Are recessions good for everyone's health? The association between mortality and the business cycle by race/ethnicity in the US

Abstract

This article studies the effect of the business cycle on the mortality rates of the major racial/ethnic groups in the USA. We use county-level data from 1999 to 2005 and employ a panel econometric approach that includes county- and year-fixed effects. We found that the mortality rates for whites and latinos are procyclical, i.e, that economic expansions (contractions) are associated with increases (decreases) in mortality. Moreover, the magnitude of this relationship is larger for latinos than for whites. However, we generally do not find a statistically significant relationship for blacks. Finally, the procyclical relationship for whites and latinos increases in magnitude as the county becomes less racially/ethnically diverse. Taken together, these findings suggest that the procyclical association identified in previous studies of the overall US population may vary by race and ethnicity.

Acknowledgements

We thank Luis Robles for his excellent research assistance and the Robert Wood Johnson Foundation Center for Health Policy at the University of New Mexico for their financial support.

Notes

¹As the number of physicians is only available for 2003 forward, the values of 1999 through 2002 are linearly extrapolated. As a robustness test, the regressions were estimated without this variable and the results were largely unchanged.

²A number of robustness checks were performed, including using the level (rather than log) of the dependent variable, revising the included control variables, and excluding year fixed effects. The main results were largely unchanged.

³An alternative approach would be to use the entire sample and interact the unemployment rate with the various racial proportions. However, this approach implies that the relationship between the explanatory variables and the mortality rate does not vary across counties. In the estimates where the sample is subset, the coefficients on the explanatory variables vary significantly across subsets. Thus, the use of interaction terms may be inappropriate.