Are Recessions Really Good For Your Health? Understanding Procyclical Mortality
Principal Investigator: Doug Miller (Department of Economics, UC Davis)
With our pilot grant funding, we were able to hire an RA for several quarters' effort. With this, we were able to push the project forward in several ways.
First, we compiled data from several large micro data sets (Vital Statistics mortality records, Cancer-SEER population denominator files, Basic Monthly Current Population Survey, and Census data) and published data sets such as the Labor force statistics from the BLS. After gathering and cleaning these data, we were able to create analysis data files that had at the state-year level our control variables, our key RHS variable, and various key LHS variables. These LHS variables include overall age-adjusted mortality, and mortality for specific age and race/sex groups.
Second, we estimated the responsiveness of mortality to the state-year unemployment rate, with controls for state and year fixed effects, state trends, and demographic controls. We did this for overall (age-adjusted) mortality, and for single-year age-specific mortality. We computed the semi-elasticity of mortality with respect to the unemployment rate, as well as the number of averted deaths for each age that would occur nationally in the most recent year of our sample due to a one percent increase in the unemployment rate.
Third, we re-estimated our main models stratifying by sex, and examining different causes of death. In doing this, we discovered that a large relevant category of death was "other deaths". With some effort, we were able to re-classify many of these deaths into interpretable categories.
Fourth, we ran a "horse race" regression, comparing the impact of "own group" unemployment rates and "state" unemployment rate. The relative importance of the coefficients in this model enable us to shed light on the competing hypotheses of "own activity" and "externalities".