طلاق، روش های یادگیری و پیشرفت تحصیلی کودکان: مطالعه طولی اثرات واسطه ای و تعدیلی
کد مقاله | سال انتشار | تعداد صفحات مقاله انگلیسی |
---|---|---|
37143 | 2014 | 13 صفحه PDF |
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Journal of School Psychology, Volume 52, Issue 3, June 2014, Pages 249–261
چکیده انگلیسی
Abstract Data from the Early Childhood Longitudinal Study — Kindergarten Cohort (ECLS-K) were used to test the hypothesis that approaches to learning (ATL) mediates the link between parental divorce and academic achievement. Fixed effects regression was utilized to test for mediation, and subsequent moderation analyses examining gender and age at time of divorce also were conducted. Results indicated that divorce was associated with less growth in test scores and that ATL mediated 18% and 12% of this association in reading and mathematics respectively. Parental divorce also was associated with larger negative effects for children who experienced divorce at an older age as well as for girls' mathematics test scores. These findings contribute to the understanding of the impact of parental divorce on children's academic achievement and underscore the importance of focusing on the variability of child outcomes following parental divorce.
مقدمه انگلیسی
Introduction Approximately half of all American youth will experience parental divorce at some point during their childhood (Lansford, 2009). In light of the prevalence of divorce and its association with negative outcomes, researchers have extensively studied the effects of divorce on children (Amato, 2001, Amato, 2010, Amato and Keith, 1991, Cherlin et al., 1991 and Emery, 1999). Amato (1993) described a general model of post-divorce outcomes that specified the effects of divorce depend on the configuration of stressors (e.g., conflict, parental remarriage, and moving) and resources (e.g., economic and parental support) present in a post-divorce situation. This model implies an understanding of divorce as a process rather than merely an event. According to the theory, as children experience stressors involved with their parents' unraveling marriage, they start to experience negative outcomes that persist and change following the divorce. Among other negative outcomes, parental divorce has been associated with lower academic achievement (Amato, 2001, Emery, 1999 and Jeynes, 2002). Few studies, however, have examined variables that potentially mediate or moderate the link between divorce and academic achievement. Some have speculated that the pathway between divorce and diminished academic achievement is primarily one of motivation rather than ability (Emery, 1999). Because divorce can result in a constellation of negative effects, including perceived guilt, blame, stressors, and diminished resources for children (Kurtz & Derevensky, 1993), it is plausible that divorce can negatively affect a child's motivation, engagement, and learning-related behavior in the classroom. This set of attitudes and behaviors have been characterized as approaches to learning (ATL; DiPerna et al., 2007 and Li-Grining et al., 2010) and could help explain the negative association between parental divorce and academic achievement. 1.1. Divorce and academic achievement Researchers have consistently shown divorce to be negatively related to academic achievement (Amato and Keith, 1991, Amato, 2001 and Jeynes, 2002). In a review of the empirical literature, Jeynes (2002) found only a small number of studies that did not show a significant negative association between divorce and academic achievement. Further, in a meta-analysis, Amato (2001) found that, after controlling for the quality of research methodology, children of divorce performed 0.17 standard deviations lower on measures of academic achievement than their peers from intact families. Similarly, in a large longitudinal study, Sun (2001) found that divorce had significant negative long-term effects on children's academic achievement. These studies support the hypothesis that parental divorce adversely affects children's academic achievement. With regard to moderators of the association between parental divorce and academic achievement, few studies have examined gender as a moderator of the impact of divorce on academic achievement. Sun and Li (2001) found no interaction effects between parental divorce and gender when predicting academic performance. This finding is consistent with several other studies as well (e.g., Allison and Furstenberg, 1989, Amato, 2001, Lansford et al., 2006 and Painter and Levine, 2000). In contrast, Neighbors, Forehand, and Armistead (1992) concluded that the negative effects of parental divorce on academic achievement were stronger for girls than boys. In sum, evidence is mixed as to whether gender moderates the impact of parental