رفع ابهام در مورد رابطه بین بزهکاری و بیش فعالی، موفقیت پایین، افسردگی و وضعیت اجتماعی اقتصادی پایین: تجزیه و تحلیل داده های تکراری طولی

Abstract Purpose To test hypotheses about causal linkages among hyperactivity, low academic achievement, depression, low SES, and delinquency. Methods 503 boys were followed up in the Pittsburgh Youth Study. Comparable measures of all variables at each age from 11 to 15 are analyzed. Cross-lagged panel models are tested. Results Hyperactivity, depression and achievement decreased with age, while SES and delinquency increased with age. The analyses suggest that hyperactivity and low SES caused low achievement, which in turn caused delinquency, which in turn caused depression. Conclusions Depression is not a risk factor for delinquency. These analyses should be repeated with larger numbers of variables. Developmental and life-course theories should propose and test sequential rather than simultaneous influences on offending. Since low achievement has the most direct influence on delinquency, interventions should target low achievement rather than hyperactivity or SES.

Introduction Hyperactivity, low academic achievement, depression, and low socioeconomic status (SES) are viewed as important risk factors for delinquency (see e.g. Derzon, 2010, Farrington et al., 2012, Farrington et al., 1990, Loeber et al., 1998a and Loeber et al., 1998b). A risk factor is defined as a variable that precedes and predicts an outcome such as delinquency. A risk factor is considered to have causal effects if changes in the risk factor are followed by changes in the outcome with high internal validity (see, e.g., Murray, Farrington, & Eisner, 2009). Therefore, longitudinal research is needed to investigate risk factors and causal risk factors. While the most important risk factors for delinquency have been well established for many years and are highly replicable across time and place (e.g., Farrington & Loeber, 1999), less is known about causal influences and intervening mechanisms. For example, if hyperactivity and low academic achievement both predict delinquency, is it that hyperactivity causes low academic achievement which in turn causes delinquency, so that hyperactivity only indirectly influences delinquency through the mediating factor of low academic achievement? Similarly, if low SES and low academic achievement both predict delinquency, is it that low SES causes low academic achievement which in turn causes delinquency, so that low SES only indirectly influences delinquency through the mediating factor of low academic achievement? The main aim of this paper is to investigate these alternative sequences of risk factors and mediating processes. The relationship between depression and delinquency is particularly perplexing. Depression is positively related to delinquency (e.g., Loeber et al., 1998a and Loeber et al., 1998b). However, depression is classified as an internalizing problem and is positively related to other internalizing problems such as anxiety and shyness/withdrawal (Achenbach & Edelbrock, 1983). Nevertheless, anxiety and shyness/withdrawal are often negatively related to delinquency and are sometimes regarded as protective factors against delinquency. For example, in the Cambridge Study in Delinquent Development, which is a longitudinal survey of over 400 London boys, Farrington, Gallagher, Morley, St. Ledger, and West (1988) found that boys from criminogenic backgrounds who did not become offenders tended to have few or no friends at age 8. Similarly, Kerr, Tremblay, Pagani, and Vitaro (1997), in the Montreal longitudinal-experimental study of over 1,000 boys, concluded that behavioral inhibition (anxiety) protected boys against delinquency. The present paper aims to advance knowledge about the relationship between depression and delinquency. There have been previous attempts to investigate causal effects and mediating mechanisms (see, e.g., Baron and Kenny, 1986 and Hayes, 2009). For example, McGloin, Pratt, and Maahs (2004), using US National Longitudinal Survey of Youth data, concluded that the relationship between low intelligence and delinquency was mediated by low school achievement, low self-control, and deviant peer pressure. Masten et al. (2005), in a Minneapolis longitudinal study of over 200 children from age 8 to age 20, concluded that childhood externalizing behavior (aggression and delinquency) led to low academic achievement in adolescence, which in turn led to externalizing and internalizing problems later in life. The present study goes beyond previous research by including more risk factors and by analyzing comparable annually collected data in the Pittsburgh Youth Study (PYS; see later). Annually collected data in the PYS was previously used to study causal effects by comparing within-individual analyses and between-individual analyses (Farrington, Loeber, Yin, & Anderson, 2002). The authors found that poor parental supervision predicted a boy's delinquency both between and within individuals, but that peer delinquency predicted a boy's delinquency between individuals but not within individuals. In other words, changes in peer delinquency within individuals (from one assessment to the next) did not predict subsequent changes in a boy's delinquency from one assessment to the next. This suggested that peer delinquency might not be a cause of a boy's delinquency but might instead be measuring the same underlying construct (perhaps reflecting co-offending). In contrast, poor parental supervision was predictive within individuals and therefore might be a causal factor. These kinds of analyses can only be carried out in a study such as the PYS with numerous comparable assessments repeated at regular intervals. The present paper uses similar PYS data to investigate causal linkages between hyperactivity, low academic achievement, depression, low SES, and delinquency. The following are plausible hypotheses which will be tested: 1. Hyperactivity, low achievement, depression, and low SES cause delinquency 2. Delinquency causes hyperactivity, low achievement, and depression. The hypothesis that delinquency of a boy causes low SES of his parents seems very unlikely and was not tested. 