Confederate Leader Names among Children Born 1850–1940 Figures

children with Confederate leader names (per 1,000) Figure 2: Confederate Leader Names among Children Born 1850–1940 Southern whites in South Southern whites outside South Non-Southern whites outside South 1.5 1 .5 0 1850 1860 1870 1880 1890 1900 birth year 1910 1920 1930 1940 Notes: Three-year moving average in Confederate leader name frequencies across birth years among different subsets of Southern and non-Southern white populations (ages 0–9) in the U.S. Census: those born in the South living in the South (dark red), those born in the South living outside the South (bright red), and those born outside the South living outside the South (light red). An individual’s name match equals one if their given, first name is highly likely to have been given in reference to a Confederate leader. This includes individuals whose first name includes a leader’s full name (e.g., “Robert Lee”), as well as distinctive nicknames like Stonewall and last names like Beauregard. The list of Confederate leader names includes those with multiple public symbols in the Confederacy in the Southern Poverty Law Center’s (SPLC) “Whose Heritage?” Project. Figure 3: Confederate Culture Index, Post-1900 Notes: Map shows Confederate cultural index (CCI) scores for counties across the conterminous United States. This score is based on the sum of county-level indicators for (i) any matched Confederate memorials (i.e., monuments, location names), (ii) any United Daughters of the Confederacy (UDC) chapters, (iii) any 2nd Ku Klux Klan (KKK) chapters, and (iv) any recorded lynchings of Black people. With the exception of memorials, for which data feature limited information on the time dimension, we restrict these outcomes to those observed after 1900, thus following the migratory period of study. Note that these scores are highest in the formerly Confederate-controlled states and territories (outlined in black) but also tend to be higher in former border states, the West, and some major population centers (e.g., Chicago, the Northeast megalopolis). Tables Table 1: Occupational Sorting by Southern White Migrants, 1900 Dependent Variable: Individual Works as. . . Any Authority Position (1) (2) Lawyer or Judge (3) Law Enforcement (4) Public Administrator (5) Religious Worker (6) Educator (7) (a) Full Sample (N = 16, 187, 176) Southern Non-Southern mean 0.595∗∗∗ (0.063) 0.388∗∗∗ (0.041) 0.269∗∗∗ (0.053) 0.004 (0.019) -0.011 (0.007) 0.394∗∗∗ (0.039) -0.061∗∗∗ (0.017) 1.58 1.58 0.39 0.26 0.09 0.36 0.49 (b) Out-of-State Migrants Only (N = 8, 207, 851) Southern Non-Southern mean 0.726∗∗∗ (0.082) 0.550∗∗∗ (0.048) 0.389∗∗∗ (0.064) 0.068∗∗∗ (0.017) 0.013∗ (0.007) 0.239∗∗∗ (0.040) 0.016 (0.019) 1.47 1.47 0.33 0.24 0.08 0.48 0.35 (c) Western Counties Only (N = 1, 348, 860) Southern Non-Southern mean 0.802∗∗∗ (0.098) 0.503∗∗∗ (0.059) 0.463∗∗∗ (0.066) 0.135∗∗∗ (0.024) 0.002 (0.018) 0.183∗∗∗ (0.042) 0.020 (0.035) 1.71 1.71 0.54 0.22 0.14 0.35 0.47 (d) Non-Western Counties Only (N = 14, 838, 316) 0.546∗∗∗ (0.072) 0.360∗∗∗ (0.049) 0.224∗∗∗ (0.064) -0.026 (0.023) -0.014∗ (0.007) 0.442∗∗∗ (0.047) -0.079∗∗∗ (0.020) Non-Southern mean 1.57 1.57 0.37 0.26 0.08 0.36 0.50 County FE Occup. Income Category FE Yes No Yes Yes Yes – Yes – Yes – Yes – Yes – Southern Notes: Regressions of occupation indicators (×100) on Southern origin among various subsamples of white male individuals outside of the South between the ages of 18 and 64 in the 1900 U.S. Census. The “any authority” occupation indicator is based on any of the five positions of authority in columns 3–7. Sample derived from the complete-count Census, with Southern origin determined using state of birth. Southern states are those of the former Confederacy and Oklahoma. Panel (b) restricts the sample to include only migrants born in a different state from the given county. Panel (c) restricts to counties in the West Census region, and panel (d) to the non-West Census region. All regressions include destination county fixed effects. Individual controls include a cubic in age, marital status, and number of children. Column 2 additionally controls for occupational income category fixed effects, based on bins of occscore width 2. These fixed effects capture the historical income score for the average worker in the chosen occupation (e.g., occscore = 32 for law enforcement and = 30 for mechanics). Occupational classifications follow those laid out by the Census Bureau’s 1950 (“occ1950”) definitions. Standard errors are clustered by county. Significance levels are denoted by * p < 0.10, ** p < 0.05, *** p < 0.01. Table 2: The Confederate Diaspora in 1900 and Confederate Memorialization Any Confederate Memorials (2) Dependent Variable: (1) % Southern Whites, 1900 Estimator State FE Baseline controls 1870 share control Observations Outcome mean Adj. R2 F-statistic Anderson-Rubin, p-val KP Underident., p-val ∗ ∗∗∗ (3) Any United Daughters of the Confederacy