Persistent offenders and adolescence‐limited offenders : differences in life‐courses

This is a self-archived version of an original article. This version may differ from the original in pagination and typographic details. Author(s): Title: Year: Version: Copyright: Rights: Rights url: Please cite the original version: CC BY 4.0 https://creativecommons.org/licenses/by/4.0/ Persistent offenders and adolescence‐limited offenders : differences in life‐courses © 2020 The Authors. Criminal Behaviour and Mental Health published by John Wiley & Sons Ltd Published version Pulkkinen, Lea; Fadjukoff, Päivi; Pitkänen, Tuuli Pulkkinen, L., Fadjukoff, P., & Pitkänen, T. (2020). Persistent offenders and adolescence‐limited offenders : differences in life‐courses. Criminal Behaviour and Mental Health, 30(4), 196-209. https://doi.org/10.1002/cbm.2157 2020 ORIGINAL ARTICLE Persistent offenders and adolescence-limited offenders: Differences in life-courses Lea Pulkkinen 1 | Päivi Fadjukoff 1 | Tuuli Pitkänen 2 1 Department of Psychology, University of Jyvaskyla, Jyväskylä, Finland 2 A-Clinic Foundation, Helsinki, Finland Correspondence Lea Pulkkinen, Sulkapolku 6 B 37, 00370 Helsinki, Finland. Email: [email protected]i Abstract Background and Aims: As our previous study indicated, almost half of juvenile delinquents continued offending in adulthood, while the rest ceased to do so. We compared these groups with each other and with non-offenders in the lifecourse use of alcohol, identity development and life situation. Methods: Based on the Jyväskylä Longitudinal Study of Personality and Social Development, four groups were formed at age 42 for men and women: persistent, adolescence-limited and adult-onset offenders and nonoffenders. Longitudinal data (N= 369; 53% males) have been collected at ages 8, 14, 20, 27, 36, 42 and 50. Results: Persistent offending, but not adolescence-limited offending, was associated with the accumulation of problems in adulthood and an elevated risk of death before age 54. For males, persistent offending was associated with heavy drinking from adolescence to middle age, diffuse identity, high divorce rates and poverty. For females, persistent offending was associated with heavy drinking in adulthood and a higher rate of being outside the workforce in middle age. In males, adolescence-limited offending was associated with more controlled drinking in adulthood, and in females, with early divorce from the partner and advanced identity. Conclusions: Detecting the risks of the accumulation of problems and potential for positive transitions of juvenile delinquents until middle age will be important for prevention and treatment. Received: 28 April 2020 Accepted: 9 May 2020 DOI: 10.1002/cbm.2157 This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. © 2020 The Authors. Criminal Behaviour and Mental Health published by John Wiley & Sons Ltd Crim Behav Ment Health. 2020;1–14. wileyonlinelibrary.com/journal/cbm 1 KEYWORDS drinking behaviour, identity, life-course, middle age, offender, selfcontrol 1|INTRODUCTION Rolf Loeber with his collaborators (Loeber, Green, & Lahey, 2003) stresses the need for prospective longitudinal studies on antisocial behaviour. The Jyväskylä Longitudinal Study of Personality and Social Development (JYLS; Pulkkinen, 2017) offers opportunities to prospectively analyse pathways of antisocial behaviour. Our present interest was in possible differences between offender groups such as the life-course-persistent and adolescence-limited offenders (ALO; Moffitt, 1993) for both genders in the use of alcohol, identity development and various life outcomes and in the accumulation of problems. Continuity in the use of alcohol is high from adolescence to adulthood (Pitkänen, Kokko, Lyyra, & Pulkkinen, 2008). Heavy drinking increases the risk of criminal behaviour: 56% of the convicted assaults, 35% of robberies and 11% of thefts in 2010 were committed under the influence of alcohol; the percentages for other drugs were 0.4, 4 and 2%, respectively (Yearbook, 2011). Due to a high connection between criminal offending and the use of alcohol already in the teenage years (Gottfredson & Hirschi, 1990), our first hypothesis was that juvenile delinquency is