مسیرهای بزهکاری اولیه: بررسی کمک های فردی و جمعی خلق و خوی دشوار، دخالت کم مادران و رفتار برون سازی

Abstract Background The purpose of this study was to investigate the impact of difficult temperament and maternal involvement, measured at ages one and three years, respectively, on externalizing behavior at age five and early delinquency at age nine. Methods Maternal- and child-reports from 4,897 members of the Fragile Families and Child Wellbeing Study (FFCWS) were included in a path analysis of four of five waves of FFCWS data: Wave 2 (difficult temperament at age 1), Wave 3 (maternal involvement at age 3), Wave 4 (externalizing behavior at age 5), and Wave 5 (delinquency at age 9). Findings Although difficult temperament at age 1 displayed a weak zero-order correlation with delinquency at age 9 and low maternal involvement at age 3 failed to correlate with delinquency at age 9, both entered into significant chained relationships with delinquency via externalizing behavior at age 5. In addition, difficult temperament at age 1 seemed to evoke low parental involvement at age 3. Conclusions The respective roles of a difficult temperament, maternal involvement, and externalizing behavior in a proximal chaining process may be partially responsible for the continuity that has been observed in antisocial behavior over time.

Introduction It is a well-known empirical fact that temperament (Lahey et al., 2008, Leve et al., 2005 and Nigg, 2006) and low parental support (Brophy and Dunn, 2002, Côté et al., 2006 and Smith et al., 2004) are associated with later externalizing behavior. There is also a well-documented relationship between early externalizing behavior and subsequent delinquency (Campbell et al., 2006, Loeber and Burke, 2011 and Timmermans et al., 2009). As interesting as these findings are, they still leave three crucial questions unanswered. First, although temperament and parenting are both linked to externalizing behavior, are they meaningfully connected to one another? Second, is the temperament/parenting-externalizing behavior relationship the result of mediation, moderation or both? Third, is there a causal connection between temperament, early parenting and subsequent delinquency that runs through externalizing behavior? In an effort to answer these and other questions the current study examined individual and collective or chained relationships among difficult temperament, maternal involvement, externalizing behavior, and early delinquency. A series of studies have identified the presence of bidirectional relationships between temperament and several less than optimal parenting styles. Eisenberg et al. (1999), for instance, determined that negative emotionality in a 6 to 8 year old child predicted parental distress two years later and that parental distress predicted further negative emotionality in the child two years after this. Correlations have also been observed between a fearful and irritable temperament and inadequate parenting (Lengua & Kovacs, 2005) and between a difficult childhood temperament and early negative parenting (Kiff, Lengua, & Zalewski, 2011). Most recently, Lee, Zhou, Eisenberg, and Wang (2012) observed a bidirectional relationship between authoritarian parenting and a difficult early temperament marked by low effortful control and high anger/frustration. Because temperament data were only available for one wave of the current study, the focus of the present investigation was on the unidirectional relationship between early temperament, subsequent parental involvement, and later externalizing and delinquent behavior. Biological sex has the potential to moderate the effect of temperament and parenting on externalizing and delinquent behavior, although this area of research has produced mixed results. Gender intensification theory holds that girls are more affected by disruptions in family relations than boys (Davies & Windle, 1997) and while some studies note that parenting factors correlate better with externalizing and delinquent behavior in girls than in boys (Leve et al., 2005 and Walters, 2013), other studies show just the opposite (McFadyen-Ketchum et al., 1996 and Miner and Clarke-Stewart, 2008) or indicate that sex does not moderate parenting at all (De Haan et al., 2012 and Eisenberg et al., 2005). Much the same situation is encountered when studies on the interactive effect of sex on temperament are reviewed (Miner and Clarke-Stewart, 2008 and Shaw et al., 1994). The current study consequently sought to ascertain whether sex interacts with either temperament or parental involvement in the development of later externalizing and delinquent behavior. Smeekens, Riksen-Walraven, and van Bakel (2007) conducted a study in which they determined that angry temperament moderated the relationship between parenting and externalizing behavior and that later