Chapters (4) (5) (6) 0.008 (0.004) 0.020 (0.007) 0.034∗∗∗ (0.009) 0.014∗∗∗ (0.004) 0.018∗∗ (0.008) 0.042∗∗∗ (0.013) OLS Yes Yes IV Yes Yes OLS Yes Yes IV Yes Yes 1,701 0.25 0.22 1,701 0.25 IV Yes Yes Yes 1,701 0.25 1,701 0.09 0.24 1,701 0.09 IV Yes Yes Yes 1,701 0.09 30.6 0.00 0.00 15.0 0.00 0.00 30.6 0.02 0.00 15.0 0.00 0.00 Notes: Regressions of (i) Confederate memorials (coded as any Confederacy leader-inspired monuments as well as matched place names, street names or school names) and (ii) any United Daughters of the Confederacy chapters after 1900 on the share of Southern whites in 1900 in non-Southern counties (sample mean of 2.2%). Southern counties are those belonging to states of the former Confederacy and Oklahoma. Columns 2–3 and 5–6 instrument the share of Southern whites using a shift-share instrument based on the 1870 cross-sectional distribution of Southern whites and the predicted change in the Southern white population living outside the South from 1870 to 1900. The latter is generated via a set of flexible LASSO regressions (see equation 1). Baseline controls include log county population in 1870 and log county area (in square miles), entered as quadratic terms. Columns 3 and 6 also control for the share of Southern whites in 1870. All regressions include state fixed effects. The Anderson-Rubin p-value corresponds to the null hypothesis that the coefficient on % Southern W hites, 1900 is zero and that the overidentifying restrictions are valid. The KP Underidentification test p-value corresponds to the Kleibergen and Paap (2006) Lagrange Multiplier (LM) test whose null hypothesis is that the equation is underidentified. Standard errors are clustered at the 60×60 square-mile grid cell level, following the approach of Bester et al. (2011). Significance levels are denoted by * p < 0.10, ** p < 0.05, *** p < 0.01. Table 3: The Confederate Diaspora in 1900 and Expressions of White Supremacy Dependent Variable: % Southern Whites, 1900 Estimator State FE Baseline controls 1870 share control Observations Outcome mean Adj. R2 F-statistic Anderson-Rubin, p-val KP Underident., p-val Any Ku Klux Klan Chapters (1) (2) (3) Any Lynchings of Black People (4) (5) (6) 0.008∗∗ (0.004) 0.018∗ (0.010) 0.035∗∗ (0.016) 0.008∗∗ (0.004) 0.015∗∗ (0.006) 0.018∗∗ (0.009) OLS Yes Yes IV Yes Yes OLS Yes Yes IV Yes Yes 1,701 0.37 0.23 1,701 0.37 IV Yes Yes Yes 1,701 0.37 1,701 0.05 0.09 1,701 0.05 IV Yes Yes Yes 1,701 0.05 30.6 0.05 0.00 15.0 0.00 0.00 30.6 0.01 0.00 15.0 0.01 0.00 Notes: Regressions of (i) any 2nd Ku Klux Klan chapters and (ii) any recorded lynchings of Black people after 1900 on the share of Southern whites in 1900 in non-Southern counties (sample mean of 2.2%). Southern counties are those belonging to states of the former Confederacy and Oklahoma. Columns 2–3 and 5–6 instrument the share of Southern whites using a shift-share instrument based on the 1870 cross-sectional distribution of Southern whites and the predicted change in the Southern white population living outside the South from 1870 to 1900. The latter is generated via a set of flexible LASSO regressions (see equation 1). Baseline controls include log county population in 1870 and log county area (in square miles), entered flexibly using quadratic terms. Columns 3 and 6 also control for the share of Southern whites in 1870. All regressions include state fixed effects. The Anderson-Rubin p-value corresponds to the null hypothesis that the coefficient on % Southern W hites, 1900 is zero and that the overidentifying restrictions are valid. The KP Underidentification test p-value corresponds to the Kleibergen and Paap (2006) LM test whose null hypothesis is that the equation is underidentified. Standard errors are clustered at the 60×60 square-mile grid cell level, following the approach of Bester et al. (2011). Significance levels are denoted by * p < 0.10, ** p < 0.05, *** p < 0.01. Table 4: Identification and Robustness Checks on IV Estimates in Tables 2 and 3 Any Confederate Memorials (1) Dependent Variable: Any Any Any Lynchings KKK UDC Chapters Chapters of Blacks (2) (3) (4) CCI Score (from 0–4) (5) Alternative Standard Errors 0.034∗∗∗ (0.009) (0.017) (0.013) (0.007) 0.042∗∗∗ (0.013) (0.018) (0.022) (0.013) 0.035∗∗ (0.016) (0.020) (0.024) (0.008) 0.018∗∗ (0.009) (0.010) (0.010) (0.005) 0.130∗∗∗ (0.039) (0.057) (0.066) (0.030) 2. Initial 1870 Share Control Only 0.030∗∗∗ (0.009) 0.039∗∗∗ (0.013) 0.028∗ (0.016) 0.019∗∗ (0.009) 0.115∗∗∗ (0.039) 3. Baseline Controls + 1870 Shares Control + Additional Sorting Correlates 0.032∗∗∗ (0.008) 0.044∗∗∗ (0.012) 0.044∗∗ (0.017) 0.021∗∗ (0.010) 0.141∗∗∗ (0.038) 4. Post-LASSO from Baseline Controls (columns 3 and 6 of Tables 2 and 3) 0.032∗∗∗ (0.010) 0.042∗∗∗ (0.012) 0.029∗ (0.016) 0.018∗∗ (0.008) 0.122∗∗∗ (0.035) 5. Post-LASSO from All Controls 0.034∗∗∗ (0.010) 0.050∗∗∗ (0.013) 0.041∗∗ (0.018) 0.019∗∗ (0.009) 0.146∗∗∗ (0.038) 0.024∗∗∗ (0.008) 0.033∗∗∗ (0.011) 0.020 (0.014) 0.018∗∗ (0.007) 0.096∗∗∗ (0.035) 0.048∗ (0.029) 0.077∗∗∗ (0.026) 0.042∗ (0.024) 0.049∗∗ (0.025) 0.217∗∗∗ (0.081) 8. SSIV Using Actual Shifts 0.031∗∗∗ (0.009) 0.038∗∗∗ (0.012) 0.035∗∗ (0.016) 0.016∗ (0.008) 0.121∗∗∗ (0.036) 9. Restricting to Counties “Unsettled” in 1860 0.039∗∗∗ (0.013) 0.039∗∗ (0.018) 0.062∗∗ (0.031) 0.021 (0.015) 0.161∗∗ (0.067) 10. Excluding Oklahoma from Sending States 0.036∗∗∗ (0.010) 0.044∗∗∗ (0.014) 0.036∗∗ (0.017) 0.019∗ (0.010) 0.135∗∗∗ (0.041) 11. Including Missouri and Kentucky in Sending States 0.018∗∗ (0.008) 0.031∗∗∗ (0.008) 0.026∗∗ (0.013) 0.014∗∗ (0.007) 0.088∗∗∗ (0.025) 12. Excluding Border States from Receiving States 0.054∗∗∗ (0.018) 0.058∗∗ (0.026) 0.069∗ (0.036) 0.025 (0.017) 0.205∗∗ (0.088) 1. Baseline (columns 3 and 6 of Tables 2 and 3) Bester et al. (2011) 60 mi2 grid-cell Conley (1999) 200 km spatial HAC Conley (1999) 500 km spatial HAC Adao et al. (2019) SSIV adjustment Varying Control Sets Other Alternative Specifications 6. No Fixed Effects 7. Weighting by 1900 Population Alternative Diaspora Variation Notes: This table re-estimates Tables 2 and 3 using a variety of robustness specifications. See the notes to those tables for the list of baseline controls. Additional sorting controls include cotton, tobacco, and overall agricultural potential, Black shares in 1870, slave shares in 1860, Union Army enlistment and mortality rates, Breckinridge vote shares in 1860, and dummies for on the frontier in 1860 and never on the frontier, based on Bazzi et al. (2020). All continuous baseline and sorting controls are entered flexibly using quadratic terms. All regressions include state fixed effects, except for row 6. All rows instrument the share of Southern whites using a shift-share instrument based on the 1870 cross-sectional distribution of Southern whites and the predicted aggregate change in Southern white population living outside the South from 1870 to 1900. The latter is generated via a set of flexible LASSO regressions (see equation 1). SStandard errors are clustered at the 60×60 square-mile grid cell level following the approach of Bester et al. (2011), with the first row also reporting standard errors based on the Conley (1999) spatial HAC with wide bandwidths of 200 km and 500 km as well as the Adao et al. (2019) adjustment for SSIV estimators. Rows 4–5 choose optimal controls from these sets using the Belloni et al. (2014) double LASSO procedure. This procedure first runs a LASSO regression of the Southern white share on the set of controls. It then does the same using the given outcome. Lastly, it runs the IV regression using all of the controls that were selected in the first two steps. Row 7 weights regressions by county population in 1900. Row 8 uses a shift-share IV based on actual rather than predicted shifts. Row 9 uses only the subsample of 626 non-Southern counties classified by the Census Bureau as “unsettled” by non-natives as of 1860 (i.e., using the Census definition of < 2 persons per square mile). Rows 10–12 consider alternative sending and/or receiving state definitions for the endogenous and instrumental variables. Column 5 considers as an additional outcome the composite Confederate Culture Index, which sums the outcomes in columns 1–4. Significance levels are denoted by * p < 0.10, ** p < 0.05, *** p < 0.01. Table 5: Comparing Southern versus Northern White Migrant Impacts in Western States Dependent Variable: % Southern Whites, 1900 % Northern Whites, 1900 Estimator State FE Baseline controls 1870 share controls Observations Outcome mean KP Joint F-statistic KP Underident., p-val % S. Whites, SW F-statistic % S. Whites, Underid. p-val % N. Whites, SW F-statistic % N. Whites, Underid. p-val Any Confederate Memorials (1) Any UDC Chapters (2) Any KKK Chapters (3) Any Lynchings of Blacks (4) CCI Score (from 0–4) (5) 0.033∗∗∗ (0.009) 0.002 (0.005) 0.044∗∗∗ (0.012) 0.004 (0.004) 0.019∗ (0.011) 0.024∗∗∗ (0.006) 0.021∗∗ (0.009) -0.001 (0.004) 0.117∗∗∗ (0.032) 0.029∗∗ (0.012) IV Yes Yes Yes 823 0.29 12.16 0.00 27.06 0.00 27.89 0.00 IV Yes Yes Yes 823 0.17 12.16 0.00 27.06 0.00 27.89 0.00 IV Yes Yes Yes 823 0.31 12.16 0.00 27.06 0.00 27.89 0.00 IV Yes Yes Yes 823 0.08 12.16 0.00 27.06 0.00 27.89 0.00 IV Yes Yes Yes 823 0.86 12.16 0.00 27.06 0.00 27.89 0.00 Notes: This table makes two changes to the specification in columns 3 and 6 from Tables 2 and 3: (i) it adds the share of Northern white migrants as an additional endogenous variable, and (ii) it restricts the sample to Western states outside the North and South where we define the “North” as the territories of the Union during the Civil War, excluding the western states (California, Oregon, Nevada). The share of Southern and Northern whites are each instrumented using a shift-share instrument based on the 1870 cross-sectional distribution of whites of each group and the predicted change in the Southern (Northern) white population living in the West from 1870 to 1900. The shift part is generated via a set of flexible