associated with the use of alcohol. The second hypothesis was that the association between criminal offending and heavy drinking remains in individuals who continue offending after adolescence but declines in those who cease offending. The third hypothesis concerned the accumulation of problems such as heavy drinking and antisocial behaviour (Rönkä & Pulkkinen, 1995), conceptualised as syndromes (Jessor & Jessor, 1977) or patterns (Magnusson & Bergman, 1988), through contemporary, interactive or cumulative continuity, as analysed by Caspi, Bem, and Elder (1989). We expected that heavy drinking, criminal arrests, health problems and poverty accumulate until middle age particularly in lifecourse persistent offenders (PO). Prospective longitudinal studies suggest that there is a higher number of risk factors associated with life-course offending than adolescent-limited offending (Jolliffe, Farrington, Piquero, Loeber, & Hill, 2017). Our fourth hypothesis was that differences exist in identity development between offender groups. Identity means a self-structure that provides a person with a sense of consistency and continuity, thus enabling personal psychosocial progress and well-being (Erikson, 1968). Studies in adolescent and student populations have demonstrated that heavy alcohol use at a young age is associated with less mature identity development, whereas higher identity achievement is related to lower levels of alcohol consumption (e.g. Bishop, Weisgram, Holleque, Lund, & WheelerAnderson, 2005; Laghi, Baiocco, Lonigro, & Baumgartner, 2013). We have found that advanced identity is associated with the length of education (Fadjukoff, Kokko, & Pulkkinen, 2007), precedes later social well-being (Fadjukoff & Pulkkinen, 2006), and is associated with adaptive adjustment in middle age (Fadjukoff, Feldt, Kokko, & Pulkkinen, 2019). Thus, we expected to find less advanced identity development in life-course PO. Our last hypothesis was that, in accordance with the general theory of crime (Gottfredson & Hirschi, 1990), low self-control conceptualised within the impulse control model (Pulkkinen, 2017) characterises both male and female offenders and that antisocial propensity is found particularly in life-course PO. 2|METHODS 2.1 |Sample The initial sample of the JYLS (Pulkkinen, 2017) consisted of 12 entire school classes (369 students; 53% males, born in 1959) randomly drawn from the second-grade classes of different schools of a town in Central Finland. The major 2PULKKINEN ET AL. data collections were conducted (with participation rates from the eligible sample in parentheses) when the participants were at the ages of 8 (100%), 14 (97%), 27 (88%), 36 (86%), 42 (83%) and 50 (84%) years. The retention rate was close to 90% until age 36 but declined to 77% at age 42 and 73% at age 50 due to deaths, illnesses and other reasons. A subsample was interviewed at ages 14 and 20 (Pulkkinen, 2017). By age 50, 10% of the initial sample had withdrawn from the study and 3.3% had died. It had been found (Pulkkinen, 2017) that there was a high continuity in drinking patterns through adulthood and that several heavy drinkers died by age 50 or were less likely than others to participate in data collection at age 50 due to their poor condition. Scores for middle age variables were supplemented by scores for age 42, if available, in case of missing information at age 50; this procedure was applied up to 35 persons (12%). In adulthood, the studied sample was compared in demographic variables with the age cohort born in Finland in the same year (1959) by using information from Statistics Finland. The demographic characteristics of the sample corresponded to the characteristics of the general population in marriage rate, family type, number of children, education and employment status (Pulkkinen, 2017). 