Does obesity lead to poor school performance? Estimates from propensity score matching Hongyun Han Department of Sociology, University of Wisconsin-Madison March 26, 2012 ABSTRACT High body weight is negatively associated with test scores among elementary and middle school students. Are these negative outcomes due to preexisting differences, or are they a casual effect of childhood obesity? To better understand the causal mechanisms underlying this pattern, I use a propensity score matching approach to control for biases from observable preexisting differences, and conduct sensitivity analysis to assess the impact of biases from unobserved variables. Using data from the Early Childhood
Longitudinal Study, the matching models reveal that obese eighth graders, on average, score 0.17 standard deviations lower in reading and 0.16 standard deviations lower in math, a reduction roughly equivalent to one sixth of the racial achievement gap. Obesity penalties are larger for girls than for boys in both subjects. Differences between obese and normal-weight children decline slightly after adjusting for missing values. Findings from sensitivity analyses indicate that unmeasured variables would need to increase the odds of becoming obese by at least 20 percent to change the conclusion. Key words: obesity, academic achievement, propensity score matching
Does obesity lead to poor school performance? Estimates from propensity score matching Childhood obesity has become a public health crisis in the United States. The rates of obesity among children and adolescents have tripled over the past four decades(Wang and Beydoun 2007). Roughly one in five children and adolescence ages 2 through 19 was obese (Ogden et al. 2010). Treatments for obesity-related conditions in the United States cost roughly $150 billion per year (Cawley 2010). Past research has revealed substantial negative impacts of obesity on public health and the health care system (Finkelstein, Ruhm and Kosa 2005). Few studies, however, have examined the causal effect of childhood obesity on academic achievement. Although past studies have consistently shown that obesity is associated with lower levels of cognitive function (Li et al. 2008; Miller et al. 2006; Shore et al. 2008), scholars disagree about the impact of obesity on standardized test scores, and associated gender differences in the impacts (Averett and Stifel 2010; Datar, Sturm and Magnabosco 2004; Kaestner and Grossman 2009). Further, the methods employed in previous studies are insufficient to establish a causal effect of obesity on academic achievement—any observed negative effect may be due to preexisting differences rather than a casual relationship. Children who gain excessive weight may come from more disadvantaged families or possess other unobserved characteristics that lead to worse outcomes. For instance, the ability to concentrate may be an unobserved
characteristic that affects both weight and school performance; drawing conclusions about the causal relationship between obesity and poor test scores is difficult because of the potential for unobserved characteristics. To identify the casual effect of childhood obesity on academic achievement, I employ propensity score matching to reduce preexisting differences associated with observed variables. I use a sensitivity analysis to evaluate the strength of the matching estimates against the bias associated with unobserved variables. To alleviate the possibility of reverse causality, I also use predictor variables measured in fifth grade to predict outcomes in eighth grade. Using data from the Early Childhood Longitudinal Study, the matching models reveal that obese eighth graders score, on average, 0.17 standard deviations lower in reading and 0.16 standard deviations lower in math, a reduction roughly equivalent to one sixth of the racial achievement gap. These estimates are robust unless an unmeasured variable would have to increase...