parenting mediated the relationship between early parenting and subsequent externalizing behavior. In a study published that same year, Simons, Simons, Chen, Brody, and Lin (2007) discerned that cognitive factors mediated the relationship between early parenting and later delinquency. Findings from these and other studies raise the possibility that mediation may be as important as moderation in explaining the chain of events running from early temperament and parenting to later externalizing and delinquent behavior. Walters (2014), for instance, ascertained that youth’s perception of parental attitudes toward deviance and youth attitudes toward deviance mediated the relationship between actual parental attitudes toward deviance and youth delinquency even though the zero-order correlation between actual parental attitudes toward deviance and youth deviance was non-significant. This suggests that a proximal chaining process may be partially responsible for the conflicting results obtained in research on the developmental precursors of antisocial behavior. Conceptualizing the variables as sequential links in a chain, it could be argued that as the distance between links increases correlations between variables decrease though they remain linked by the intervening effect of other variables in the chain. Extending the proximal chaining hypothesis to the relatively weak relationships observed when temperament and early parenting are correlated with later delinquency (see van der Voort, Linting, Juffer, Bakersman-Kranenburg, & van IJzendoorn, 2013), we can see that temperament and early parenting, by virtue of their position on the chain, could potentially affect delinquency even though their zero order correlations with delinquency are weak. Proper evaluation of this hypothesis, however, requires that certain variables be controlled. This should include a range of pre-existing factors that could potentially shape child temperament and parenting behavior; factors such as low birth weight, family structure, family income, and parental race. Hence, a model capable of simultaneously testing two chains, one running from difficult temperament to maternal involvement to externalizing behavior to early delinquency and the other running from difficult temperament to externalizing behavior to early delinquency, and of controlling for potentially important pre-existing variables was employed in this study as a means of testing the chaining hypothesis. DeLisi and Vaughn (2014) offer a temperament-based theory of antisocial behavior that may provide the conceptual framework that is required to make sense of the temperament-offending relationship. According to this theory, low effortful control and high negative emotionality are temperamental traits that contribute to self-regulatory difficulties in childhood and adolescence, all of which then lead to an evolving pattern of adult criminality. An early age of onset of offending behavior is said to be a direct consequence of these early self-regulatory problems (DeLisi, Neppl, Lohman, Vaughn, & Shook, 2013). In addition, difficult temperament can evoke certain environmental reactions which, in turn, contribute to future self-regulatory and behavioral problems (Barrett & Fleming, 2011). The weak relationship commonly observed between temperament and later criminality (van der Voort et al., 2013) would seem to present a problem for the DeLisi and Vaughn (2014) model. That is, until one considers the role of chaining in the overall process and the ability of proximal events like temperament and early maternal involvement to link to more distal outcomes like delinquency by way of the chaining process. The current study The purpose of this study was three-fold: 1. ascertain whether difficult temperament is capable of evoking low maternal involvement even after controlling for pre-existing levels of maternal involvement; 2. check for moderation effects by evaluating interactions between sex and difficult temperament/maternal involvement; 3. determine whether variables like difficult temperament and maternal involvement are linked to early delinquency by a series of mediating variables. The first hypothesis stated that difficult temperament at age 1 would predict maternal involvement at age 3, after controlling for maternal involvement at age 1. The second hypothesis held that sex would moderate difficult temperament and maternal involvement in their relationships with each other, externalizing behavior, and early delinquency. The third hypothesis predicted that difficult temperament and maternal involvement would be linked to early delinquency via externalizing behavior and that the difficult temperament → maternal involvement → externalizing behavior → delinquency and difficult temperament → externalizing behavior → delinquency chains would both be significant.

