Lee, Kenneth T. H. (2016). The impact of adolescent physical activity on educational attainment. 2016 Add Health Users Conference.
Although reviews of observational and experimental studies show positive associations between physical activity (PA) and academic achievement (Centers for Disease Control and Prevention, 2010; Fedewa & Ahn, 2011; Howie & Pate, 2012), the majority of the studies in these reviews is focused on children and further and more rigorous examination for adolescents with causal implications is important. Drawing data from Add Health, I examine the extent to which physical activity, measured using a composite of PA items in Wave 1, predicts educational attainment, measured by years of completed schooling in Wave 4. I find a one standard deviation increase in PA is associated with a .132 year increase in completed schooling in the presence of the host of child, family, and other demographic controls in an OLS regression. However, an OLS regression of educational attainment on PA may not produce a consistent estimate of the impact of PA. Among others, issues such as selection into and misreporting of PA may generate a correlation between PA and educational attainment even if PA does not have a causal impact. To account for the presence of these issues, I use the number of different PA resources in a 8.05 km network radius about an adolescent’s home from Obesity and Neighborhood Database in Wave 1—conditioned on total annual snowfall, and the total number of crimes and the proportion aged 25+ with no high school diploma or equivalency in a respondent’s county as an instrumental variable- to generate a consistent estimate of the impact of additional PA on educational attainment. The results from the 2SLS model using the PA resource counts as an instrumental variable and aforementioned conditions shows that a standard deviation increase in PA is predictive of an additional .587 years of schooling. Additional analyses to be completed include examining the heterogeneous impacts by gender and racial subgroups, as well as by health. Mediation and moderation by income, gender, and race are also additional factors to consider in this analysis. Robustness checks using different Euclidean and network radii (1, 3, and 5km) for the PA resource counts will be used to confirm the 2SLS results. It would also be fruitful to examine grades in curricular classes as an achievement outcome.
2016 Add Health Users Conference
Lee, Kenneth T. H.
City of Publication