divorce on children's academic achievement. There also is mixed evidence regarding the moderating influence of age on the impact of parental divorce on academic achievement. Amato (2001) reported that divorce more negatively impacted the academic achievement of elementary students than high school students. Jeynes (2002) also concluded that most studies show a more negative effect for children when divorce occurs earlier in childhood. In contrast, Lansford (2009) posited that divorce may more negatively affect adolescents than elementary school students in the area of academic achievement due to the developmental salience of academic achievement during adolescence. Still other studies (e.g., Mednick, Backer, Reznick, & Hovecar, 1990) have reported that age does not significantly moderate the effect of divorce on academic achievement. In sum, evidence is inconclusive as to whether or how age moderates the negative association between parental divorce and academic achievement. In addition to studying moderating variables, recent research also has examined potential mediators between divorce and negative consequences of divorce. For example, Potter (2010) studied the association between divorce, academic achievement, and psychosocial well-being. Specifically, Potter hypothesized that psychosocial well-being acts as a mediator between divorce and academic achievement. Results indicated that the academic differences between children from divorced and intact families widened as time progressed, indicating that divorce exacerbated differences that were present prior to legal recognition of the divorce. Potter also found that psychosocial well-being following divorce only partially mediated the association between parental divorce and academic achievement. As such, there is need for further research on mediators of the association between parental divorce and academic achievement to more completely explain this association. 1.2. Approaches to learning Several authors have identified key student variables that are related to academic achievement. Such variables could potentially mediate the associations between divorce and achievement. DiPerna (2006) and DiPerna et al., 2002 and DiPerna et al., 2005 have proposed and tested a model of academic enablers hypothesized to contribute to academic achievement. DiPerna and Elliott (2002) defined academic enablers as “attitudes and behaviors that allow a student to participate in, and ultimately benefit from academic instruction in the classroom” (p. 294). Given the observed association between parental divorce and diminished academic achievement (e.g., Amato, 2001 and Jeynes, 2002), models such as those by DiPerna and colleagues, may provide insight regarding potential mediators of this association. As such, the DiPerna (2006) model was used as a guiding framework of the current study. Specifically, the DiPerna (2006) model emphasizes the importance of students' motivation, engagement, and study skills. These academically-related attitudes and behaviors are often referred to as ATL. Li-Grining et al. (2010) defined ATL as individual characteristics and observable behaviors related to the learning process. Specifically, it includes motivation (Duncan et al., 2007) and engagement behaviors such as staying on task, persisting in difficult activities, and asking for assistance (DiPerna et al., 2007). Academic achievement has been shown to be related to both motivation (Bandura, 1997, Gottfried, 1990, Mitchell, 1992 and Peterson, 1990) and academic engagement (Finn, 1993, Greenwood et al., 2002 and Singh et al., 2002), and both of these variables are important parts of the DiPerna (2006) model of academic enablers. Given that motivation and engagement behaviors are reflected within ATL, it is not surprising that ATL has been shown to be related to academic achievement in both mathematics (DiPerna et al., 2007) and reading (Li-Grining et al., 2010). 1.3. Divorce and approaches to learning Although there are no previous studies examining the association between ATL and parental divorce, ATL has been linked with family involvement (Fantuzzo, McWayne, Perry, & Childs, 2004). In addition, there are a small number of studies examining the association between parental divorce and constructs (motivation and engagement) reflected by ATL. Specifically, researchers have documented divorce's effects on several motivational sub-constructs as well as the overarching construct of motivation (Bertram, 2006, Ivanova and Israel, 2005 and Kurtz and Derevensky, 1993). Further, some have speculated that the pathway between divorce and diminished academic achievement is primarily one of motivation rather than ability (Emery, 1999). If divorce negatively impacts children's motivation, it is plausible that such impairment negatively affects