Results Descriptive statistics and evocation analysis Descriptive statistics and correlations for the 10 variables included in this study are listed in Table 1. A review of the final column of the table reveals that difficult temperament and externalizing behavior both achieved significant zero-order correlations with early delinquency. Zero-order correlations between difficult temperament and maternal involvement showed that difficult temperament at Wave 2 predicted maternal involvement at Wave 3 (r = .10, p < .001) in support of the evocation hypothesis. Wave 2 difficult temperament continued to predict Wave 3 maternal involvement even after Wave 2 maternal involvement was controlled (rp = .04, p < .05). Table 1. Descriptive Statistics for and Correlations between the Independent, Dependent, Mediating, and Control Variables from this Study Variable n Wave M SD Range B C D E F G H I J A. Sex 4897 W1 1.48 0.50 1–2 -.01 .04* .00 .00 .00 -.01 -.02 -.08** -.20** B. Parents Married 4882 W1 0.25 0.43 0–1 -.08** .26** .30** .44** -.11** -.04* -.10** -.09** C. Low Birth Weight 4759 W1 0.10 0.30 0–1 -.06** -.06** -.05** .04* .01 .03 .01 D. Mother White 4807 W1 0.31 0.46 0–1 .82** .26** -.09** -.13** .02 -.10** E. Father White 4773 W1 0.28 0.45 0–1 .29** -.10** -.14** -.00 -.12** F. Family Income 4897 W1 31.99 31.56 0–133.8 -.12** -.09** -.09** -.12** G. Temperament W2 4333 W2 8.47 3.20 3–15 .10** .23** .04* H. Involvement W3 4167 W3 50.58 9.84 0–63 .15** .01 I. Externalizing 3521 W4 3.20 2.56 0–12 .16** J. Delinquency 3344 W5 1.25 1.78 0–17 Note. n = number of non-missing values; Wave = wave when variable measured; M = mean; SD = standard deviation; Range = range of scores in current sample; Temperament = difficult temperament at age 1 year (Wave 2); Sex = biological sex (male = 1, female = 2); Parents Married = child’s parents married at the time of child’s birth (yes = 1, no = 0); Low Birth Weight = child weighed less than 2,500 grams at birth (yes =1, no = 0); Mother White = racial status of mother (white = 1, nonwhite = 0); Father White = racial status of father (white = 1, nonwhite = 0); Family Income = estimated yearly family income in thousands of dollars; Internalizing = internalizing behavior at age 5 years (Wave 4); Involvement = maternal involvement at age 3 years (Wave 3); Externalizing = externalizing behavior at age 5 years (Wave 4); Delinquency = early delinquent behavior at age 9 years (Wave 5). *p < .05; **p < .001. Table options Path analysis The path analysis results are summarized in Table 2 and the path diagram is reproduced in Fig. 1. Regarding the second hypothesis, there was no evidence of sex moderation in any of the three regressions included in the path analysis. With respect to the third hypothesis, the Temperament-2 → Involvement-3 → Externalizing → Delinquency chain satisfied the first two criteria for chaining. This chain achieved a significant indirect effect (Criterion 1) and the direct effect of Wave 2 temperament on Wave 5 delinquency was non-significant (Criterion 2). With respect to the third criterion, however, the four-stage mediator chain had a lower AVG-β than the Externalizing → Delinquency chain (.121 vs. .146). The Temperament-2 → Externalizing → Delinquency chain, on the other hand, satisfied all three criteria for chaining. The Temperament-2 → Externalizing → Delinquency chain achieved a significant indirect effect, the direct effect of Wave 2 temperament on delinquency was non-significant, and the two-leg mediator chain had a higher AVG-β than the Externalizing → Delinquency chain (.180 vs .146). Table 2. Results of the Path Analysis with Total, Direct, and Indirect Effects Predictor b(95% CI) β t p Involvement-3 on Temperament-2 0.249(0.154, 0.346) 0.081 5.07 .000 Sex − 0.333(− 0.919, 0.248) − 0.017 − 1.12 .264 Temper-2 x Sex − 0.084(− 0.396, 0.228) − 0.009 − 0.52 .600 Parents Married 0.876(0.099, 1.642) 0.038 2.24 .025 Low Birth Weight 0.020(− 1.035, 1.074) 0.001 0.04 .970 Mother White − 1.269(− 2.342, − 0.149) − 0.059 − 2.26 .024 Father White − 1.566(− 2.711, − 0.398) − 0.071 − 2.68 .007 Family Income 0.000(0.000, 0.000) − 0.062 − 3.67 .000 Externalizing on Involvement-3 0.036(0.027, 0.045) 0.137 7.82 .000 Temperament-2 0.171(0.143, 0.199) 0.213 12.10 .000 Sex − 0.384(− 0.548, − 0.226) − 0.075 − 4.62 .000 Temper-2 x