Background:
Obesity prevalence among schoolchildren substantially increases as they move from 1^st to 5^th grade. Physical education (PE) classes in primary schools promote physical activity and may mitigate the development of child obesity. Estimates of PE class effectiveness are thus important for schools to design appropriate physical activity policies.

Objectives:
This study investigates the relationship between PE instruction time and the development of child obesity. More specifically it uses longitudinal data to assess how the length of PE instruction time is associated with individual body mass index (BMI) and obesity status after controlling for child sociodemographics and school characteristics.

Methods:
We use the Early Childhood Longitudinal Survey (ECLS) for this analysis. Our sample follows a cohort of 7,819 children as they progress from 1^st grade to 5^th grade (1997-2001). These children are sampled from 998 schools across the United States.

Body mass index (BMI), defined as weight in kilograms divided by height in meters squared, is constructed from measured child weight and height collected in the spring of 1^st grade, 3^rd grade and 5^th grade. We compute the change in BMI, the dependent variable of our analysis, between 1^st and 3^rd grades, 3^rd and 5^th grades and 1^st and 5^th grades. The number of hours of PE instruction time during the past week, the explanatory variable of interest, is provided by the school administrator at the same times that child BMI is collected for all three waves.

Multivariate linear regression is used to discern the effect of PE instruction time on children’s BMI growth trajectory while controlling for child sociodemographics and school characteristics. We regress the change in a child’s BMI between 1^st and 3^rd grade (and between any two of the three available years) on the duration of PE time in 1^st and 3^rd grade. A linear probability model is used when obesity status, a dichotomous variable, is the outcome measure. Our models use robust standard errors to correct for heteroskedasticity and cluster on schools to control for shared characteristics of sampled students within the same school.

Results:
We find that past PE instruction time is a significant predictor of future BMI change and obesity status for males but not females. The average change in BMI in the sample between 1^st and 3^rd grades or between 3^rd and 5^th grades is approximately 1.8 units while the average BMI in 1^st grade is about 17.0 for males. We find that an increase in one hour of PE instruction time per week in 1^st grade is associated with a 15.7% decrease in the change in BMI between 1^st and 3^rd grade for males (coefficient=-0.157, p<0.01). We find an effect of similar magnitude but less significance for an increase in one hour of PE instruction time per week in 3rd grade on BMI change between 3rd and 5th grade for males (coefficient=-0.127, p<0.05). An increase of one hour of PE in 1^st grade is associated with a 24% decrease in BMI change between 1^st and 5^th grade (coefficient=-0.243, p<0.02).

As the definition of obesity depends not just on the BMI but also on the age and sex of the child, a dichotomous indicator of obese status may be easier to interpret. In our sample, the rate of obesity among males is 14% in 1^st grade, 20% in 3^rd grade and 24% in 5^th grade. We find that an increase of one hour in PE instruction time per week in 1^st grade decreases the probability of being obese by 1.7% in 3^rd grade (coefficient= -0.017, p<0.05) and by 3.9% in 5^th grade for males (coefficient=-0.039, p<0.001). An increase in one hour of PE instruction time per week in 3^rd grade decreases the probability of being obese in 5^th grade by 3.0% for males (coefficient= -0.029, p<0.01). Our results would suggest that if PE class were 1 hour longer in 1^st grade, the obesity rate among our 5^th grade sample of males would be 20% instead of 24%.

Conclusions:
This study finds that PE instruction time can have a significant impact on the development of child obesity for males. The longitudinal approach in this paper is preferable to cross-sectional analyses since the length of past, rather than current, PE classes appears to be a better predictor of current obesity status. The effect of PE classes on the development of childhood obesity among females and within certain groups is still unclear. We will explore this heterogeneity in future analyses.

Support:
Robert Wood Johnson Foundation via Active Living Research