A Bayesian Analysis of Sibling Correlations in Health
Timothy Hallliday, University of Hawaii at Manoa
Bhashkar Mazumder, Federal Reserve Bank of Chicago
We investigate sibling correlations in health status using the Panel Study of Income Dynamics and Bayesian methods that allow us to estimate the covariance structure of a system of latent variable equations. Across a battery of outcomes, we estimate that between 50% and 60% of health status can be attributed to familial or neighborhood characteristics. Taking the principal component across all outcomes, we obtain a sibling correlation of about 53%. These estimates, which are larger than previous estimates of sibling correlations in health that rely on linear models, are more in-line with sibling correlations in income and suggest that health status, like other measures of socioeconomic success, is strongly influenced by family background. Therefore, efforts to improve the circumstances of families and communities may potentially lead to improved childhood health today and also reduce future health disparities.