LASSO regressions (see equation 1). Baseline controls include log county population in 1870 and log county area (in square miles), entered flexibly using quadratic terms. Column 5 considers as an additional outcome the composite Confederate Culture Index, which sums the outcomes in columns 1–4. The KP Underidentification test p-value corresponds to the Kleibergen and Paap (2006) LM test whose null hypothesis is that the equation is underidentified. The SW F-statistics and Underidentification test p-values are based on Sanderson and Windmeijer (2016) first-stage F statistics and LM tests, respectively, for the individual endogenous regressors. Standard errors are clustered at the 60×60 square-mile grid cell level, following the approach of Bester et al. (2011). Significance levels are denoted by * p < 0.10, ** p < 0.05, *** p < 0.01. Table 6: The Confederate Diaspora, Sundown Towns, and the “Great Retreat” Dependent Variable: % Southern Whites, 1900 Number of Sundown Towns in County (1) (2) ∗ 0.102 (0.058) Yes Yes County County 1,701 0.47 15.0 0.00 (4) (5) (6) 1.096 (1.293) 2.355∗∗∗ (0.854) 1.501 (2.182) 2.603∗∗ (1.277) 0.599 (0.733) 2.060∗∗∗ (0.771) 0.868 (1.314) 2.329∗ (1.254) IV Yes Yes Yes IV Yes Yes Yes Yes Town Town 33,904 65.57 6.1 0.00 12.3 0.00 51.6 0.00 IV Yes Yes Yes IV Yes Yes Yes Yes Town County 33,904 65.57 6.9 0.00 14.5 0.00 53.5 0.00 0.106 (0.061) β≥25 : % Southern Whites, 1900 × ≥ 25 Blacks, 1870 IV Yes (3) ∗ β<25 : % Southern Whites, 1900 × < 25 Blacks, 1870 Estimator State FE Town controls Baseline controls 1870 share control Unit of analysis Diaspora regressor at . . . level Observations Outcome mean (KP Joint) F-statistic KP Underident., p-val β<25 , SW F-statistic β<25 , Underid. p -val β≥25 , SW F-statistic β≥25 , Underid. p -val No Blacks in Town After 1900 = 1 IV Yes Yes Yes Yes County County 1,701 0.47 12.1 0.00 Town Town 33,904 65.57 12.8 0.00 25.6 0.00 80.9 0.00 Town County 33,904 65.57 15.4 0.00 31.6 0.00 85.0 0.00 Notes: Columns 1–2 show SSIV regressions of the number of sundown towns in a county c on the share of Southern whites in 1900 in all non-Southern counties. Columns 3–6 alternatively show SSIV regressions of an indicator (×100) of whether a given non-Southern town had no Black residents at some point after 1900 (through 1940) on the share of Southern whites in 1900 in all non-Southern towns (columns 3–4) or counties (columns 5–6). Columns 3–6 include an interaction term for whether a town had over 25 Blacks in 1870 and report the coefficient estimates for those two subsamples. All columns instrument the share of Southern whites (and their interactions in columns 3–6) using a shift-share instrument based on the 1870 cross-sectional distribution across counties of Southern whites and the predicted change in Southern white (or Southern former slaveholder) population living outside the South from 1870 to 1900. The latter is generated via a set of flexible LASSO regressions (see equation 1). Baseline controls include log county population in 1870 and log county area (in square miles), entered flexibly using quadratic terms. Columns 3–6 also control for town longitude, latitude, a dummy for whether it had over 25 Blacks in 1870, and a dummy for whether it had over 1,000 residents in 1870, while column 2 controls for the county-level aggregates for these factors. All regressions include state fixed effects. The sample of confirmed sundown towns used to construct the numerator in columns 1–2 is originally from Loewen (2005) (who coined the term “Great Retreat” to capture the Black exodus from towns across America in the early 1900s) and taken from Taylor (2020) via its complementary GIS resource. The sample of towns used in columns 3–6 is based on Berkes et al. (2022). The KP Underidentification test p-value corresponds to the Kleibergen and Paap (2006) LM test whose null hypothesis is that the equation is underidentified. The SW F-statistics and Underidentification test p-values are based on Sanderson and Windmeijer (2016) first-stage F statistics and LM tests, respectively, for the individual endogenous regressors. Standard errors are clustered at the 60×60 square-mile grid cell level, following the approach of Bester et al. (2011). Significance levels are denoted by * p < 0.10, ** p < 0.05, *** p < 0.01. Table 7: Heterogeneous Effects of the Confederate Diaspora in Malleable and Oppositional Places Dependent Variable: Any Confederate Any UDC Any KKK Any Lynchings CCI Score Memorials Chapters Chapters of Blacks (from 0–4) (1) (2) (3) (4) (5) (a) Interacting with a County’s Frontier Status (as of 1860) β1 : % Southern Whites, 1900 β2 : % Southern Whites × On Frontier, 1860 β1 + β2 : % Southern Whites in Frontier Counties Estimator State FE Baseline controls 1870 share control Observations Outcome mean KP Joint F-statistic KP Underident., p-val β1 , SW F-statistic β1 , Underid. p-val β2 , SW F-statistic β2 , Underid. p-val 0.013 (0.009) 0.035∗∗ (0.014) 0.022∗ (0.012) 0.032 (0.021) 0.006 (0.012) 0.047∗ (0.026) 0.015∗ (0.008) 0.006 (0.015) 0.056∗∗ (0.027) 0.119∗ (0.064) 0.047∗∗∗ (0.013) 0.054∗∗∗ (0.020) 0.053∗∗ (0.026) 0.020 (0.014) 0.175∗∗∗ (0.064) IV Yes Yes Yes 1,701 0.25 4.0 0.02 65.6 0.00 91.1 0.00 IV Yes Yes Yes 1,701 0.09 4.0 0.02 65.6 0.00 91.1 0.00 IV Yes Yes Yes 1,701 0.37 4.0 0.02 65.6 0.00 91.1 0.00 IV Yes Yes Yes 1,701 0.05 4.0 0.02 65.6 0.00 91.1 0.00 IV Yes Yes Yes 1,701 0.76 4.0 0.02 65.6 0.00 91.1 0.00 (b) Interacting with a County’s Union Army Enlistment Rate β1 : % Southern Whites, 1900 β2 : % Southern Whites × % Union Enlistment β1 + β290th : % Southern Whites at 90th β1 + β290th : percentile of Union Enlistment Estimator State FE Baseline controls 1870 share control Observations Outcome mean KP Joint F-statistic KP Underident., p-val β1 , SW F-statistic β1 , Underid. p-val β2 , SW F-statistic β2 , Underid. p-val 0.041∗∗∗ (0.011) -0.045∗∗∗ (0.016) 0.046∗∗∗ (0.015) -0.034∗∗ (0.014) 0.046∗∗ (0.020) -0.062∗∗∗ (0.020) 0.014∗ (0.008) -0.006 (0.009) 0.148∗∗∗ (0.042) -0.147∗∗∗ (0.042) 0.001 (0.001) 0.023∗∗ (0.010) 0.003 (0.013) 0.010 (0.007) 0.044 (0.027) IV Yes Yes Yes 1,701 0.25 14.5 0.00 18.3 0.00 49.1 0.00 IV Yes Yes Yes 1,701 0.09 14.5 0.00 18.3 0.00 49.1 0.00 IV Yes Yes Yes 1,701 0.37 14.5 0.00 18.3 0.00 49.1 0.00 IV Yes Yes Yes 1,701 0.05 14.5 0.00 18.3 0.00 49.1 0.00 IV Yes Yes Yes 1,701 0.76 14.5 0.00 18.3 0.00 49.1 0.00 Notes: This table augments the specification in columns 3 and 6 from Tables 2 and 3 to include interactions of the Southern white share with pre-migration characteristics of the destination counties: an indicator for frontier status in 1860 from Bazzi et al. (2020) (panel a), and Union Army enlistment rates during the Civil War (panel b). The own-term in the interaction is also included as an additional regressor in both stages. All columns instrument the share of Southern whites using a shift-share instrument based on the 1870 cross-sectional distribution of Southern whites and the predicted change in the Southern white population living outside the South from 1870 to 1900. The latter is generated via a set of flexible LASSO regressions (see equation (1)). The interaction terms are instrumented separately by the interaction of the SSIV and the given variable. The specifications are otherwise identical to those in Tables 2 and 3; see the notes therein for additional details. Column 5 considers as an additional outcome the composite Confederate Culture Index, which sums the outcomes in columns 1–4. In panel (b), the first coefficient, β1 , captures the effect in incorporated counties with zero Union enlistment rate, and the bottom row coefficient reports the effect evaluated at the 90th percentile of Union enlistment rates, which is 70.2%. We include a separate interaction and own-term indicator for counties, mostly unincorporated, with missing Union enlistment rate; the interaction term is instrumented by the SSIV times the missing indicator. The KP Underidentification test p-value corresponds to the Kleibergen and Paap (2006) LM test whose null hypothesis is that the equation is underidentified. The SW F-statistics and Underidentification test p-values are based on Sanderson and Windmeijer (2016) firststage F statistics and LM tests, respectively, for the individual endogenous regressors. Standard errors are clustered at the 60×60 square-mile grid cell level, following the approach of Bester et al. (2011). Significance levels are denoted by * p < 0.10, ** p < 0.05, *** p < 0.01. Table 8: The Distinctive Influence of Former Slaveholders in the Confederate Diaspora (1) (2) (3) (4) (5) (6) (7) (a) Dep. Var.: All Confederate Cultural Activity (CCI Score, from 0–4) % Southern Whites, 1900 0.130∗∗∗ (0.039) 1.055∗∗∗ (0.402) % Southern Former Slaveholders, 1900 1.140∗∗∗ (0.402) -0.042 (0.063) 2.694∗∗ (1.099) -0.036 (0.064) 2.609∗∗ (1.120) -0.028 (0.047) 2.497∗∗∗ (0.931) -0.010 (0.063) 2.386∗∗ (1.122) (b) Dep. Var.: Any Confederate Memorials % Southern Whites, 1900 ∗∗∗ 0.034 (0.009) 0.291∗∗∗ (0.108) % Southern Former Slaveholders, 1900 0.318∗∗∗ (0.106) -0.015 (0.021) 0.769∗∗ (0.345) -0.013 (0.021) 0.739∗∗ (0.351) -0.010 (0.016) 0.698∗∗ (0.279) -0.005 (0.021) 0.574∗ (0.321) (c) Dep. Var.: Any United Daughters of the Confederacy Chapters % Southern Whites, 1900 ∗∗∗ 0.042 (0.013) 0.250∗ (0.131) % Southern Former Slaveholders, 1900 0.304∗∗ (0.130) -0.013 (0.020) 0.869∗∗ (0.350) -0.008 (0.019) 0.785∗∗ (0.338) -0.006 (0.014) 0.762∗∗∗ (0.291) 0.004 (0.018) 0.631∗∗ (0.311) (d) Dep. Var.: Any Ku Klux Klan Chapters % Southern Whites, 1900 ∗∗ 0.035 (0.016) 0.304∗ (0.181) % Southern Former Slaveholders, 1900 0.284 (0.177) 0.002 (0.023) 0.557 (0.395) -0.003 (0.025) 0.624 (0.427) -0.002 (0.019) 0.614 (0.373) 0.002 (0.028) 0.687 (0.488) (e) Dep. Var.: Any Lynchings of Blacks % Southern Whites, 1900 ∗∗ 0.018 (0.009) -0.012 (0.015) 0.461∗ (0.246) -0.010 (0.011) 0.423∗∗ (0.204) -0.011 (0.016) 0.494∗ (0.279) IV Yes Yes Yes Yes 0.209∗∗ (0.101) 0.233∗∗ (0.100) IV