Source
Acharya, Blackwell and Sen (2016), LIT-1388, read in full in NOTE-1089 from The Journal of Politics 78(3): 621–641, the typeset PDF posted by the second author. The online appendix, which holds many of the robustness checks, was not read.
What was actually shown
- The association. Pooled CCES surveys (over 40,000 white respondents in 1,329 Southern counties) and ANES thermometer scores. A 20 point increase in 1860 slave share goes with a 2.3 point fall in the share of whites identifying as Democrats and comparable shifts on affirmative action and racial resentment: “approximately a 0.09–0.13 standard deviation change in the CCES outcomes” (p. 626). With state fixed effects and 1860 economic, demographic and geographic controls (Table 1).
- The causal reading. Instrumenting slave share with cotton suitability gives larger estimates (Table 2). The authors call the exclusion restriction “a strong assumption” (p. 628). Holding present black share fixed by sequential g-estimation leaves the direct effect about as large (Table 3), against racial threat. Migration in 1935–40 and present income inequality do not account for it (Figure 3, Table 4).
- What could have come out otherwise. The timing could have shown the difference before the war; it did not. High- and low-slave counties voted alike before the war, and in the 1851 Georgia and Mississippi elections fought over slavery and secession (Figures 4–5); the difference grew with disenfranchisement after Reconstruction. The mechanization test could have shown no attenuation; slavery’s effect on partisanship nearly vanishes where tractors spread fastest in 1930–40 (Table 5). Lynchings in 1882–1930 were higher where slavery had been.
- Transmission. Southern parents’ 1965 racial thermometer gaps correlate with their children’s into 1997, and the slavery effect is as large among whites born after 1965 (Table 6). The authors conclude that political culture, “rather than exclusively institutions”, plays a role (p. 638), and do not rule out institutions.
What this does not say
- Not that slavery’s attitudes were carried forward unchanged. On the paper’s own timing the difference was produced after emancipation. The persistence is of a post-emancipation divergence, not of antebellum opinion.
- Not that most Black Belt whites differ. The effects are about a tenth of a standard deviation.
- Not party persistence. “it is sociopolitical attitudes, rather than partisanship, that are passed down from parent to children” (fn. 23); the same counties were more Democratic before about 1950 and are more Republican now.
- Not a family-level finding. Respondents are assigned the slavery of the county where they live now. Family descent is not observed.
- Not that institutions played no part. Families and local institutions are not separated.
Connections
- CLAIM-420. The claim names this work as the hard case for telling association from transmission. The association is strong; transmission is argued with thin evidence; and the transmitted attitude was formed after the institution’s abolition, so this is causal descent from slavery joined to perspectival transmission of a later reaction.
- CASE-133. An attitude that kept its place and changed its party is the case’s carrier-and-character distinction in vote data.
- CLAIM-329. A test of persistence by transmission against persistence by resemblance (similar present conditions), for an attitude rather than a collective.
- LIT-1372 (Pierson). Cited by the paper for the path dependence of institutions. The mechanism here is not increasing returns but repression incentives followed by socialization.