Methods: Using data from waves three and five of the Fragile Family and Child Wellbeing Study, we used a cross-lagged panel design to test the direction of causality between latent variables representing parenting stress (PS) and maternal depression (DEP). Structural Equation Modeling (SEM) was used to test four specific models: (1) a baseline model with autoregressive paths; (2) a model with autoregressive effects and PS wave three predicting DEP at wave five; (3) a model with autoregressive effects and DEP wave three predicting PS at a wave five; and, (4) a fully cross-lagged model with the autoregressive effects and both PS and DEP at wave three predicting each other at wave five. The sample for this study consisted of 2166 mothers who had complete data for both waves. In each of the models, the baseline latent variables representing PS and DEP were hypothesized as correlated. Direct paths were then hypothesized between the latent variables. The error terms associated with the latent variables representing PS and DEP at wave five were also hypothesized as correlated, and the error terms of the same PS and DEP items measured at wave three and wave five were also hypothesized as correlated. These error terms were hypothesized as correlated because we assumed that factors contributing to measurement error in any particular item or latent construct would be consistent across the two time periods.
Results: The baseline model was of interest because it allowed us to evaluate the underlying measurement models as well as the stability of the constructs over time. The fit indices indicated that the measurement models provided an adequate fit to the data in our study. In addition, the autoregressive paths for both latent variables in the baseline model were statistically significant, indicating that the constructs were stable over the two time periods. More important for this study, however, are the differences in chi-square values between the models that were compared in our analysis. The chi-square difference test was of particular interest because it allowed us to compare these models that were hierarchically nested. Results of the chi-square difference test showed that Model 4, the fully cross-lagged model, provided a significantly better fit to the data than the other competing models tested in this study. These results supported the hypothesis that a reciprocal relationship existed between PS and DEP.
Implications for Interventions: Based on the findings of this study, interventions that address both parenting stress and maternal depression are warranted.