The Impact of SNAP and Cash Public-Assistance Income on Age at Onset of Adverse Health Conditions: A Latent Variable Approach To Control for Family Background
Research Center: Institute for Research on Poverty, University of Wisconsin-Madison
Investigator: Fitzgerald, John and Rachel Connelly
Institution: Bowdoin College
9700 College Station
Brunswick, ME 04011
A substantial body of research reveals that growing up in a low-income family is associated with poor health in later life. Some of that research points to the effect of food insecurity on negative health outcomes. Other research examines the effect of poverty more generally. This study looks at USDA’s Supplemental Nutrition Assistance Program (SNAP) benefits and other forms of public assistance that provide increased resources for income-distressed families that may improve children’s health outcomes later in life, counteracting a lack of food or other financial resources experienced as a child that could have a lasting negative effect on one’s health.
Establishing that increased household resources provided by SNAP and other public-assistance programs improves health outcomes is a nontrivial task because unobserved factors that correlate with parental recipiency may also correlate with later poor-health outcomes for children. Estimation of the causal relationship between SNAP benefits and other cash-welfare amounts while a child and later health outcomes requires a methodology to deal with the endogeneity of welfare income within a model of health outcomes.
This study uses a two-part empirical strategy to avoid endogeneity bias: an instrument for public-assistance benefits based on exogenous variation in benefits that are not correlated with child background is estimated; and then a latent variable framework is employed, which jointly estimates the health onset hazard and the proportion of time between the ages of 0 and 8 that one’s family received any public assistance dollars. The latent variable framework absorbs the endogenous correlation between realized public assistance income (proportion of time receiving public assistance times the simulated average potential benefit amount) and the health onset outcome variable. A more standard two-stage residual inclusion (2SRI) model is also estimated to compare with the latent variable framework. The models are estimated using data from the Panel Study of Income Dynamics (PSID), which includes direct observation of the family’s fortunes when the child was between 0 and 8 years of age and measures of the age of onset of three chronic conditions: the age of onset of any chronic metabolic condition that includes high blood pressure, heart disease, or diabetes; the age of onset of any chronic condition listed in the survey; and the age of onset of asthma. The sample is limited to children born between 1966 and 1986 whose average family income during the years between ages 0 and 8 is less than 300 percent of the poverty line and who can be observed in sample until at least age 24. Almost 1,600 children satisfy these stringent criteria.
Thirty-two percent of the sample reported having a chronic metabolic syndrome by the end of the panel. Seventeen percent reported having asthma by the end of the panel and 58 percent reported having any chronic condition. Average total benefit dollars among those receiving any public assistance was $6,053, of which 62 percent came from SNAP. The sample was 55 percent black and 7 percent Hispanic and 8 percent were low-birth-weight babies. Comparing the PSID data on health with larger and more comprehensive National Health Survey statistics provides reassurance that the PSID health data is well measured.
The results on the age of onset hazard of chronic conditions show that modeling the endogeneity of public assistance recipiency is essential to analyzing the effect of public assistance recipiency on the age of onset of negative health outcomes. The null hypothesis of an exogenous proportion of one’s childhood spent living in a household that receives public assistance is rejected in both the 2SRI and the latent variable frameworks for all three health outcome models. Once the endogeneity of annual public assistance dollars is accounted for, the effect of an additional dollar of public assistance increases the age of onset of both any chronic condition and any metabolic condition. The age of onset of asthma is not found to be affected by an additional public assistance dollar. The difference in the results provides indirect evidence that the causal pathway between increased public assistance benefits and a reduction of the speed to metabolic conditions is through increased food security that public assistance brings.
Increasing the amount of benefit income by $100 a month (for example) would have two effects on the hazard of onset: increases in benefit income increase the probability of take-up, that is, it increases the estimated proportion of time one’s family received public assistance and it has a direct negative effect on the hazard of onset. Together the marginal effect of an extra $100 a month is estimated to reduce the hazard of metabolic syndrome by -.011, where the mean hazard at age 40 is .027.
Extending the model to allow a direct effect of the proportion of time one’s family received public assistance on health outcomes does not change the basic results. Estimating the model with sibling data allows the researchers to control for fixed family effects but increases issues with measurement error and reduces sample size substantially. The sibling model estimates also find a protective impact of benefit income on slowing the onset of chronic disease. However, these effects appear to be too large relative to those estimated in the other two models.
Overall, the empirical results from this careful specification of age of onset hazard demonstrates that what happens to a child between the ages of 0 and 8 has a significant effect on their lifetime health profile. That public assistance dollars can be shown to have an impact on adult health up to 40 years later indicates that public debate on benefit levels should include these long-term positive consequences of relaxing a family’s budget constraint in the short term.