Table 2 Fit statistics for the null and causal model Models df χ 2 RMSEA SRMR CFI AIC Model comparison ∆df ∆χ 2 Selleck INCB024360 Auto-regressions 222 2,294.224* .0525 .0438 .978 2,450.224       Causal model 219 2,230.428* .0521 .0369 .979 2,392.428 1 vs. 2 3 53.80* Reversed causal model 219 2,231.221* .0521 .0358 .979 2,393.221 1 vs. 3 3 63.00* Reciprocal model 216 2,189.406* .0519 .0334 .979 2,357.406 2 vs. 4 3 51.02*               3 vs. 4 3 41.82* * p < .05 Comparing the different models (Models 2, 3, 4) to the stability model (Model 1) revealed

that all three models show a significant decrease in chi-square, indicating a better fit. Model 4 shows, however, the largest decrease in chi-square (Δχ 2 = 104.82, df = 6, p < .05). In order to test further which of the models is the most parsimonious, these models were compared to each other and Model 4 showed even in comparison with Models 2 (Δχ 2 = 51.02, df = 3, p < .05) and 3 (Δχ 2 = 41.82, df = 3, p < .05), a significant decrease in

chi-square. Additionally, this was also confirmed by comparison of the other fit indices (RMSEA, SRMR and AIC; Table 2). Consequently, the reciprocal model (Model 4) was accepted as the best fitting model. Figure 1 shows the reciprocal model and the standardized paths estimates. In the best fitting model (Model 4), higher levels of work–family IWR-1 purchase conflict at time 1 are associated with performance-based self-esteem (β = .06, p < .05) at time 2 after control for children, gender, education and age. However,

no relationship between work–family conflict at time 1 and emotional exhaustion at time 2 could be established. Emotional exhaustion GDC973 at time 1 was related to work–family conflict (β = .09, p < .05) and performance-based self-esteem (β = .04, p < .05) at time filipin 2. Moreover, performance-based self-esteem at time 1 was related to work–family conflict (β = .10, p < .05) and emotional exhaustion (β = .04, p < .05) at time 2. In addition, some covariates were related to the constructs of interest at time 1, children living at home (β = .07, p < .05), university education (β = .14, p < .05) and age (β = −.07, p < .05) were positively related to work–family conflict; gender (β = .05, p < .05), university education (β = .11, p < .05) and age (β = −.11, p < .05) were related to performance-based self-esteem; and gender was positively related to emotional exhaustion (β = .13, p < .05). Further, we tested in the best fitting model whether the structural paths were different for men and women. Multiple-group analysis did not show differences in the relations of the tested constructs over time for men and women (Δχ 2 = 87.12, Δdf = 21, p > .05). Discussion The study had two overall aims; first, we tested the prospective associations between emotional exhaustion, performance-based self-esteem and work–family conflict; secondly, we wanted to investigate possible gender differences in the relations between the three constructs.