Further Evidence of Cultural Transmission - Confederate Leader Names

E Further Evidence of Cultural Transmission: Confederate Leader Names In this Appendix, we explore whether postbellum migrants contributed to the steady expansion in Confederate affinity among non-Southern whites in the early 1900s, as measured using children’s names (see Figure 2). To ensure sufficient variation on the right-hand side, we construct and use for this analysis a less restrictive measure of “Confederate names” among children, based on first names that exactly match a Confederate leader’s legal or adopted first name, last name, or entire full name (e.g., enumerated first name of “Robert Lee”).1 For identification, we examine how non-Southern white parents named their children before and after moving, as a function of the share of Southern whites in their destination counties as of 1900.2 We pool all children born to white non-Southern parents in the first two Census rounds of the 20th century when Confederate culture became widespread nationally and many first-generation postbellum migrants remained alive. We consider households with ≥ 2 children: ≥ 1 born in the state of residence at time τ and ≥ 1 born in a different state by τ + 10. To avoid double counting children across Census periods, we restrict sample ages to 0–9. We then estimate the time of household move, τe, as the midpoint between the birth years of the children born in different states, where child year of birth is given by τe + j for j = −9, ..., 9. Overall, this yields 1,404,641 white children in 467,040 households. We proceed to estimate the following equation, which relates the given name of child i to whether their household h had yet moved to non-Southern county c at their time of birth τe + j: Confederate nameihcτ = θh + β · % Southern Whitesc,1900 × Born After Movei + X0iτ γ + εihcτ , (E.1) where Confederate nameihcτ is multiplied by 100 for interpretive ease. Household fixed effects, θh , absorb various characteristics of h including its cultural attitudes, its place of origin, and factors affecting its destination choice.3 The Xiτ vector includes dummies for child gender, birth order, birth period, and birth year relative to move in j. Standard errors are clustered by contemporaneous destination county. Our first set of results in Appendix Table E.1 uses a pooled sample of male and female children in multi-children mover households. The estimates in columns 1–2 suggest that a 1 p.p. increase in Southern white migrant shares in 1900 is associated with roughly a 0.09 p.p. increase in the likelihood that parents give their child a Confederate name, relative to a child born prior to the move. Though somewhat small, these estimates are statistically significant at conventional levels. In the remaining columns of Appendix Table E.1, we show that these results are, reassuringly, driven by boys. Most Confederate heroes whose names had become popular in the postbellum South—generals, politicians, and veterans—were men. Narratives also often focused on men. Historian David W. Blight 1 This of course entails a tradeoff between including names that are almost certain to reference Confederate leaders but are rarely given (e.g., “Stonewall”) and those shared by Confederate leaders but also commonly given otherwise (e.g., “Lee”). This broad measure ensures sufficient variation in the outcome, at the possible expense of attenuation bias in estimates. As a validation check, men from former slaveowning households (both fathers and sons) in the South are significantly more likely to name their children after Confederate leaders. Using the 1900 Census, we find that conditional on father’s birthplace, current county FE, and cohort FE, children aged 0–20 with fathers from former slaveowning households are (i) nearly 1 p.p. more likely to have a loosely defined Confederate leader’s name (as used in this Appendix section) relative to a mean of 22% among children without slaveowning fathers, and (ii) nearly 0.04 p.p. more likely to have a strictly defined Confederate leader’s name (as in Figure 2) relative to a mean of 0.2% among children without slaveowning fathers. 2 Bazzi et al. (2020) develop this strategy to identify effects of frontier exposure on individualism, and Bazzi et al. (forthcoming) adapt it to identify effects of exposure to Southern whites during the Great Migration on religious conservatism. 