The relatively large literature on the association between aggregate economic activity, for example, as measured by unemployment, and health (mortality) has produced a number of mixed results, which often vary by time period and model specification. In this fascinating article, Stevens et al. try to move the literature forward by identifying which demographic groups contribute most to the relationship, for example, is the relationship strongest for elderly. They show that, in fact, the association between unemployment and mortality is strongest for the elderly and secondly for the young. For people in the prime years of work life there is little association between economic activity and mortality.
However, the tenuousness of the association reflected by the varied results reported in the literature is apparent in this study too. Footnote 18 in the text is illustrative of the problem. When the period of analysis is limited to 1989 to 2002 there is no significant association between unemployment and mortality, or unemployment and mortality by place of death. This is an important fact because the paper concludes that nurse staffing in nursing homes may bean important cause of the pro-cyclical nature of mortality. The question is why that cause only manifests during the some periods and not others. The 1989 to 2002 period included two recessions.
Finally, the premise that staffing levels in nursing homes experience "shortages" seems inconsistent with the notion that markets clear. Perhaps, the fixed reimbursement for nursing home patients covered by Medicaid prevent wage adjustments and may cause "shortages", but this problem would cause a reduction in all inputs and substitution of private paying patients for Medicaid patients. The problem would also affect acute-care hospitals that treat large number of Medicaid patients and a much wider age range and more numerous number of persons. Moreover, the literature on the association between nurse staffing and patient outcomes is quite mixed (see for example Spetz et al. Medical Care Research and Review 2013) and not particularly supportive of the hypothesis that staffing has such large effects on patient outcomes.