3. Hyperactivity causes low achievement which in turn causes delinquency. The alternative hypothesis that low achievement causes hyperactivity which causes delinquency was also tested. 4. Low achievement causes depression which in turn causes delinquency. The alternative hypothesis that depression causes low achievement which causes delinquency was also tested. 5. Low SES causes hyperactivity, low achievement, and depression, which in turn cause delinquency. These hypotheses were tested in cross-lagged panel models

Results Changes with age Table 1 shows the means and standard deviations of all variables. The N's typically reduced from about 475 at age 11 to about 445 at age 15. Generally, hyperactivity and depression decreased from age 11 to age 15, while low academic achievement, SES, and delinquency increased with age. Table 1. Means and standard deviations of the variables Variable M SD HIA 11 (PT) 9.01 4.00 HIA 12 (PT) 8.83 4.10 HIA 13 (PT) 8.64 4.15 HIA 14 (PT) 8.15 4.26 HIA 15 (PT) 7.60 4.38 LAC 11 (BPT) 2.13 0.66 LAC 12 (BPT) 2.23 0.71 LAC 13 (BPT) 2.24 0.69 LAC 14 (BPT) 2.29 0.72 LAC 15 (BPT) 2.34 0.73 DEP 11(PT) 2.60 2.36 DEP 12(PT) 2.55 2.26 DEP13(PT) 2.45 2.34 DEP 14(PT) 2.13 2.10 DEP 15(PT) 1.54 1.82 SES 11(P) 38.24 11.59 SES 12(P) 38.10 11.29 SES 13(P) 38.28 11.36 SES 14(P) 39.44 11.82 SES 15 (P) 39.25 11.77 DEL 11 (B) 6.29 27.88 DEL 12 (B) 7.43 39.90 DEL 13 (B) 17.91 83.60 DEL 14 (B) 33.06 154.61 DEL 15 (B) 37.50 167.68 Notes. HIA = hyperactivity-impulsivity-attention deficit problems; LAC = low academic achievement; DEP = depressive symptoms; SES = socioeconomic status; DEL = delinquency. 11–15 indicates ages 11–15. Informants are indicated in parentheses: P = Parent, T = Teacher; B = Boy. Statistical tests: HIA decreased with age (Wilks' Lambda = .91, F(4,424) = 10.89, p < .0001). LAC increased with age (Wilks' Lambda = .91, F(4, 418) = 10.91, p < .0001) DEP decreased with age (Wilks' Lambda = .80, F(4, 442) = 25.63, p < .0001) SES increased with age (Wilks' Lambda = .98, F(4, 410) = 2.55, p = .038) Ln(DEL) increased with age (Wilks' Lambda = .94, F(4, 418) = 6.18, p < .0001). Table options Because it was a count based on numbers of acts, delinquency was a highly skewed variable. In order to reduce the skewness, Ln(DEL) was calculated. At all ages, the skewness of this variable was between 1.7 and 2.0, and its kurtosis was between 2.2 and 4.0. These values allow normality to be assumed (Kline, 2005). The other variables were not skewed. All intercorrelations between variables were inspected. Only the correlations between hyperactivity and low academic achievement at the same age, and between hyperactivity and depression at the same age, exceeded .40. Only one correlation, between hyperactivity at age 12 and depression at age 12, exceeded .50, and this was .52. It was considered that these correlations were not sufficiently large to invalidate the analyses. Correlations between the same variable at different ages were taken account of in the models. Analytic strategy Structural equation analyses were conducted using Mplus 6.1 (Muthén & Muthén, 1998–2010). Several cross-lagged panel analyses were conducted in Mplus to investigate the interrelations between delinquency and hyperactivity, low academic achievement, depression, and SES. In order to do this a series of autoregressive cross-lagged models (Jöreskog, 1970) were fitted in Mplus. In each model, change in the variables is accounted for by regressing each repeatedly assessed variable on its immediate prior value (i.e. one-year stability paths). In addition, two-year stability paths were also estimated in the models. The stability paths signify continuity within variables. The cross-lagged, across-time paths represent associations between the repeated assessments. All forward and backward paths (e.g. from X to delinquency at all ages and from delinquency to X at all ages) were tested. The models also allowed for cross-sectional correlations between parallel assessed variables. To determine model fit, the Comparative Fit Index and Tucker Lewis Index (CFI and TLI; acceptable values > .90) (Bentler, 1990) and the root mean squared error of approximation (RMSEA; acceptable values < .08) (Browne & Cudeck, 1993) were used. As mentioned, the delinquency variable was log transformed to deal with its extreme non-normality. In addition, the non-normal distribution of the delinquency variable was also accounted for by using a MLR (Maximum Likelihood with Robust Standard Errors) estimator in Mplus. Boys with partially missing data were included in the model estimation by using a FIML (Full Information Maximum Likelihood) procedure (Enders & Bandalos, 2001). The first set of cross-lagged panel models was used to investigate whether hyperactivity, low