*Names in bold indicate Presenter
In this research, we use a sample of 4440 low-income families from the randomized control trial of Head Start Impact Study (HSIS) to explore whether maternal engagement in education predicts parenting practices. Because Head Start offers comprehensive services to families, mothers whose children attend such programs may increase their education at higher rates than those whose children do not. Therefore, the HSIS holds promise as a way to estimate causal effects of maternal education on parenting practices, if certain assumptions are accepted. Models explored in this study use a mediation framework, as maternal engagement in education is considered one mechanism through which random assignment to Head Start increases positive parenting practices.
Preliminary analyses suggest there are significant effects of assignment to Head Start on: a) maternal engagement in education in the year following random assignment (b = .03, se = .01, p =.02); and, b) family engagement in cultural activities two years after treatment assignment (b =.04, se = .01, p <.001). However, commonly used mediation procedures do not allow for causal estimates of the effect of the mediator on the outcome. We will use a regression-based subgroup approach (Peck, 2007, 2012) to expand on these analyses and explore the association between maternal engagement in education and a number of parenting practices, including cognitive stimulation, family engagement in cultural activities and involvement in schooling.
First, using a random subset of the treatment group, we will calculate the probability that mothers engaged in education based on substantial pre-treatment covariates. Second, we will apply these probabilities to the control group and remainder of the treatment group, so that everyone in the sample has a probability of engaging in education had they been assigned to treatment. Third we will create groups of participants who are predicted to engage in education across treatment and control conditions; the randomization of Head Start means that these groups can be assumed to be similar on observed and unobserved characteristics. We will then check the accuracy of predictions using the treatment group. Finally, we will conduct regression analyses comparing the parenting outcomes of those predicted to engage in education across the treatment and control groups. Implications for supporting policies that focus on improving maternal education for low-income families will be discussed in light of findings.