Indiana University SPEA Edward J. Bloustein School of Planning and Public Policy University of Pennsylvania AIR American University

Panel Paper: The Impact of Public Assistance on Age at Onset of Adverse Health Conditions: A Latent Variable Approach to Control for Family Background

Friday, November 13, 2015 : 8:30 AM
Merrick I (Hyatt Regency Miami)

*Names in bold indicate Presenter

Rachel Connelly and John Fitzgerald, Bowdoin College
Public assistance including Food Stamps/SNAP provides increased resources for families that may improve children’s health outcomes.   This project focuses on the role of public assistance in delaying or preventing the onset of certain adverse health conditions, namely diabetes, asthma, high blood pressure, and heart problems. This study estimates how family income and public assistance recipiency during childhood affect the length of time before the onset of these chronic health conditions. Estimation of the causal relation between assistance benefits and later health outcomes accounts for the endogeneity due to unmeasured family background variables that affect both public assistance usage and later child health. This project uses a hazard model for the time from birth to onset of certain adverse health conditions allowing for endogenous income and benefit covariates. Non-assistance income and assistance benefits are jointly modeled and linked to the health onset hazard by correlation in unobserved latent variables.  The model is estimated using data from the Panel Study of Income Dynamics (PSID) for a sample of low-income children for whom we observe family background as a child age 0 to 8 who are followed to at least age 25 to observe later health outcomes.

                  The model includes three equations (a discrete-time onset hazard, public assistance benefits, and non-benefit income) linked by a common latent variable reflecting unmeasured heterogeneity in family background.  We estimate the model by maximum likelihood, estimating each equation conditional on the latent random effect, which is integrated out numerically.  The model exploits state and time variation in benefits (by family size) to get exogenous variation in benefits, conditioning on non-benefit income. Preliminary results indicate that ignoring family background  heterogeneity and assuming exogenous income and benefit sometimes results in insignificant  impacts of public assistance on onset hazards, or even spurious positive impacts, speeding the onset of disease. In models where we allow and control for endogeneity of benefits in the hazard, we find that income and benefit income have significant impacts in postponing the onset of adverse health conditions.