Estimating the Effects of Maternal Education on Summer Learning: How Method Matters
*Names in bold indicate Presenter
The limited research on summer learning suggests that children with more-educated mothers tend to gain skills over the summer whereas those with less-educated mothers tend to lose. Children who maintain or build skills over the summer return to school better positioned to take on the challenges of their new grade compared to children who lose academic knowledge over the summer. Thus, it is important to further examine the relation between maternal education and summer learning.
In determining how best to estimate the relation between maternal education and summer learning, it is necessary to take into account that different modeling techniques will yield different results. As noted by Quinn (2014), estimates of summer learning are sensitive to analytic strategies, which is important to consider given that policymakers are likely to accept the estimates at face value without understanding that the analytic approach affects the interpretation of the results. To address this issue, this study examines the effect of maternal education on summer learning using two different analytic strategies, highlighting the different estimates obtained from each.
Using nationally representative data from the Early Childhood Longitudinal Study, Kindergarten Class of 1998-99 (ECLS-K), the present study documents the differences in summer learning between kindergarten and first grade in reading and math using (1) change score regression models (namely, subtracting the reading/math score from Spring of Kindergarten from the reading/math score from the Fall of 1st grade), and (2) regressor variable models (namely, predicting Fall 1st grade scores using Spring of Kindergarten scores as a covariate). Models using the regressor variable method answer the question, “Overall, do students whose mothers have different levels of education earn different scores at Fall 1st grade if they shared the same score at Spring of Kindergarten?” whereas models using a the change score method answers the question, “On average, do students whose mothers have different levels of education gain different amounts over the summer?”
Preliminary results suggest that estimates are, in fact, different across modeling strategy. Although the pattern that children with more educated mothers tend to gain more over the summer than those with less educated mothers holds across models, the factors associated with these findings differ by analytic technique. For instance, in-home learning activities account for much of the variance in summer learning by maternal education using the regression models than the change score models. In essence, the different estimates suggest distinct and potential solutions to increasing summer learning for children whose mothers have varying levels of education.