2.2 |Offender categories For the study of antisocial behaviour of the participants, the national register of criminal convictions was searched every 5 years between ages 21 (1981) and 47 (2006). The local police register was searched at age 20 (in 1980) for information about antisocial behaviour between ages 15 and 20. Personal interviews at ages 27, 36 and 42 were utilised to supplement the register-based information about antisocial paths. The Self-Report Delinquency Scale (Junger-Tas, Terlow, & Klein, 1994) was administered at age 36 and the Life History Calendar (Caspi et al., 1996) at age 42. Antisocial behaviour of the male participants was assessed and categorised using different sources of information (Pulkkinen, Lyyra, & Kokko, 2009); the corresponding procedure was applied to women for the present study. We formed four offender groups (Figure 1) as suggested by Glueck and Glueck (1950, p. 285): ‘delinquent boys and girls who abandon their antisocial behaviour’(ALO), those who ‘continue in careers of crime’(PO), non-delinquents who ‘later develop into delinquents’(adult-onset offenders, AOO) and ‘the rest of the non-delinquents’(nonoffenders, NO). Speeding and arrests for drunkenness were not used as criteria for the categorisation of offending if they had appeared only once. The peak age for the onset of delinquency was 16–17 years (Pulkkinen, 1988). In adulthood, the incidence of crimes decreased. For the PO, the range of the duration of the criminal career extended FIGURE 1 Offender groups and their proportions of the initial sample [Colour figure can be viewed at wileyonlinelibrary.com] PULKKINEN ET AL.3 from about 7 to 25 years including juvenile delinquency before age 21 and convicted crime in adulthood. The prevalence of the offender categories is higher if both self-reports and official record are considered (Jolliffe, Farrington, Piquero, MacLeod, & van de Weijer, 2017), as it was the case in our study. 2.3 |Measures Data were collected using several methods (Pulkkinen, 2017), such as teacher rating and peer nomination at ages 8 and 14; personal interviews by trained interviewers at ages 27, 36, 42 and 50; a Life Situation Questionnaire (LSQ) mailed to the participants and inventories. Socioemotional behaviour: Peer nomination at age 14 was used (Pitkänen, 1969) for assessing same sex classmates in variables indicating low self-control (aggression, anxiety and emotional lability) and high self-control (constructiveness, emotional stability and compliance) within a two-dimensional impulse control model (Pulkkinen, 2017, pp. 24–26). Problem behaviour: Teachers' reports on their students' conduct problems at age 14 were collected by asking how frequently the student had been disciplined at school, involved in truancy, smoking and drinking and if the student had been in trouble with the police or other enforcement authorities (Savolainen, Mason, Lyyra, Pulkkinen, & Kokko, 2017). The variable was standardised with all participants (n= 351). Antisocial behaviour: The number of types of offences committed from age 15 to 20 was coded in the scale from 0 = no offences to 3 = three types of offences (alcohol, theft and violent offences; Pulkkinen, 1983). The age of onset of drinking was determined at ages 14, 20, 27, 36 and 42, with the priority given to the information received closest to the actual onset age (Pitkänen, Lyyra, & Pulkkinen, 2005). Alcohol-related offences between ages 15 and 20 were coded using police register data. The sum of them included offences such as drunk driving, selling alcohol and detention for intoxication. Drinking to intoxication at age 20 was coded using interview data at age 20. The categories were: 0 = does not use alcohol, 1 = does not drink to intoxication, 2 = drinks sometimes to intoxication and 3 = drinks often to intoxication. Heavy drinking at ages 27, 36 and 50 was operationalised on the basis of the reported times of drinking to intoxication but controlled with the frequency of drinking at least five portions per occasion as reported in the quantity– frequency table (Pitkänen et al., 2008) included in the mailed LSQ. Heavy drinking was coded from 0 = not at all to 5 = several times a week. Risky annual drinking at age 50 was counted using a quantity–frequency table (Pitkänen, 2006). Separate coding was used for females (0 = less than 7,000 g of pure alcohol per year, 1 = 7,000–10,000 g per year and 2 = over 10,000 g per year) and for males (0 = less than 10,000 g per year, 1 = 10,000–15,000 g per year and 2 = over 15,000 g per year). Two screening tests were used (Pitkänen et al., 2005): the Cut-down, Annoyed, Guilt, Eye-opener (CAGE) questionnaire developed by Ewing and Rouse (Ewing, 1984) by coding the responses in a three-point scale (0 = no, 1 = sometimes and 2 = often) and the modified version of the Michigan Alcoholism Screening Test (MAST) developed by Selzer (1971). The latter consisted of 11 items including 2 items concerning alcohol dependence (Kaprio, 2006). Problems due to drinking at age 50 were coded on the basis of the interview in which the participants were asked if they had experienced any of seven difficulties, such as being involved in a (physical) fight, an accident, absence from work and personal relationship because of drinking. The response options for each question were 0 = no, 1 = sometimes and 2 = often. Treatment for substance use disorder by age 50 was based on data at health interviews at ages 36, 42 and 50 and at the life-history calendar. It was coded by 1 = yes or 0 = no. 4PULKKINEN ET AL. Self-rated health was measured with a widely used single-item scale (Smith & Goldman, 2011) at ages 27, 36, 42 and 50. The options were 1 = very good, 2 = good, 3 = moderate, 4 = poor and 5 = very poor. However, at age 27, the options were 1 = good, 2 = moderate and 3 = poor. Death: Data from the Statistics Finland have been obtained about the deaths of the participants of the longitudinal study until 2014 (age 54). It was coded 1 = yes or 0 = no. Unemployed or retired at age 50 referred to people who were outside the work force because of unemployment or early retirement due to chronic disability (Savolainen et al., 2017). The data on unemployment and disability pension were collected by the mailed LSQ and validated and supplemented by the retrospective Life History Calendars administered at ages 42 and 50. It was coded 1 = yes or 0 = no. Persistent poverty at age 50 (43) was operationalised as having annual income below 60% of the median (Savolainen et al., 2017) for two consecutive data collections (1 = yes; 0 = no). The data on income were obtained from government tax records at ages 43 and 50. Divorce at least once until ages 28, 37 and 50 was assessed using all data available from the LSQ, personal interview and Life History Calendar. It was coded 1 = yes or 0 = no. Identity status: The identity interview, based on the Erikson (1968) traditions and utilising Marcia's (1966) identity status paradigm, was conducted at ages 27, 36, 42 and 50 (Fadjukoff, 2007; Fadjukoff, Pulkkinen, & Kokko, 2016). Five domains were considered: religious beliefs, political ideology, occupational career, intimate relationships and lifestyle. On the basis of two dimensions (commitment and exploration), four identity statuses were defined: diffusion (no exploration, no commitment), moratorium (yes exploration, no commitment), foreclosure (no exploration, yes commitment) and identity achievement (yes exploration, yes commitment). 2.4 |Data analyses Differences between the means of the four offender groups with 95% confidence intervals (95% CI) for all study variables enabling cross-sectional and longitudinal comparisons were reported online in Supporting Information Tables S1 and S2. Separately for males and females. One-way ANOVA and Bonferroni tests were used for comparing differences between the reported means, p< .05. Only significant differences between the means of the offender groups are presented in the text and tables. Additionally, ttests were used for studying the cross-sectional differences between the means of the offender groups and chi-squared tests for cross-tabulations. For effect size, Cramér's Vwas used, with 0.1–0.3 for small, 0.3–0.5 for medium and over 0.5 for large effect sizes. 