Yes Yes Yes IV Yes Yes Yes IV Yes Yes Yes Yes IV Yes Yes Yes IV Yes Yes Yes Yes IV Yes Yes Yes Yes Yes 1,701 15.0 0.00 1,679 19.8 0.00 1,679 20.2 0.00 1,679 2.5 0.01 7.1 0.01 5.1 0.02 1,679 2.4 0.02 6.5 0.01 4.8 0.03 1,679 3.9 0.00 6.5 0.01 4.8 0.03 % Southern Former Slaveholders, 1900 Estimator State FE Baseline controls 1870 share control(s) % Non-Southern slaveholders control Linked Census measures only? Sorting controls Observations KP Joint F-statistic KP Joint Underident., p-val % S. Whites, SW F-statistic % S. Whites, Underident., p-val % Slaveholders, SW F-statistic % Slaveholders, Underident., p-val -0.015 (0.015) 0.499∗∗ (0.240) Yes 1,677 2.6 0.01 6.8 0.01 5.2 0.02 Notes: This table re-estimates Tables 2 and 3 using an alternative explanatory variable, based on migrating Southern white former slaveholders as of 1900, as well as alternative shift-share instrument, based on the 1870 cross-sectional distribution of Southern whites and the predicted aggregate change in the Southern former slaveholder population living outside the South from 1870 to 1900. The latter is generated via a set of flexible LASSO regressions (see equation 1). Baseline controls include log county population in 1870 and log county area (in square miles). Columns 3, 5, and 6 also control for the share of non-Southern former slaveholders in a county, defined as a share of all individual linked to non-Southern states as of the 1860 Slave Schedule. All regressions control for the share of Southern whites and/or the share of Southern former slaveholders in 1870 as well as state fixed effects. Column 6 uses an alternative endogenous variable for Southern whites in 1900 based on the same linked Census dataset as the slaveholder measure. Column 7 includes additional sorting controls for cotton, tobacco, and overall agricultural potential, Black shares in 1870, slave shares in 1860, Union Army enlistment and mortality rates, Breckinridge vote shares in 1860, and dummies for on the frontier in 1860 and never on the frontier, based on Bazzi et al. (2020). All continuous baseline and sorting controls are entered flexibly using quadratic terms. The KP Underidentification test p-value corresponds to the Kleibergen and Paap (2006) LM test whose null hypothesis is that the equation is underidentified. The SW F-statistics and Underidentification test p-values are based on Sanderson and Windmeijer (2016) first-stage F statistics and LM tests, respectively, for the individual endogenous regressors. Standard errors are clustered at the 60×60 square-mile grid cell level, following the approach of Bester et al. (2011). Significance levels are denoted by * p < 0.10, ** p < 0.05, *** p < 0.01. Table 9: Institutional Foundations of Confederate Diaspora Influence (1) (2) (3) (4) (5) (6) Any Confederate Memorials (7) -0.048 (0.064) 2.731∗∗ (1.105) -0.042 (0.063) 2.705∗∗ (1.102) -0.041 (0.062) 2.675∗∗ (1.092) -0.042 (0.062) 2.648∗∗ (1.116) -0.041 (0.063) 2.729∗∗ (1.160) -0.047 (0.062) 2.724∗∗ (1.250) -0.014 (0.021) 0.764∗∗ (1.357) -0.015 (0.020) 0.872∗∗ (0.409) -0.002 (0.022) 0.578 (0.422) -0.016 (0.015) 0.510∗ (0.470) -0.160 (0.341) 0.727∗∗∗ (0.237) 1.141∗∗∗ (0.182) 0.270 (0.354) 0.227 (0.214) -0.258 (0.345) 0.060 (0.130) 0.049 (0.055) 0.248∗ (0.127) 0.026 (0.076) -0.097 (0.113) 0.223∗∗ (0.094) 0.043 (0.061) 0.088 (0.134) 0.087 (0.075) -0.082 (0.112) 0.340∗∗∗ (0.109) 1.010∗∗∗ (0.066) -0.069 (0.113) 0.073 (0.075) -0.014 (0.119) 0.103 (0.111) 0.039 (0.036) 0.002 (0.092) 0.041 (0.056) -0.065 (0.077) IV Yes Yes Yes 1,679 0.77 2.4 0.02 IV Yes Yes Yes 1,679 0.77 2.3 0.02 IV Yes Yes Yes 1,679 0.25 2.3 0.02 IV Yes Yes Yes 1,679 0.09 2.3 0.02 IV Yes Yes Yes 1,679 0.38 2.3 0.02 IV Yes Yes Yes 1,679 0.05 2.3 0.02 Dependent Variable: % Southern Whites, 1900 % Southern Former Slaveholders, 1900 Confederate Cultural Activity (CCI Score, from 0–4) Any UDC Chapters (8) Any KKK Chapters (9) Any Lynchings of Blacks (10) Any Linked . . . from Southern Slaveholder Households? Public Administrators 0.735∗∗∗ (0.227) 1.140∗∗∗ (0.182) Religious Workers Law Enforcement 0.192 (0.329) Lawyers or Judges 0.222 (0.212) Educators Estimator State FE Baseline controls 1870 share controls Observations Outcome mean KP Joint F-statistic KP Underident., p-val IV Yes Yes Yes 1,679 0.77 2.5 0.01 IV Yes Yes Yes 1,679 0.77 2.5 0.01 IV Yes Yes Yes 1,679 0.77 2.5 0.01 IV Yes Yes Yes 1,679 0.77 2.5 0.02 Notes: This table re-estimates Table 8 alongside county-level indicators of whether any Southern white (male) migrants linked to former slaveholder households (as of 1860) work in a given occupation in that non-Southern county (as of 1900). See the notes to Table 8 for all other details on instruments and controls. Southern counties are those belonging to states of the former Confederacy and Oklahoma. Standard errors are clustered at the 60×60 square-mile grid cell level, following the approach of Bester et al. (2011). Significance levels are denoted by * p < 0.10, ** p < 0.05, *** p < 