خودکارآمدی و پیشرفت تحصیلی در دانش آموزان دبیرستانی استرالیا: اثرات واسطه ای آرزوهای دانشگاهی و بزهکاری

Abstract Studies have shown that self-efficacy, aspirational, and other psychosocial influences account for considerable variance in academic achievement through a range of mediational pathways, although no research to date has tested the mediational relationships identified. The present research investigated the structural relations among self-efficacy, academic aspirations, and delinquency, on the academic achievement of 935 students aged 11–18 years from ten schools in two Australian cities. The Children's Self-Efficacy Scale, Adapted Self-Report Delinquency Scale (Revised), and Children's Academic Aspirations Scale were administered to participants prior to academic achievement being assessed using mid-year school grades. Structural equation modeling was employed to test three alternative models for the relationships from academic, social, and self-regulatory efficacy on academic achievement. A partial mediation model showed the best overall fit to the data. Academic and self-regulatory efficacy had an indirect negative effect through delinquency and a direct positive effect on academic achievement. Academic and social self-efficacy had positive and negative relationships, respectively, with academic aspiration and academic achievement; however, the relationship between academic aspiration and academic achievement was not significant in the final model.

Results Confirmatory factor analyses The first analysis of the measurement model assessed the second-order structure of the measures. To begin, assessment of the delinquency and self-efficacy measures was undertaken. Three of the subscales were not clearly separate from the remaining subscales – school misdemeanors, vehicle-related offences, and soft drug use. Items from these subscales did not factor cleanly as they represented “softer” misdemeanors that perhaps occurred alongside the “harder” delinquency behaviors. These three subscales were omitted from future analyses such that delinquency was operationalized by hard drug use, property offences, physical aggression, and stealing offences. The initial analysis of the self-efficacy measures also produced conceptual overlaps amongst the scales. Although the model fitted the data adequately (c2 = 2641.35; RMSEA = .08; CFI = .86; NFI = .84), examination of the findings revealed cross-loadings. It was decided, therefore, to identify the most optimal items to form a subscale for the self-efficacy measures. Three subscales were identified from the academic self-efficacy scale: Learning (4 items; e.g., “Learn social studies”, “Learn English”); Process (6 items; e.g., “Organise school-work”; “Take class notes”); and Academic Expectations (2 items; “Live up to teacher expectations” and “Live up to parent expectations”). Four subscales were identified from the social self-efficacy scale: Friends (3 items: e.g., “Make friends with other sex”; “Make friends with same sex”); Social Expectations (2 items; “Live up to peer expectations” and “Live up to own expectations”); Dealing with Social Issues (2 items; “Stand up for self” and “Dealing with annoying things”); and Sport (2 items; “Learn sport skills” and “Learn team sports”). This model produced a significantly better fit to the data (c2 = 1558.97; RMSEA = .05; CFI = .93; NFI = .93) and so was used in the subsequent analyses. The tested measurement model assessed the fit of the latent scale scores (delinquency, academic self-efficacy, self-regulatory self-efficacy, social self-efficacy, and academic aspirations) on their respective latent subscales which loaded on to the individual items. This model had an acceptable fit to the data (c2 = 5160.87, df = 1066, p < .001; RMSEA = .06; CFI = .96; NFI = .95). The loadings of the items, and the second-order factor loadings of the subscales on to the scales are presented in Fig. 2. Bivariate correlations between constructs are presented in Table 1. Factor loadings for the items on the first- and latent-variables on the ... Fig. 2. Factor loadings for the items on the first- and latent-variables on the second-order factors. Figure options Table 1. Correlations of study variables. Mean Standard deviation Age Sex School SES Academic self-efficacy Social self-efficacy Self-regulatory self-efficacy Academic aspirations Delinquency English Achievement AGE 14.40 1.43 SEX .01 School SES .04 .03 Academic self-efficacy 4.08 .95 −.19a .01 .06 Social self-efficacy 4.45 .90 −.05 .03 .03 .54a Self-regulatory self-efficacy 4.77 1.15 −.05 .04 −.01 .50a .34a Academic aspirations 4.67 .91 −.11a .02 .04 .54a .30a .32a Delinquency 1.67 .81 .06 −.18a −.06 −.27a −.12b −.39a −.25a English Achievement 4.47 .97 −.09a .27a −.16a .25a .03 .25a .14a −.24a a Correlation is significant at the .01 level (2-tailed). b Correlation is significant at the .05 level (2-tailed). Table options Structural model The hypothesized mediated model was tested and the fit was compared to a partially-mediated model. These models used a simplified first-order structure of the measures by computing scale scores for the subscales loading on to the latent scales (delinquency, academic self-efficacy, and social self-efficacy). These composites were calculated using the factor score regression weights from one-factor congeneric models (Bagozzi & Heatherton, 1994). Maximised reliability coefficients were also calculated; all except stealing (rc = .64) were found to have acceptable levels above .70. The other constructs (academic aspirations, self-regulatory self-efficacy) remained as first-order factors with item-level data. The hypothesized mediated model showed good fit to the data (c2 = 2247.43, df = 241, p < .001; RMSEA = .08; NFI = .84; CFI = .85). However, the partially-mediated model showed a significantly better fit to the data (c2 = 2181.86, df = 238, p < .001; RMSEA = .07; NFI = .84; CFI = .86) than the fully mediated model (Dc2 = 65.57, df = 3, p < .05). When non-significant paths were removed, the final model fit was satisfactory (c2 = 1994.89, df = 219, p < .001; RMSEA = .07; NFI = .85; CFI = .87). Thus, the partially-mediated model without non-significant paths was accepted over the hypothesized mediated model. The final model is presented in Fig. 3. Structural estimates of final model controlling for sex and school ... Fig. 3. Structural estimates of final model controlling for sex and school socio-economic status. Figure options As expected, academic self-efficacy was related to both academic aspirations and delinquency, and directly related to academic achievement. The indirect effect of academic self-efficacy on achievement via delinquency was significant (Sobel z = 2.02, p < .05) Self-regulatory self-efficacy, also as expected, was negatively related to delinquency and positively related to academic achievement. Delinquency was negatively related to academic achievement. Again, the indirect effect of self-regulatory self-efficacy on achievement via delinquency was significant (Sobel z = 2.48, p < .05). Of the control variables, gender was significantly associated with delinquency (loading = −.21, p < .05), indicating that females were less likely to report delinquent behavior than males, and with academic achievement (loading = .23, p < .05), indicating that females were more likely to do better in academic achievement than males. School socio-economic status was also significantly related to academic achievement (loading = −.17, p < .05). However, examination of the results showed some statistically non-significant paths. Age was not significantly related to either delinquency (loading = −.03, n.s.) or academic achievement (loading = −.02, n.s.), and school socio-economic status was not significantly related to delinquency (loading = −.02, n.s.). More importantly, academic aspirations was not significantly related to academic achievement as hypothesized. This path was significant in the fully mediated model, thus the direct relationship between academic self-efficacy and academic achievement appears to be stronger than that between academic aspirations and academic achievement. Our findings, therefore, suggest that academic aspirations does not mediate the effect of academic self-efficacy on academic achievement. The second unexpected finding was the negative relationship between social self-efficacy and academic achievement. Given that the bivariate correlation between the two was non-significant, it is likely that this negative relationship was due to controlling for the strong positive effects of academic self-efficacy and self-regulatory self-efficacy on academic achievement.