3|RESULTS 3.1 |Adolescent behaviour 3.1.1 |Socioemotional behaviour For the male groups of offenders and NO, the PO differed from the ALO in higher aggression, lability and behaviour problems at school at age 14, and in more varied offences from age 15 to 20 (Table 1). The PO had been lower in anxiety than the NO. For females, the AOO had been more aggressive than the NO and less anxious than the ALO. No significant differences were found between the groups in the variables for high self-control. There were, however, differences in compliance when the NO were compared with the rest. The NO were more compliant than their counterparts, t(5.1) = 2.10, p= .039 for males; t(160.1) = 2.54, p= .012 for females, as well as more passive, t (184) = 3.67, p< .001 for males; t(163) = 2.30, p= .023 for females. PULKKINEN ET AL.5 TABLE 1 Significant differences between the means of the offender groups in variables for adolescent behaviour, use of alcohol and adult outcomes, separately for men and women; one-way ANOVA a Men 0 NO 1 AOO 2 ALO 3 PO FpBonferroni Non-offender offender Adult-onset offender Adolescence-limited offender Persistent offender Mean Mean Mean Mean Socioemotional behaviour Age 14 Aggression 186 11.1 23.9 12.7 29.1 6.20 .001 0,2 < 3 14 Anxiety 186 24.7 19.6 13.9 8.1 4.42 .005 3 < 0 14 Lability 186 14.8 19.8 11.4 25.6 4.20 .007 2 < 3 14 Activity 186 13.0 25.8 15.5 20.1 3.06 .03 0 < 1 14 Passivity 186 24.7 12.5 11.3 10.1 5.26 .002 2,3 < 0 Anti-social/problem behaviour 14 Problem behaviour 189 −0.25 −0.01 0.18 2.14 5.37 .001 0,1,2 < 3 15–20 Types of offences 196 0.15 0.19 0.60 0.98 14.88 .001 0 < 2,3;1 < 3 Alcohol use 14–42 Onset age of drinking 191 16.2 15.6 14.9 14.9 3.79 .011 0 < 2,3 15–20 Alcohol offences 196 0.18 0.28 0.68 2.05 9.47 .001 0,1,2 < 3 20 Drinking to intoxication 66 1.59 1.64 2.06 2.33 4.49 .006 0 < 3 27 Heavy drinking 169 2.07 3.12 2.46 3.18 7.49 .001 0 < 1,3;2 < 3 36 Heavy drinking 160 2.08 2.70 2.27 3.27 6.52 .001 0,2 < 3 50 Heavy drinking 162 2.10 2.54 2.18 3.42 6.31 .001 0,2 < 3 50 Risky annual drinking 157 0.35 0.45 0.34 0.77 3.15 .027 2 < 3 50 MAST, 11 questions 143 2.00 3.68 2.32 4.19 4.70 .004 0.2 < 3 50 CAGE, 4 questions 160 1.41 2.48 1.98 3.47 8.03 .000 0.2 < 3 50 Problems due to drinking 156 0.36 1.17 1.04 2.50 7.00 .000 0,2 < 3 6PULKKINEN ET AL. TABLE 1 (Continued) Men 0 NO 1 AOO 2 ALO 3 PO FpBonferroni Non-offender offender Adult-onset offender Adolescence-limited offender Persistent offender Mean Mean Mean Mean By 50 Treatment for substance use 169 0.00 0.22 0.02 0.29 4.40 .005 0,2 < 3 Social and economic problems By 36 Divorce (1 = yes, 2 = no) 172 0.16 0.45 0.35 0.47 3.77 .012 0 < 3 By 50 Divorce 185 0.25 0.47 0.56 0.53 4.01 .009 0 < 2,3 50 (43) Persistent poverty 179 0.06 0.29 0.10 0.35 6.42 .000 0,2 < 3 Self-rated health and death By 54 Death by 2014 196 0.02 0.13 0.04 0.14 2.74 .045 Women 0 NO 1 AOO 2 ALO 3 PO FpBonferroni Non-offender offender Adult-onset offender Adolescence-limited offender Persistent offender Mean Mean Mean Mean Socioemotional behaviour Age 14 Aggression 165 11.9 26.3 21.9 21.4 3.30 .022 0 < 1 14 Anxiety 165 14.7 3.3 19.5 5.4 3.20 .025 1 < 2 14 Lability 165 15.2 29.1 26.5 27.6 3.75 .012 14 Activity 165 19.0 29.3 25.8 39.5 2.64 .051 14 Stability 165 22.6 21.6 16.1 23.7 0.68 .566 Anti-social/problem behaviour 15–20 Types of offences 173 0.05 0.04 0.41 0.55 10.37 .001 0,1 < 2,3 Alcohol use 14–42 Onset age of drinking 165 15.9 15.6 14.5 14.9 2.65 .051 (Continues) PULKKINEN ET AL.7 TABLE 1 (Continued) Women 0 NO 1 AOO 2 ALO 3 PO FpBonferroni Non-offender offender Adult-onset offender Adolescence-limited offender Persistent offender Mean Mean Mean Mean 15–20 Alcohol offences 173 0.05 0.00 0.19 1.55 6.18 .001 0,1,2 < 3 20 Drinking to intoxication 67 1.43 1.33 1.91 2.33 3.69 .016 27 Heavy drinking 154 1.24 1.48 1.67 2.63 4.35 .006 0 < 3 36 Heavy drinking 150 1.38 1.50 1.69 2.80 4.06 .008 0,1 < 3 50 Risky annual drinking 114 0.09 0.27 0.05 0.60 4.23 .007 0,2 < 3 50 MAST, 11 questions 108 1.50 0.90 1.15 3.10 3.49 .018 1,2 < 3 50 CAGE, 4 questions 115 1.18 1.13 1.19 2.80 2.80 .043 0 < 3 50 Problems due to drinking 112 0.28 0.41 0.76 1.80 5.68 .001 0,1 < 3 By 50 Treatment for substance use 122 0.01 0.00 0.00 0.20 6.50 .000 0,1,2 < 3 Social and economic problems By 27 Divorce (1 = yes 0 = no) 162 0.15 0.24 0.44 0.40 4.34 .006 0 < 2 By 36 Divorce 152 0.32 0.58 0.63 0.70 5.05 .002 0 < 2 By 50 Divorce 162 0.35 0.56 0.70 0.70 5.23 .002 0 < 2 50 Unemployed or retired 135 0.14 0.09 0.09 0.50 3.89 .011 0,1,2 < 3 Self-rated health and death Abbreviations: CAGE, Cut-down, Annoyed, Guilt, Eye-opener; MAST, Michigan Alcoholism Screening Test. a Means of the offender groups with 95% confidence intervals and F-ratios for all study variables are presented in a Supporting Information Table S1. 8PULKKINEN ET AL.