0.01. Table 10: Black Incarceration in the Southern (Former) Slaveholder Diaspora (1) Black ∗∗∗ 0.289 (0.029) Dependent Variable: Person is Incarcerated in 1920 (2) (3) (4) (5) ∗∗∗ 0.291 (0.030) ∗∗∗ 0.284 (0.028) ∗∗∗ 0.283 (0.028) ∗∗∗ (6) 0.331 (0.042) 0.334∗∗∗ (0.045) 0.005 (0.011) 0.379∗∗∗ (0.099) 0.372∗∗∗ (0.102) Any . . . from Southern Slaveholder Households in 1900 County of Residence? Law Enforcement Law Enforcement × Black 0.002 (0.005) 0.339∗∗ (0.143) 0.335∗∗ (0.140) Lawyers or Judges 0.000 (0.003) 0.199∗ (0.113) Lawyers or Judges × Black 0.210∗ (0.113) Public Administrators Public Administrators × Black Estimator Initial state FE Initial county FE Observations Outcome mean (whites) OLS Yes 4,228,050 0.06 OLS Yes 4,228,050 0.06 OLS Yes 4,228,050 0.06 OLS Yes 4,228,050 0.06 OLS Yes 4,228,050 0.06 OLS Yes 4,228,050 0.06 Notes: OLS regressions of a binary indicator of incarceration status (×100) among Black and white men between the ages of 20 and 45 in the 1920 U.S. Census on an individual being Black, interacted with an indicator for whether they lived in a non-Southern county in 1900 with any Southern white (male) migrants linked to former slaveholder households (as of 1860) working in a given governance occupation in that non-Southern county (as of 1900). Excluded Southern counties are those belonging to states of the former Confederacy and Oklahoma. We also exclude white men born in any of these states. Individual controls include a cubic in age, marital status, and number of children. Odd columns include fixed effects for state of residence in 1900, while even columns include fixed effects for county of residence in 1900. Note that these county FE absorb the effects of Southern white and slaveholder migrant population shares in the individuals’ counties as of 1900. Standard errors are clustered by initial county. Significance levels are denoted by * p < 0.10, ** p < 0.05, *** p < 0.01. Table 11: The Confederate Diaspora and Klan Membership in the Early 20th Century Dependent Variable: Matched to KKK Member Records (2) (3) (4) (5) (1) (6) (a) All White Men Southern-Born Second-Generation Non-Southern-Born w/ Southern Parent ∗∗∗ 0.032 (0.007) 0.032 (0.007) 0.037∗∗∗ (0.007) 0.029∗∗∗ (0.005) Non-Southern-Born w/ Southern Mother – 129,248 0.24 0.029∗∗∗ (0.005) 0.035∗∗∗ (0.005) 0.031∗∗∗ (0.009) 0.021∗∗ (0.009) Non-Southern-Born w/ Southern Father Fixed Effects Observations Outcome mean (control) ∗∗∗ – 129,248 0.24 0.024∗∗∗ (0.004) 0.026∗∗∗ (0.004) 0.033∗∗∗ (0.006) 0.019∗∗∗ (0.006) County 241,298 0.21 0.024∗∗∗ (0.005) County 241,298 0.21 0.026∗∗∗ (0.006) 0.014∗∗ (0.006) District 241,297 0.21 District 241,297 0.21 (b) Only White Men with no Southern Heritage/Parentage Neighbors in Houses Next Door 1st or 2nd Gen. Southern White Neighbor 0.019∗∗∗ (0.006) 0.018∗∗∗ (0.004) 0.018∗∗ (0.007) 0.021∗∗∗ (0.008) 1st Gen. Southern White Neighbor 2nd Gen. Southern White Neighbor 0.008∗∗ (0.003) 0.015∗∗∗ (0.005) 0.021∗∗∗ (0.005) Neighbors in Enumeration District % 1st Gen. Southern Whites in District % 2nd Gen. Southern Whites in District Fixed Effects Observations Outcome mean (control) Estimator Sample Counties 0.006 (0.004) 0.010∗∗ (0.004) 0.005∗∗∗ (0.001) 0.008∗∗∗ (0.001) 0.005∗∗∗ (0.001) 0.008∗∗∗ (0.001) – 93,654 0.25 – 93,654 0.25 County 189,263 0.22 County 189,263 0.22 County 189,263 0.22 County 189,263 0.22 OLS Denver only OLS Denver only OLS All metro OLS All metro OLS All metro OLS All metro Notes: The dependent variable is a binary indicator for whether a white male in the Denver, CO metropolitan area as of the 1920 U.S. Census can be found in Denver KKK membership records from the 1920s. Linking based on first and last names, using the ABE algorithm from Abramitzky et al. (2019) with NYSIIS standardized names. In panel (a), the sample includes all white men in the given area, and the regressors include indicators for whether they were born in the South and whether they were born outside the South but their parents were born in the South. In panel (b), the sample includes only those born outside the South to parents who were also born outside the South, and the regressors include an indicator for whether one’s next-door neighbors have first- and/or second-generation Southern white migrants and the share of first- and/or- second-generation Southern white migrants in one’s enumeration district. The latter is leave-out, excluding one’s next-door neighbors from the district total. There are 313 enumeration districts in Denver county and 527 in the greater metro area, which spans 14 counties. All regressions include controls for a cubic polynomial in age, marital status, and number of children. Standard errors are clustered by enumeration district. Significance levels are denoted by * p < 0.10, ** p < 0.05, *** p < 0.01. Table 12: Occupational Sorting by 2nd Generation Southern White Migrants, 1900 Dependent Variable: Individual Works as. . . Any Authority Position (1) (2) Lawyer or Judge (3) Law Enforce. (4) Public