Abstract
As children grow up, they may encounter changing family structures and poverty status. Any attempt to measure the effects of these statuses on children's school achievement runs the risk of spurious effects due to child- and family-heterogeneity. An analytic strategy for avoiding these spurious effects is to use longitudinal data on families, children, and their academic achievement to estimate random coefficient growth-curve models in which a large number of causally prior control variables are allowed to impact both the intercept and slope of the child's achievement trajectory, while poverty and family structure at each point in time enter the model as time-varying covariates. An even more powerful strategy is to use Allison's [Allison, P. D. (2005). Fixed effects regression methods for longitudinal data using SAS. Cary, NC: SAS Institute, Inc.] “hybrid” version of this model, in which a fixed-effects specification differences away all unchanging child and family characteristics. We use CNLSY79 data to estimate both types of models for the effects of poverty status and family structure on children's mathematics achievement between ages 5 and 14. We find that poverty status exerts a modest, statistically significant negative effect on math achievement, but only among younger children. The correlation between family structure and children's mathematics achievement is largely spurious, due instead to child- and family-heterogeneity on causally prior variables.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have