Necessary Conditions for the Imprinting Explanation . | Analysis . | Result . |
---|---|---|
1. Reduced aggregate disparities in 1918 were driven by cohorts that could have been exposed to the 1890 virus during critical developmental periods. | Empirical | Yes: Reduced disparities were driven by cohorts aged 20–39 in 1918. |
2. Either a. relevant cohorts of urban White 1918 populations had greater exposure to the 1890 flu than relevant cohorts of urban non-White 1918 populations or b. relevant cohorts of these populations had similar exposures, which swamped other factors that tended to produce higher non-White influenza mortality in other years. | Empirical | Suggestive yes (2a): Using urban origins as a proxy for childhood influenza exposure, young adult urban White populations had greater exposure than young adult urban non-White populations. Suggestive no (2b): City-level factors were not less predictive of disparities in 1918 compared to prior years, failing to provide evidence of imprinting “swamping” other factors (analysis reported in the online appendix). |
3. Populations with greater 1890 flu exposure had higher mortality in 1918. | Empirical | Suggestive yes: Using urban origins as a proxy for childhood influenza exposure, city populations with greater exposure had higher 1918 mortality. |
4. 1890 influenza exposure was sufficiently a. prevalent and b. deleterious to survival in 1918 to account for reduced racial disparities in the relevant cohorts. | Simulation | Not disproven: Simulations find that, to account for reduced disparities, imprinting would need to be highly prevalent among urban Whites in relevant cohorts, but that this prevalence could result from plausible attack rates. |
Necessary Conditions for the Imprinting Explanation . | Analysis . | Result . |
---|---|---|
1. Reduced aggregate disparities in 1918 were driven by cohorts that could have been exposed to the 1890 virus during critical developmental periods. | Empirical | Yes: Reduced disparities were driven by cohorts aged 20–39 in 1918. |
2. Either a. relevant cohorts of urban White 1918 populations had greater exposure to the 1890 flu than relevant cohorts of urban non-White 1918 populations or b. relevant cohorts of these populations had similar exposures, which swamped other factors that tended to produce higher non-White influenza mortality in other years. | Empirical | Suggestive yes (2a): Using urban origins as a proxy for childhood influenza exposure, young adult urban White populations had greater exposure than young adult urban non-White populations. Suggestive no (2b): City-level factors were not less predictive of disparities in 1918 compared to prior years, failing to provide evidence of imprinting “swamping” other factors (analysis reported in the online appendix). |
3. Populations with greater 1890 flu exposure had higher mortality in 1918. | Empirical | Suggestive yes: Using urban origins as a proxy for childhood influenza exposure, city populations with greater exposure had higher 1918 mortality. |
4. 1890 influenza exposure was sufficiently a. prevalent and b. deleterious to survival in 1918 to account for reduced racial disparities in the relevant cohorts. | Simulation | Not disproven: Simulations find that, to account for reduced disparities, imprinting would need to be highly prevalent among urban Whites in relevant cohorts, but that this prevalence could result from plausible attack rates. |
Note: The leftmost column lists necessary conditions for the explanation to hold. Conditions 1–3 are evaluated, respectively, by tests D1–D3 of Table 2. Conditions 4a and 4b are jointly evaluated through test D4, and the outcome of test D4 is further evaluated through test D5.