children's academic engagement as well (DiPerna et al., 2002). Especially given that student engagement has been shown to be changeable and linked with academic achievement (Appleton et al., 2008 and Fredricks et al., 2004), it is possible that divorce's impact on children affects their classroom experience, which negatively affects their academic motivation, engagement, and ultimately, their academic achievement. However, there is little research examining the association between parental divorce and engagement. Teachman (2008) found that living situations other than a stable biological family have a negative effect on student engagement, even after controlling for turbulence, parenting context (e.g., parental religious beliefs, volunteering in the community, and mental health), and income. In addition, Abd-El-Fattah (2006) found that non-intact family structures were negatively related to parental involvement, which in turn was related to student disengagement. These findings suggest that an association between parental divorce and ATL could potentially help explain the negative association between parental divorce and academic achievement. The small number of studies in this area, however, highlights the need for further research on the connection between divorce and motivation, especially given its potential as a mediator between parental divorce and academic achievement. There are also several theoretical reasons that parental divorce could affect children's ATL. First, increased mobility after divorce (Booth & Amato, 1992) could mean that children of divorce have to cope with a new classroom, teacher, and set of friends in addition to new living situations and decreased access to one parent. In addition, the impact of parental divorce on children's psychological adjustment, social relations, and self-concept (Amato, 2001) could impact his or her experience of the classroom. Finally, effects of parental divorce on children's behavior could lead to disruptions in the classroom that both distract from learning and strain relationships with teachers and other school professionals. In sum, the constellation of consequences resulting from divorce could negatively influence a child's ability or willingness to participate in, and benefit from classroom instruction, lowering his or her ATL and ultimately academic achievement. 1.4. Rationale and purpose Given the association between divorce and diminished academic achievement (Amato, 2001, Amato and Keith, 1991 and Emery, 1999), the paucity of studies examining mediators of this association, and the potential for ATL to mediate the association (Bertram, 2006, Emery, 1999 and Teachman, 2008), there is a need to examine whether ATL can help explain divorce's effects on academic achievement. Further, there is a need to examine potential moderators of the negative association between parental divorce and academic achievement (Amato, 2010). As such, the purpose of the current study was to test the hypothesis that ATL partially mediates the association between divorce and children's academic achievement. A secondary purpose of this study was to examine potential moderating effects of a child's age at the time he or she experiences parental divorce (i.e., age at time of divorce) and gender on the association between parental divorce and academic achievement.
نتیجه گیری انگلیسی
Results 3.1. Tests of assumptions Descriptive statistics for key variables are displayed in Table 2. Because fixed effects regression with more than two time points involves mean deviation scores rather than raw scores, the assumptions of regular ordinary least squares regression were tested on distributions of mean deviation scores. First, P–P and Q–Q plots were examined to assess the assumption of normality. Slight deviations from normality were detected in the distributions of change scores for both reading and mathematics test scores; however, the high number of cases assured that these deviations from normality did not affect the results (Field, 2009). Upon examination of scatter and residual plots, the assumptions of linearity and homoscedasticity also held, although some deviant patterns emerged. These deviations further supported the use of robust clustered standard errors, which also account for clustering in the data. The assumption of normality of residual also held upon examination of P–P and Q–Q plots, and excessive multicollinearity was not present. All analyses were conducted in STATA 12. Table 2. Means, and standard deviations for key variables by wave (grade) and divorce status. Fall kindergartena Spring kindergarten Grade 1 Grade 3 Grade 5 Variable No divorce (n = 9,794) Divorce (n = 164) No divorce (n = 9,202) Divorce (n = 304) No divorce (n = 7,384) Divorce (n = 461) No divorce (n = 5,712) Divorce (n = 575) No divorce (n = 5,316) Reading Score 37.29 44.99 49.22 78.67 82.77 128.25 135.03 150.36 156.68 (11.22) (9.94) (15.21) (22.93) (23.88) (27.41) (25.87) (25.51) (24.22) Math Score 28.46 36.23 39.54 63.15 65.72 100.83 105.73 124.52 129.94 (9.57) (10.23) (12.38) (17.76) (18.09) (23.11) (23.18) (22.99) (22.63) Approaches to Learning 3.10 3.00 3.24 2.95 3.16 2.97 3.19 3.01 3.20 (0.64) (0.70) (0.64) (0.71) (0.66) (0.68) (0.63) (0.69) (0.63) Age at Assessment 68.55 74.98 74.74 87.19 86.91 111.44 110.98 134.93 134.74 (4.27) (4.25) (4.38) (4.22) (4.22) (4.19) (4.18) (5.32) (5.32) Note. Age reported in months. Standard deviations reported under means in parentheses. a Divorce cases not available in this wave due to exclusion criteria. Table options A Hausman (1978) test was performed to determine whether fixed effects regression was preferable to random effects regression. The Hausman test compares the coefficients calculated in fixed and random effects models. Because this test could not be computed with the data adjusted for clustering, it was computed on the data unadjusted for clustering. Results indicated that coefficients for each predictor (divorce, age at time of assessment, and ATL) were significantly different in random and fixed effects models for both reading and mathematics. This result indicates that if a random effects model was used, bias may be introduced due to the restrictions of a random effects model (Allison, 2009) leading to overestimation of the effects of parental divorce. Thus, the Hausman test further supported the use of a fixed effects approach. 3.2. Models of direct effects All tests were conducted at the p < .05 level, and all reported regression coefficients are unstandardized. Also, statistically significant unstandardized regression coefficients were divided by the pooled standard deviation across waves (19.93 for reading test scores, 16.97 for mathematics test scores, and 0.64 for ATL) to provide some indication of effect size (ES). All unstandardized regression coefficients are presented in Table 3, displayed sequentially according to the order in which the analyses were conducted. First, the direct effects of divorce and ATL on reading and mathematics test scores were tested. Controlling for age at time of assessment, divorce was not a statistically significant predictor of reading test scores, b = − 1.14, p = .06, but was a statistically significant predictor of mathematics test score, b = − 0.90, p = .003, ES = − 0.05 and ATL, b = − 0.11, p > .001, ES = − 0.18 ( Table 3, Direct Effect — Divorce). Furthermore, controlling for age at time of assessment, ATL was a statistically significant predictor of test scores for both reading, b = 1.84, p < .001, ES = 0.09 and for mathematics, b = 0.98, p < .001, ES = 0.06 ( Table 3, Direct Effect — ATL). Table 3. Fixed effects regression coefficients of models for reading test scores, mathematics test scores, and approaches to learning scores. Parameter Direct effects Mediation Moderation Divorce ATL Divorce × Gender Divorce × Age at divorce b SE b SE b SE b SE b SE Reading test scores Age at Assessment 1.91⁎⁎⁎ 0.003 1.91⁎⁎⁎ 0.004 1.91⁎⁎⁎ 0.003 1.91⁎⁎⁎ 0.003 1.91⁎⁎⁎ 0.003 Divorce − 1.14 0.60 − 0.94 0.62 − 1.85 1.02 34.84⁎⁎⁎ 4.39 ATL 1.84⁎⁎⁎ 0.29 1.82⁎⁎⁎ 0.30 Divorce × Gender 1.45 1.29 Divorce × Age at divorce − 0.38⁎⁎⁎ 0.04 Mathematics test scores Age at assessment 1.57⁎⁎⁎ 0.004 1.57⁎⁎⁎ 0.004 1.57⁎⁎⁎ 0.004 1.57⁎⁎⁎ 0.004 1.58⁎⁎⁎ 0.004 Divorce − 0.90⁎⁎ 0.30 − 0.79⁎⁎ 0.30 1.41 1.19 27.21⁎⁎⁎ 3.07 ATL 0.98⁎⁎⁎ 0.22 0.97⁎⁎⁎ 0.22 Divorce × Gender − 4.71⁎ 2.22 Divorce × Age at divorce − 0.30⁎⁎⁎ 0.03 Approaches to learning scores Age at Assessment 0.00005 0.0002 0.00005 0.0002 0.00005 0.0002 Divorce − 0.11⁎⁎⁎ 0.02 − 0.14⁎⁎⁎ 0.02 − 0.05 0.07 Divorce × Gender 0.06⁎⁎ 0.02 Divorce × Age at divorce − 0.0007 0.0007 Note. N = 9,794. ATL = Approaches to learning. ⁎ p < .05. ⁎⁎ p < .01. ⁎⁎⁎ p < .001. Table options 3.3. Mediation results In order to determine whether statistically significant mediation was present, Sobel's test was conducted on each model. The results of the Sobel's tests indicated that ATL was a statistically significant mediator of the effect of parental divorce on academic achievement for both reading test scores, a ∗ b = − 0.21, z = − 4.42, p < .001, and mathematics test scores, a ∗ b = − 0.11, z = − 3.65, p < .001. Asymmetrical confidence intervals did not include 0 for both reading, 95% CI [− 0.31, − 0.12], and mathematics, 95% CI [− 0.17, − 0.06], indicating the mediation was statistically significant. To determine the proportion of the mediated effect to the total effect according to procedures outlined by Fairchild and McQuillin (2010), a model of the effects of divorce on reading test scores and mathematics test scores after controlling for ATL was tested ( Table 3, Mediation). Controlling for ATL, the unstandardized regression coefficient for divorce decreased in magnitude for both reading and mathematics test scores. This divorce coefficient was not statistically significant for reading test scores, b = − 0.94, p = .13, but was statistically significant for mathematics test scores, b = − 0.79, p = .009, ES = − 0.05. Based on the decrease in divorce coefficient magnitude when controlling for ATL, ATL mediated 18% of the effect of divorce on reading test scores and 12% of the effect of divorce on mathematics test scores. 