TFA
Recession-induced mortality declines are driven primarily by external effects of reduced aggregate economic activity on mortality, and recession-induced reductions in air pollution appear to be a quantitatively important mechanism.
The authors mathematically correlate the recession with a reduction in mortality, this is non-controversial (that is if we agree that the mortality statistics are valid). What happens next is where I agree with Junta - this conclusion is some pretty broad hand waiving speculation that is certainly persuasive, but not actually justified as a proven fact:
(1) from Page5 of the article, the source data is CDC (for young people and all cause mortality) and Medicare data (for retiree mortality), Bureau of Labor Statistics for actual employment, EPA Air Quality data as a proxy for drivers of mortality from pollution, Behavioral Risk Factor Surveillance Survey (BRFSS) to identify mortality driven by work activities, and Medicare Health and Retirement Survey for 2002-2014 to account for drivers of mortality amongst nursing home residents.
(2) they (page 10) then attempt to measure the "Shock" impact of the Great Recession in terms of mortality overall, in an attempt to remove confounders, they look at regional variations (between states) and level them out so as to approximate only the median effect of the recession upon Mortality (the claimed reduction),
(3) finally by page 20 the authors begin to make hypothesis about the causation of the decline (which pre-supposes their math in the above sections are sufficiently robust, as I am not a statistician I'll leave that debunking to others). The authors report first on "internal" effects which is about non-aggregate single-person behavior, like seeing your own doctor and eating healthily:
Moreover, when we look directly for evidence of internal effects, we find no evidence of a substantive role for these channels. We find no evidence of a statistically significant impact of the Great Recession on self-reported health behaviors
Then they look at external effects, communicable diseases, quality of healthcare, and then _finally_ pollution. They only consider these factors based on prior papers that suggest correlation, they don't provide a rational in the paper itself as to why you discount other possible sources of the change in mortality, so here's the first point where I think this is quite broad speculation masquerading as "hard statistical analysis":
We find little support for a role for the first two classes of external effects, but evidence consistent with a quantitatively important role for recession-induced reductions in air pollution in explaining over one-third of the recession-induced mortality declines.
Essentially they restrict the analysis to three possible causes, and with lack of proof of either of the first two analyzed causes they pull a Sherlock Holmes and "ergo the cause must be item number 3". While we can agree that there is correlation, and it seems valid to assume pollution does in fact lead to mortality, the nature of the causation is left as an "a-priori" statement and they proceed immediately to calculation of the magnitude of the connection.