Admin. (5) Religious Worker (6) Educator (7) (a) Full Sample (N = 16, 187, 176) Southern-Born (First-Generation) Non-Southern-Born w/ Southern Parent (Second-Generation) Non-Southern mean 0.669∗∗∗ (0.065) 0.471∗∗∗ (0.040) 0.451∗∗∗ (0.042) 0.397∗∗∗ (0.023) 0.307∗∗∗ (0.055) 0.241∗∗∗ (0.030) 0.009 (0.020) 0.034∗∗∗ (0.007) -0.007 (0.007) 0.023∗∗∗ (0.005) 0.407∗∗∗ (0.040) 0.080∗∗∗ (0.012) -0.046∗∗∗ (0.017) 0.093∗∗∗ (0.015) 1.58 1.58 0.39 0.26 0.09 0.36 0.49 (b) Out-of-State Migrants Only (N = 8, 207, 851) First-Generation Second Generation Non-Southern mean 0.785∗∗∗ (0.082) 0.773∗∗∗ (0.059) 0.592∗∗∗ (0.049) 0.544∗∗∗ (0.031) 0.417∗∗∗ (0.065) 0.363∗∗∗ (0.044) 0.072∗∗∗ (0.017) 0.054∗∗∗ (0.011) 0.016∗∗ (0.007) 0.034∗∗∗ (0.008) 0.254∗∗∗ (0.040) 0.196∗∗∗ (0.021) 0.026 (0.019) 0.125∗∗∗ (0.018) 1.47 1.47 0.33 0.24 0.08 0.48 0.35 (c) Western Counties Only (N = 1, 348, 860) First-Generation Second-Generation Non-Southern mean 0.865∗∗∗ (0.099) 0.932∗∗∗ (0.098) 0.541∗∗∗ (0.060) 0.565∗∗∗ (0.051) 0.496∗∗∗ (0.068) 0.499∗∗∗ (0.069) 0.143∗∗∗ (0.024) 0.121∗∗∗ (0.025) 0.005 (0.018) 0.040∗∗ (0.016) 0.188∗∗∗ (0.042) 0.074∗∗∗ (0.026) 0.033 (0.035) 0.199∗∗∗ (0.034) 1.71 1.71 0.54 0.22 0.14 0.35 0.47 (d) Non-Western Counties Only (N = 14, 838, 316) 0.619∗∗∗ (0.077) 0.406∗∗∗ (0.043) 0.426∗∗∗ (0.049) 0.370∗∗∗ (0.025) 0.260∗∗∗ (0.067) 0.204∗∗∗ (0.033) -0.023 (0.024) 0.020∗∗∗ (0.007) -0.010 (0.007) 0.020∗∗∗ (0.005) 0.456∗∗∗ (0.049) 0.083∗∗∗ (0.013) -0.065∗∗∗ (0.020) 0.079∗∗∗ (0.017) Non-Southern mean 1.57 1.57 0.37 0.26 0.08 0.36 0.50 County FE Occup. Income Category FE Yes No Yes Yes Yes – Yes – Yes – Yes – Yes – First-Generation Second-Generation Notes: This table re-estimates the specifications in Table 1 with an additional binary regressor indicating those individuals in the secondgeneration Confederate diaspora, i.e., men with at least one parent born in the South but who are themselves born outside the South. The dependent variables are binary indicators (×100) for the given occupational choice. See the notes to Table 1 for details on the specification. Standard errors are clustered by county. Significance levels are denoted by * p < 0.10, ** p < 0.05, *** p < 0.01. Table 13: An Enduring Legacy of the Confederate Diaspora Economic Dependent Variable: % Southern Whites, 1900 Estimator State FE Baseline controls Sorting controls 1870 share control Observations Outcome mean F-statistic Anderson-Rubin, p-val KP Underident., p-val Residential Black to Non-Black Earnings Ratio, 1940 (1) (2) Black Residential Segregation, 1940 (3) (4) Penal Black Residential Segregation, 2000 (5) (6) Black County Jail Shares, 2010–18 (7) (8) Any Police Killings of Blacks, 2013–20 (9) (10) -0.039∗∗∗ (0.014) -0.030∗∗ (0.012) 0.007∗∗ (0.003) 0.006∗ (0.003) 0.009∗∗∗ (0.003) 0.008∗∗∗ (0.002) 0.757∗∗ (0.313) 0.735∗∗∗ (0.276) 0.034∗∗ (0.013) 0.037∗∗∗ (0.013) IV Yes Yes IV Yes Yes Yes Yes 1,043 0.77 14.8 0.00 0.00 IV Yes Yes IV Yes Yes Yes Yes 1,693 0.08 16.3 0.05 0.00 IV Yes Yes IV Yes Yes Yes Yes 1,695 0.07 15.3 0.00 0.00 IV Yes Yes IV Yes Yes Yes Yes 1,603 13.78 15.6 0.01 0.00 IV Yes Yes IV Yes Yes Yes Yes 1,699 0.14 15.3 0.00 0.00 Yes 1,045 0.77 12.0 0.00 0.00 Yes 1,695 0.08 15.8 0.02 0.00 Yes 1,697 0.07 14.9 0.00 0.00 Yes 1,604 13.78 14.9 0.01 0.00 Yes 1,701 0.14 15.0 0.00 0.00 Notes: SSIV regressions of various Black-related socioeconomic outcomes between 1940 and 2020 on the share of Southern whites in 1900 in non-Southern counties. Excluded Southern counties are those belonging to states of the former Confederacy and Oklahoma. All columns instrument the share of Southern whites using a shift-share instrument based on the 1870 cross-sectional distribution of Southern whites and the predicted change in the Southern white population living outside the South from 1870 to 1900. The latter is generated via a set of flexible LASSO regressions (see equation 1). Baseline controls include log county population in 1870 and log county area (in square miles). Additional sorting controls include cotton, tobacco, and overall agricultural potential, Black shares in 1870, slave shares in 1860, Union Army enlistment and mortality rates, Breckinridge vote shares in 1860, and dummies for on the frontier in 1860 and never on the frontier, based on Bazzi et al. (2020). All continuous baseline and sorting controls are entered flexibly using quadratic terms. All regressions control for the share of Southern whites in 1870 and include state fixed effects. The Anderson-Rubin p-value corresponds to the null hypothesis that the coefficient on % Southern W hites, 1900 is zero and that the overidentifying restrictions are valid. The KP Underidentification test p-value corresponds to the Kleibergen and Paap (2006) LM test whose null hypothesis is that the equation is underidentified. Standard errors are clustered at the 60×60 square-mile grid cell level, following the approach of Bester et al. (2011). Significance levels are denoted by * p < 0.10, ** p < 0.05, *** p < 0.01.