3.4. Moderation and follow-up Subsequent analyses were conducted to test (a) whether gender moderated the effect of parental divorce on academic achievement and if such an effect was mediated by ATL, or (b) whether mediation of ATL between parental divorce and academic achievement was moderated by gender. First, there was no significant interaction between parental divorce and gender predicting reading test scores, b = 1.45, p = .27 ( Table 3, Moderation — Divorce × Gender). This model was subsequently tested for moderated mediation. The interaction between parental divorce and gender predicting ATL was statistically significant, b = 0.06, p = .004 indicating that girls who experienced parental divorce gained 0.06 more points in ATL than their all other groups. Testing regression models for girls and boys separately proved that this difference was negligible as divorce was a statistically significant predictor of ATL for both boys, b = − .12, p < .001, ES = − 0.18, and girls, b = − .11, p < .001, ES = − 0.17. The interaction between ATL and gender predicting reading test scores controlling for age at time of assessment was also not statistically significant, b = 0.07, p = .80. Thus, although there was evidence that the association between divorce and ATL was moderated by gender, the difference between this association for girls and boys was negligible. Ultimately, although there was evidence of gender moderation of the mediation of the effect of divorce on reading achievement through ATL, the magnitude of this moderation was not large. In the mathematics model, there was a statistically significant interaction between parental divorce and gender, b = − 4.71, p = .037 ( Table 3, Moderation — Divorce × Gender), indicating that girls who experienced parental divorce gained 4.71 fewer points than all other groups (girls who did not experience divorce, boys who experienced divorce, and boys who did not experience divorce; Fig. 1). The model was subsequently tested for mediated moderation. The magnitude of the unstandardized regression coefficient for the interaction between parental divorce and gender predicting achievement did not decrease as a result of the inclusion of ATL as a potential mediator, b = − 4.77, p = .034. Furthermore, the interaction between gender and ATL predicting mathematics achievement was not statistically significant, b = − 0.30, p = .18. Therefore, the results do not support the presence of significant mediated moderation. Graph of the interaction between parental divorce and gender predicting academic ... Fig. 1. Graph of the interaction between parental divorce and gender predicting academic achievement in mathematics. Figure options A second set of analyses was conducted to test whether children's age at time of divorce moderated the impact of parental divorce on academic achievement (Table 3, Moderation — Divorce × Age at divorce) and whether any moderated mediation or mediated moderation was present. First, there was a significant interaction between divorce and age at time of divorce predicting reading test scores, b = − 0.38, p < .001, and mathematics test scores, b = − 0.30, p < .001, indicating that as children's age at time of divorce increased by 1 month, those who experienced divorce gained 0.38 fewer points on the ECLS-K reading test and 0.30 fewer points on the ECLS-K mathematics test. These interactions are evident in Fig. 2. There was not a statistically significant interaction between divorce and age at time of divorce predicting ATL, b = − 0.0007, p = .34 ( Table 3, Moderation — Divorce × Age at divorce, Table 3). This model was subsequently tested for mediated moderation. The magnitude of the unstandardized regression coefficient for the interaction between parental divorce and age at time of divorce did not change significantly as a result of the inclusion of ATL as a potential mediator in the reading model, b = − 0.38, p < .001, and the mathematics model, b = − 0.30, p < .001. Thus the pattern of findings did not support the presence of mediated moderation. Graph of the interaction between parental divorce and age at time of divorce ... Fig. 2. Graph of the interaction between parental divorce and age at time of divorce predicting academic achievement in reading and mathematics.