The Effect of Unexpected Income Loss on Weight: Temporary Unemployment and BMI

More than one-third of adult Americans are obese. Obesity causes adverse health outcomes that could lead to lower productivity and higher health care costs. Stylized facts show that low-income women are more likely to be obese than high-income women. Most previous studies examining the obesity-income relationship focus on the impact of weight on economic status, with only two recent studies (Schmeiser, 2008; Cawley, Moran, & Simon, 2010) estimating the causal impact of income on weight. While one of these two papers (Schmeiser, 2008) found a statistically significant (positive) impact of income on body weight, both studies used identification strategies that limited the representativeness of results to low-income and/or elderly segments of the U.S. population.

The goal of the present analysis is to estimate the causal effect of income on weight using exogenous variation in household income created by transitory and unanticipated unemployment spells experienced by the spouse. A unique feature of this identification strategy is that it allows for estimation of the causal impact of income on body mass index (BMI) for a wider segment of adult population than previous research. The study draws upon the 1998-2013 waves of the Panel Study of Income Dynamics (PSID) to identify unanticipated, transitory and thus exogenous changes of household income resulting from an unemployment spell of the spouse. This identification strategy necessarily limits the analysis sample to couples. Changes in household income are assumed to be exogenous if they result from an unemployment spell of the spouse that occurred due to a layoff or being fired, illness or injury, employer bankruptcy, or employer selling the business. This excludes separations when the spouse voluntarily left his or her job, which could be endogenous to the individual’s weight or consumption decisions of the household. In order to identify unemployment episodes that are less likely to be voluntary and more likely to have a significant impact on total household income, only unemployment spells that last one month or longer are used. Additionally, to ensure that the unemployment spells are not voluntary, or that they are not permanent or seasonal, the analysis sample is restricted to households with strong labor force attachment.

Instrumental variable models of BMI (continuous outcome) and obesity (dichotomous outcome) are estimated separately for the samples of husbands and wives. The estimations control for individual and time fixed effects, as well as a rich set of demographic and socio-economic variables. Findings indicate no causal effect of income on BMI of men. For women, results reveal a quantitatively small and marginally statistically significant positive impact of income on BMI. A loss of $1,000 of household income decreases women’s BMI by about 0.02 points. Additional estimations reveal that the effect of income on women’s weight is concentrated in low-income households.