., & Carlsson, C. (2015). Continuity, change, and contradictions: Risk and agency in criminal careers to age 59. Criminal Justice and Behavior, 42(4), 382–411. Sohoni, T., Paternoster, R., McGloin, J. M., & Bachman, R. (2014). ‘Hen's teeth and horse's toes’: The adult onset offender in criminology. Journal of Crime and Justice, 37(2), 155–172. Svensson, R. (2002). Strategic offences in the criminal career context. British Journal of Criminology, 42(2), 395–411. Wiecko, F. M. (2014). Late-onset offending: Fact or fiction. International Journal of Offender Therapy and Comparative Criminology, 58(1), 107–129. Wikström, P. O. H. (1990). Age and crime in a Stockholm cohort. Journal of Quantitative Criminology, 6(1), 61–84. Zara, G., & Farrington, D. P. (2009). Childhood and adolescent predictors of late onset criminal careers. Journal of Youth and Adolescence, 38(3), 287–300.

The study was in part financed by the Swedish Research Council; grant number 2015-01201. Declarations of interest None. Acknowledgements I am grateful to Anders Nilsson, Felipe Estrada and Christoffer Carlsson at the Department of Criminology, Stockholm University, and Olof Bäckman at the Swedish Institute for Social Research (SOFI), Stockholm University for having provided me with valuable comments. References Allison, P. D. (2014). Event history and survival analysis: Regression for longitudinal event data. Vol. 46. Sage publications. Andersson, F. (2013). The female offender: Patterning of antisocial and criminal behaviour over the life-course. Faculty of Health and Society: Malmö University. Andersson, F., Levander, S., Svensson, R., & Levander, M. T. (2012). Sex differences in offending trajectories in a Swedish cohort. Criminal Behaviour and Mental Health, 22(2), 108–121. Arnett, J. J. (2000). Emerging adulthood: A theory of development from the late teens through the twenties. American Psychologist, 55(5), 469. Beckley, A. L., Caspi, A., Harrington, H., Houts, R. M., Mcgee, T. R., Morgan, N., & Moffitt, T. E. (2016). Adult-onset offenders: Is a tailored theory warranted? Journal of Criminal Justice, 46, 64–81. Belknap, J. (2014). The invisible woman: Gender, crime, and justice. Nelson Education. Bergman, L. R., & Andershed, A. K. (2009). Predictors and outcomes of persistent or agelimited registered criminal behavior: A 30-year longitudinal study of a Swedish urban population. Aggressive Behavior, 35(2), 164–178. Block, C. R., Blokland, A. A., van der Werff, C., van Os, R., & Nieuwbeerta, P. (2010). Long-term patterns of offending in women. Feminist Criminology, 5(1), 73–107. Blumstein, A., & Cohen, J. (1987). Characterizing criminal careers. Science, 237(4818), 985–991. Carlsson, C. (2011). Using ‘turning points’ to understand processes of change in offending: Notes from a Swedish study on life courses and crime. The British Journal of Criminology, 52(1), 1–16. Carr, N. T., & Hanks, R. S. (2012). If “60 is the new 40,” is 35 the new 15? Late onset crime and delinquency. Deviant Behavior, 33(5), 393–411. DeLisi, M. (2006). Zeroing in on early arrest onset: Results from a population of extreme career criminals. Journal of Criminal Justice, 34(1), 17–26. DeLisi, M., & Piquero, A. R. (2011). New frontiers in criminal careers research, 2000–2011: A state-of-the-art review. Journal of Criminal Justice, 39(4), 289–301. DeLisi, M., Tahja, K. N., Drury, A. J., Elbert, M. J., Caropreso, D. E., & Heinrichs, T. (2018). De novo advanced adult-onset offending: New evidence from a population of federal correctional clients. Journal of Forensic Sciences, 63(1), 172–177. 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(4), 349–375. Eggleston, E. P., & Laub, J. H. (2002). The onset of adult offending: A neglected dimension of the criminal career. Journal of Criminal Justice, 30(6), 603–622. Estrada, F., & Nilsson, A. (2012). Does it cost more to be a female offender? A life-course study of childhood circumstances, crime, drug abuse, and living conditions. Feminist Criminology, 7(3), 196–219. Farrington, D. P. (2003). Developmental and life-course criminology: Key theoretical and empirical issues-the 2002 Sutherland Award address. Criminology, 41(2), 221–225. Farrington, D. P., Piquero, A. R., & Jennings, W. G. (2013). Offending from childhood to late middle age: Recent results from the Cambridge Study in Delinquent Development. Springer Science & Business Media. Gomez-Smith, Z., & Piquero, A. R. (2005). An examination of adult onset offending. Journal of Criminal Justice, 33(6), 515–525. Grimm, K. J., Ram, N., & Hamagami, F. (2011). Nonlinear growth curves in developmental research. Child Development, 82(5), 1357–1371. Hayford, S. R., & Furstenberg, F. F. (2008). Delayed adulthood, delayed desistance? Trends in the age distribution of problem behaviors. Journal of Research on Adolescence, 18(2), 285–304. Hofer, v. H. (2014). Crime and reactions to crime in 34 Swedish birth cohorts: From historical descriptions to forecasting the future. Journal of Scandinavian Studies in Criminology and Crime Prevention, 15(2), 167–181. Hörnqvist, M. (1994). Att bli vuxen i olika generationer. In J. Fritzell, & O. Lundberg

70