Sex − 0.029(− 0.118, 0.060) − 0.011 − 0.65 .513 Involve-3 x Sex 0.057(− 0.030, 0.142) 0.022 1.28 .201 Parents Married − 0.461(− 0.668, − 0.259) − 0.077 − 4.44 .000 Low Birth Weight 0.139(− 0.138, 0.416) 0.016 0.98 .328 Mother White 0.606(0.265, 0.955) 0.109 3.45 .001 Father White − 0.077(− 0.429, 0.279) − 0.013 − 0.43 .670 Family Income 0.000(0.000, 0.000) − 0.041 − 2.38 .017 Delinquency on Externalizing 0.101 (0.074, 0.131) 0.146 7.09 .000 Involvement-3 − 0.006(− 0.012, 0.000) − 0.036 − 2.14 .033 Temperament-2 − 0.007(− 0.027, 0.014) − 0.012 − 0.64 .522 Sex − 0.691(− 0.812, − 0.580) − 0.194 − 11.60 .000 Temper-2 x Sex 0.013(− 0.049, 0.074) 0.007 0.41 .684 Involve-3 x Sex − 0.015(− 0.075, 0.042) − 0.009 − 0.52 .603 Parents Married − 0.090(− 0.242, 0.065) − 0.022 − 1.15 .250 Low Birth Weight 0.041(− 0.153, 0.241) 0.007 0.41 .683 Mother White − 0.040(− 0.297, 0.241) − 0.010 − 0.29 .770 Father White − 0.362(− 0.635,-0.114) − 0.091 − 2.71 .007 Family Income 0.000(0.000, 0.000) − 0.070 − 4.05 .000 Temp-2 → Involve-3 → Externalizing → Delinquency Total Effect 0.010(− 0.010, 0.030) 0.018 0.99 .324 Direct Effect − 0.007(− 0.027, 0.014) − 0.012 − 0.64 .522 Indirect Effect 0.0174(0.011, 0.023) 0.030 5.73 .000 Involve-3 − 0.002(− 0.003, 0.000) − 0.003 − 1.95 .051 Externalizing 0.017(0.012, 0.024) 0.031 6.12 .000 Involve-3 → External 0.001(0.001, 0.001) 0.002 3.68 .000 Note. Involvement-3 on = regression equation with Wave 3 maternal involvement as the outcome; Externalizing on = regression equation with Wave 4 externalizing behavior as the outcome; Delinquency on = regression equation with Wave 5 early delinquency as the outcome; Temperament = difficult temperament; Involvement = maternal involvement; Sex = biological sex (male = 1, female = 2); Temper-2 x Sex = interaction between Wave 2 difficult temperament and sex; Involve-3 x Sex = interaction between Wave 3 maternal involvement and sex; Parents Married = child’s parents married at the time of child’s birth (yes = 1, no = 0); Low Birth Weight = child weighed less than 2,500 grams at birth (yes = 1, no = 0); Mother White = racial status of mother (white = 1, nonwhite = 0); Father White = racial status of father (white = 1, nonwhite = 0); Family Income = estimated yearly family income at Wave 1; b(95% CI) = unstandardized coefficient and the lower and upper limits of the 95% confidence interval for the unstandardized coefficient (in parentheses); β = standardized coefficient; t = asymptotic t-test; p = significance level of the asymptotic t-test; N = 4897. Table options Path diagram of temperament, maternal involvement, externalizing behavior, and ... Fig. 1. Path diagram of temperament, maternal involvement, externalizing behavior, and delinquency relationships over four waves of the FFCWS. Note. Standardized beta coefficients are reported; effects of Wave 1 control variables (child sex, parents married, low birth weight, mother white, father white, and family income) and interactions between child sex and temperament and between child sex and maternal involvement are not included in the diagram; N = 4897. *p < .05, **p < .001. Figure options The sensitivity of each of the significant mediating effects observed in this study to unobserved extraneous variables was assessed using Kenny’s “failsafe ef” procedure. Sensitivity testing of the Temperament-2 → Involvement-3 → Externalizing → Delinquency chain disclosed that an unobserved covariate confounder would need to correlate .35 with both the mediator (Involve-3) and outcome (Externalizing) of the first leg of the chain or .41 with the mediator (Externalizing) and outcome (Delinquency) of the second leg of the chain to reduce the respective mediating effects to zero. Sensitivity testing of the Temperament-2 → Externalizing → Delinquency chain revealed that an unobserved covariate confounder would need to correlate .39 with both the mediator (Externalizing) and outcome (Delinquency) to completely eliminate the mediating effect of externalizing behavior on the temperament–delinquency relationship. Given that a correlation of .30 represents a medium effect size and a correlation of .50 represents a large effect size (Cohen, 1988), these results suggest that the mediating effects obtained in the current study were moderately robust to the effects of extraneous variables.