3 As there may be positive correlations between Southern white shares across origins and destinations, household fixed effects are important insofar as they control for households’ origin county Southern white shares. E-1 (2001, p. 266) argues that “by the sheer virtue of losing heroically, the Confederate soldier provided a model of masculine devotion and courage” to Southern whites. Thus, it is likely that boys’ names would contain a relatively stronger signal of exposure-based transmission of Confederate culture, if such transmission is indeed taking place. Columns 3–4 establish as much: a 1 p.p. increase in Southern white migrant shares in 1900 is associated with a statistically significant 0.15 p.p. increase in the likelihood that parents give their male child a Confederate leader name, relative to a male child born prior to the move. In contrast, estimates for girls, in columns 5–6, are small and insignificant at conventional levels. A causal interpretation implies that greater exposure to Southern white migrants induced a shift towards Confederate naming among other white migrants without Southern heritage. The parallel trends assumption here holds that, within a household, the likelihood of Confederate name-giving should not have changed across children had that household not moved to a county with a large Confederate diaspora. Appendix Figure E.1 provides event-study estimates in support of this assumption, showing limited evidence of pre-trends and gradual increases in post-move coefficients. This pattern is consistent with an exposure-based mechanism, in which it takes time for Southerners and non-Southerners to make contact and for cultural transmission to occur. Overall, these results suggest that migrants transmitted Confederate culture to non-Southern populations during a period of resurgent Confederate pride in the early 20th century. This analysis above provides an additional, granular perspective on the sort of localized, contact-based transmission of culture underlying our broader county-level findings throughout the paper. Together with our analysis of the Denver-area KKK in Section 5.3, our findings reveal an important role for non-Southern whites in helping grow the Confederate cultural diaspora, including both Confederate memorialization and racial norms, throughout the country. Table E.1: Cultural Transmission and Exposure Effects on Non-Southern White Names Dependent Variable: Child has Confederate Name Full Sample Non-Southern Boys Non-Southern Girls (1) (2) (3) (4) (5) (6) % Southern Whites, 1900 ×Born After Move 0.087∗∗∗ (0.027) 0.091∗∗∗ (0.027) 0.144∗∗ (0.059) 0.154∗∗∗ (0.059) 0.008 (0.031) 0.008 (0.031) Household FE Birth Year - Move Year FE Birth Order FE Birth Period FE Observations Households Outcome mean (pre-move) Yes Yes Yes Decade 1,404,641 467,040 20.14 Yes Yes Yes 5-Year 1,404,641 467,040 20.14 Yes Yes Yes Decade 536,052 222,686 34.56 Yes Yes Yes 5-Year 536,052 222,686 34.56 Yes Yes Yes Decade 514,282 214,091 5.06 Yes Yes Yes 5-Year 514,282 214,091 5.06 Notes: OLS regressions of an indicator for whether the non-Southern-ancestry child has a Confederate name (×100) on a dummy for whether that child in mover household h was born in its post-move county c × the share of Southern-born whites in county c in 1900. The full sample in columns 1–2 includes mover households from non-Southern origin states with at least 1 child born before moving and 1 child born after. Columns 3–4 and 5–6 further limit to mover households with at least 2 boys and 2 girls, respectively. Excluded Southern areas are those belonging to states of the former Confederacy and Oklahoma. All regressions include fixed effects for child gender, birth order, birth period (decade or 5-year), and birth year minus household year of move. In odd columns, the birth period FE are decadal and in even columns fiveyearly. Standard errors are clustered by contemporaneous destination county. Significance levels are denoted by * p < 0.10, ** p < 0.05, *** p < 0.01. E-2 Figure E.1: Cultural Transmission: Exposure to Southern Whites and Confederate Child Names βj +/- 2x std. error: % Southern whites .4 .2 0 -.2 -4 -3 -2 -1 0 1 year of birth relative to arrival 2 3 4 Notes: This figure isolates within-household, cross-child variation in parental exposure to Southern white migrants in the destination county based on the following equation: Confederate nameihcτ = θh + 9 X βj [% Southern Whitesc,1900 × 1(Born in τe + j)] + X0iτ γ + εihcτ , (E.2) j=−9 which allows β in equation (E.1) to vary with birth year relative to the household move, j = −9, ..., 9. Each graph reports estimates of βj and 95% confidence intervals for the relatively balanced event years j = −4, ..., 4. Each βj can be interpreted as the differential effect of exposure to postbellum Southern white migrants on the likelihood of a Confederate name given to a child born j years before/after their non-Southern-born parents moved to the county, relative to a child born one year prior to the household move. Estimates control for household fixed effects as well as birth order, birth decade fixed effects, and child gender. Standard errors are clustered by contemporaneous destination county. E-3