achievement, depression and SES individually predicted juvenile delinquency. When a direct relationship was found between delinquency and a variable low achievement was consequently added to that model, to investigate whether the established direct relationship held after testing and controlling for effects of low achievement. Finally, based on the consistent significant results, an additional model was estimated in which the relationship between hyperactivity, low achievement, and SES was investigated. In all of the above-mentioned models, possible reverse effects were also taken into account (except for implausible reverse influences on SES), and the cross-lagged paths were constrained to be equal over time. Cross-lagged panel models The first model tested all paths linking hyperactivity and delinquency. The results were clear-cut (Fig. 1). Only paths from hyperactivity at age x to delinquency at age x + 1 were significant, not the reverse paths from delinquency to hyperactivity. Therefore, it might be concluded that hyperactivity is a possible cause of delinquency. Hyperactivity and delinquency path model. Fig. 1. Hyperactivity and delinquency path model. Figure options The second model tested all paths linking low achievement and delinquency. The results showed that paths from low achievement to delinquency were all more significant (with larger beta values) than paths from delinquency to low achievement (Fig. 2). The beta values for the latter paths were very small. Therefore, it might be concluded that low achievement is a possible cause of delinquency. Low achievement and delinquency path model. Fig. 2. Low achievement and delinquency path model. Figure options In light of these results, the third model tested all paths linking hyperactivity, low achievement, and delinquency. Fig. 3 shows that all paths from hyperactivity at age x to low achievement at age x + 1, and all paths from low achievement at age x to delinquency at age x + 1, were significant. No reverse paths were significant, and no paths from hyperactivity at age x to delinquency at age x + 1 were significant. Therefore, it might be concluded that hyperactivity is a possible cause of low achievement, that low achievement is a possible cause of delinquency, and that any influence of hyperactivity on delinquency is indirect and mediated by low achievement. Hyperactivity, low achievement, and delinquency path model. Fig. 3. Hyperactivity, low achievement, and delinquency path model. Figure options The fourth model tested all paths linking depression and delinquency. The results were clear-cut (Fig. 4). Only paths from delinquency at age x to depression at age x + 1 were significant, not the reverse paths from depression to delinquency. Therefore, it might be concluded that delinquency is a possible cause of depression. Depression and delinquency path model. Fig. 4. Depression and delinquency path model. Figure options In light of these results, the fifth model tested all paths linking depression, low achievement, and delinquency. Fig. 5 shows that all paths from low achievement at age x to delinquency at age x + 1, and all paths from delinquency at age x to depression at age x + 1, were significant. Two paths from delinquency to low achievement were significant, but the beta values were extremely small and much less than the reverse beta values. Therefore, it might be concluded that low achievement is a possible cause of delinquency, and that delinquency is a possible cause of depression. Depression, low achievement, and delinquency path model. Fig. 5. Depression, low achievement, and delinquency path model. Figure options The sixth model tested all paths linking SES and delinquency. The results were clear-cut (Fig. 6). Only paths from SES at age x to delinquency at age x + 1 were significant. The negative beta values indicate that SES is negatively related to delinquency, or conversely that low SES is a possible cause of delinquency. Socioeconomic status and delinquency path model. Fig. 6. Socioeconomic status and delinquency path model. Figure options In light of these results, the seventh model tested all paths linking SES, low achievement, and delinquency. Fig. 7 shows that all paths from SES at age x to low achievement at age x + 1 were significant (and negative), that no reverse paths were significant, and that no paths from SES at age x to delinquency at age x + 1 were significant. All paths from low achievement at age x to delinquency at age x + 1 were more significant (had higher beta values) than the reverse paths. Therefore, it might be concluded that low SES is a possible cause of low achievement, that low achievement is a possible cause of delinquency, and that any influence of low SES on delinquency is indirect and mediated by low achievement. Socioeconomic status, low achievement, and delinquency path model. Fig. 7. Socioeconomic status, low achievement, and delinquency path model. Figure options The eighth model tested all paths linking hyperactivity, low achievement, and SES. The results were clear-cut (Fig. 8). Only paths from hyperactivity and SES at age x to low achievement at age x + 1 were significant. There were no significant paths between SES and hyperactivity. Therefore, it might be concluded that hyperactivity and low SES are possibly independent causes of low achievement. Hyperactivity, low achievement, and socioeconomic status path model. Fig. 8. Hyperactivity, low achievement, and socioeconomic status path model.