I hear a lot about deaths rising with recessions. What does the data indicate about the robustness of such a relationship?
Below is a plot of the deaths/population ratio, in logs.
Figure 1: Log ratio of deaths to population. NBER defined recession dates shaded gray. Source: CDC, Census via FRED, NBER, and author’s calculations.
Estimating a log-log regression of deaths on population, 1999-2018, one finds that the semi-elasticity of deaths with respect to recession is -1.5%, although it is not statistically significantly different from zero.
Interestingly, augmenting the regression with a “Trump dummy” leads to a Trump effect of 6.4% (statistically significantly different from zero at the .01%).